Aktuelles

Financial Stability Report 48

Recent developments and macroprudential policy update

Austrian economy experiences second year of recession in 2024

Austria’s economy has been in recession almost continuously for two years. Economic output contracted by a total of 2.1% from its peak in the second quarter of 2022 to the second quarter of 2024. This downturn has been primarily driven by two factors: an industrial recession and a notable decline in consumer spending. The industrial sector has been particularly affected by the global economic slowdown, with the downturn in Germany – Austria’s key trading partner – having a significant impact on Austrian industry. In addition to weak foreign demand, domestic demand has underperformed across various sectors. The OeNB’s September 2024 Interim Economic Outlook highlights that energy-intensive and construction-related industries have been the main drivers of the industrial recession. Despite strong income growth, consumer spending has fallen short of expectations due to persistently low consumer confidence, which has led to a sharp increase in the saving rate. In light of the OeNB’s revised outlook for the second half of the year, the forecast for real GDP growth in 2024 has been downgraded by one percentage point to –0.7%, and by 0.8 percentage points to 1.0% for 2025. As a result of the weaker economic activity, the unemployment rate is projected to rise to 7.1% in 2024 and 7.5% in 2025.

The inflation shock is coming to an end, and HICP inflation is to fall below 3% in 2024. HICP inflation peaked at 11.6% in January 2023 and has since steadily declined, reaching 1.8% in September – a level last seen in early 2021. The drop in HICP inflation from 2023 to 2024 has been driven by all major components of the index, particularly industrial goods (excluding energy), energy and food. According to the OeNB’s latest forecasts, the annual average HICP inflation rate is expected to fall from 7.7% in 2023 to 2.9% in 2024. However, disinflation is anticipated to slow in subsequent years due to the expiration of fiscal measures in the energy sector. The OeNB projects HICP inflation to be 2.3% in 2025 and 2.2% in 2026.

Table 1: OeNB September 2024 outlook for Austria – main results  
2023 2024 2025 2026
Annual change in % (real)
Gross domestic product (GDP) –0.7 –0.7 1.0 1.5
Harmonised Index of Consumer Prices (HICP) 7.7 2.9 2.3 2.2
HICP excluding energy 7.8 3.8 2.6 2.2
%
Unemployment rate (national definition) 6.4 7.1 7.5 7.3
Source: 2023: Statistics Austria; 2024 to 2026: OeNB September 2024 outlook.

Austria’s budget deficit will be higher than 3%, which is why the country is likely to face an excessive deficit procedure. The ongoing recession, coupled with declining inflation, is causing a further worsening of public finances. Without additional corrective measures, the budget deficit will exceed the 3% target in 2024 and in the coming years. This increases the likelihood of the European Commission initiating an excessive deficit procedure against Austria.

The ECB lowered its deposit rate to 3.25% in October 2024. Inflation in the euro area has fallen more quickly than anticipated at the start of the year. According to the ECB’s September forecast, inflation is projected to reach 2.2% in 2025 and 1.9% in 2026. In response, the ECB Governing Council decided to lower interest rates in June, September and October by 25 basis points each, marking the start of an easing cycle.

The Austrian banking sector’s profitability and capitalization remain strong, but risks from commercial real estate loans intensify

The consolidation of the Austrian banking sector continued in the first half of 2024, while the sector’s total assets registered moderate growth. The number of banks in Austria fell only marginally, as mergers typically take place in the second half of a year. However, total assets continued to grow moderately, reaching EUR 1,243 billion at the end of June 2024. On the asset side, loans and debt securities contributed to the increase, while banks slightly reduced their cash balances. On the liability side, growth was driven by both interbank and customer deposits.

Despite ongoing quantitative tightening, the liquidity position of the Austrian banking sector has improved and remained comfortable. Recent trends in Austrian banks’ funding and liquidity positions have continued in the first half of 2024. Banks have responded to the continued reduction of excess reserves in the Eurosystem (quantitative tightening) by further substituting their cash and reserve holdings with government and covered bond holdings. The banking sector managed to keep its overall liquidity risk metrics fairly constant: The liquidity coverage ratio, for instance, which measures the amount of high-quality liquid assets – like reserves or government bonds – that banks hold against expected short-term outflows in a liquidity stress scenario, stood at 174% on a consolidated level as of mid-2024, up 1.1 percentage points from end-2023 and up 9.2 percentage points year on year. On the liability side, the relative importance of sight deposits, which markedly decreased as interest rate hikes started in mid-2022, has stabilized in the first half of 2024. This could be a consequence of the narrowing interest rate differential between term and sight deposits due to recent interest rate cuts.

This is Chart 1.

The Austrian banking system earned a EUR 7.0 billion profit in the first half of 2024, which is only slightly below the record set in the same period of 2023. Net interest income, which makes up around two-thirds of all income, was up by 4%, while fees and commissions recorded only a marginal decline. Consequently, operating income increased to EUR 18.8 billion in the first half of 2024. As operating costs rose in line with income – with noticeable reductions in contributions to resolution and deposit guarantee funds but higher impairments on participations – the operating profit stayed flat at EUR 9.1 billion. The cost-to-income ratio was almost unchanged at 51%. As risk provisioning rose by 12% to EUR 0.9 billion, and tax payments increased by 13% to EUR 1.7 billion, the profit of the Austrian banking system fell slightly to EUR 7.0 billion. This translated into a return on assets of 1.2%, some 8 basis points lower than in the first half of 2023, but still highlighting banks’ strong profitability.

The Austrian banking sector’s capitalization further increased in the first half of 2024, and large Austrian banking groups improved their capitalization compared to their competitors in the SSM. With the tailwind of high profitability, Austrian banks further increased their capitalization in the first half of 2024. Although risk-weighted assets grew by EUR 14 billion, the retention of profits raised the common equity tier 1 (CET1) ratio by a further 10 basis points to 17.7%. The leverage ratio – which is not risk-weighted – improved to 8.5%. Compared to the EU banking sector, the Austrian banking sector is well capitalized with a CET1 ratio 130 basis points above the average. Besides, the CET1 ratio of large Austrian banking groups in the SSM increased to 16.3%, some 50 basis points above the average of their SSM peers.

As for macroprudential capital requirements, the buffer for other systemically important institutions (O-SII) addresses risks that large systemic banks pose to the financial system. 1 The buffer requires systemic banks in Austria to hold additional CET1 capital and thereby lowers their probability of failure. On an annual basis, the OeNB identifies banks that are systemically important and thus have to hold more capital. In October 2024, Austria’s Financial Market Stability Board (FMSB) recommended removing a temporary limit on O-SII buffer rates that had been in place since 2022 due to uncertainty over Russia’s war against Ukraine, increased energy prices and high inflation. As a result of this final phase-in step, four large banks will see their buffer requirements increase slightly.

In order to create a stronger link between systemic importance and buffer rates, the FMSB recommended increasing the number of buckets in the O-SII assessment methodology. 2 Banks that are systemically important but at the lower end of the distribution will be required to hold a buffer of 0.45% (after taking the overlap with the systemic risk buffer (SyRB) into account). Currently, two banks fall into this category and will see a slight reduction in their O-SII buffers. At the other end of the spectrum, banks exceeding a high threshold will need to hold a buffer of 2.2% (after taking the overlap with the SyRB into account). This bucket is deliberately left empty and ensures that if the largest banks further increase their systemic importance, they do so with a commensurate increase in resilience.

Table 2: Allocation of scores to buffer levels  
Previously New
Bucket O-SII buffer O-SII buffer Scores Bucket O-SII buffer O-SII buffer Scores
(pre-overlap) (post-overlap) (pre-overlap) (post-overlap)
% of CET1 capital % of CET1 capital
Bucket 1 0.5 0.45 Only additional
indicators and <275
Bucket 1 1.0 0.90 275–636 Bucket 2 1.0 0.90 275–636
Bucket 2 1.5 1.30 637–999 Bucket 3 1.5 1.30 637–999
Bucket 3 2.0 1.75 ≥1,000 Bucket 4 2.0 1.75 1,000–3,399
Bucket 5 2.5 2.20 ≥3,400
Source: OeNB.
Note: Changes in bold. “Pre-overlap” means prior to adjusting for overlap with systemic risk buffer; “post-overlap” means after adjustment.

Results of the OeNB’s 2024 solvency stress test

Background

The OeNB conducts annual stress tests for all Austrian banks under its dual mandate for banking supervision and financial stability. The solvency stress test is designed to assess banks’ resilience to adverse macroeconomic shocks and provides insights on both bank-specific and system-wide vulnerabilities. Conducted in a top-down fashion, it relies on the OeNB’s ARNIE stress testing framework, which is well-established and continuously improved. The stress test covers both significant and less significant institutions at the highest consolidated level. It focuses on risks faced by the Austrian banking sector, including spillover effects among banks, which are particularly relevant to the decentralized sector. The most recent stress test is based on data as of year-end 2023 and covers the period from 2024 to 2026.

Scenario

The adverse scenario assumes a severe macroeconomic downturn combined with a decline in inflation and interest rates. The baseline scenario projects a cumulative GDP growth of 3.6% for the Austrian economy over the stress test horizon (2024–26). In the adverse scenario, characterized by spillover effects of a credit bust in China, euro area GDP contracts by a cumulative 6.0% (–5.0% in Austria). Further escalation of Russia’s war against Ukraine leads to an idiosyncratic shock on energy prices in Austria, Hungary and Slovakia, triggering prolonged inflationary effects in these countries reliant on Russian gas. Austrian inflation falls from 7.7% in 2023 to 3.1% in 2026, while euro area inflation declines to 1.9% in 2026, leading to interest rate cuts. Short-term rates (3M EUR Swap) fall from 3.4% in 2023 to 2.1% in 2026, sharper and further than in the baseline scenario (2.7% in 2026).

Results and risk drivers

While the aggregate CET1 ratio increases by 1.4 percentage points in the baseline scenario, it declines by 5.4 percentage points in the adverse scenario, landing at 12.2% at year-end 2026. The following waterfall charts show the most important risk drivers and their contribution to changes in the capital ratio for both the baseline and the adverse scenario.

This is Chart 2.

Credit risk remains the main risk driver and reduces capital by 6.4 percentage points in the adverse scenario (baseline: –1.8 percentage points). Approximately one-tenth of these credit risk losses is attributable to Austrian commercial real estate (CRE) exposures, which were specifically subjected to a shock in this year’s exercise. Moreover, the contribution of net interest income drops from 11.2 percentage points in the baseline to 9.8 percentage points in the adverse scenario. Net interest margins face a double compression: With falling interest rates, banks’ loan book income declines, while interest paid on deposits does not. During the recent period of rising interest rates, banks could keep deposit rates at comparably low levels, so that current rates are still below those assumed to materialize in the adverse scenario. Interest expenses are therefore modeled to increase slightly, and net interest margins return to levels seen around 2020. Finally, gains and losses from equity participations remain significant. In the baseline scenario, banks participate in the profits of entities they are invested in and build up capital (+0.6 percentage points), but the picture reverses in the adverse scenario (–0.7 percentage points), reflecting reduced dividend income and write-downs of equity participations. On aggregate, the difference between gains and losses in the baseline and in the adverse scenario is less pronounced than in previous years due to methodological improvements, while for some banks the impact is now more material.

Compared to last year’s exercise, the 2024 OeNB stress test projects a greater impact (–5.4 vs. –4.2 percentage points in 2023). This reflects higher credit risk (–6.4 percentage points vs. –5.6 percentage points in 2023), with exposures linked to CRE being especially hard hit, and lower net interest income (9.8 percentage points vs. 10.3 percentage points in 2023). At the same time, record profits allowed Austrian banks to increase capital by 1.2 percentage points in 2023. Therefore, the final CET1 ratio after stress remains practically unchanged from last year’s stress test.

Conclusions

Overall, the stress test indicates that the Austrian banking system is well placed to withstand substantial macroeconomic shocks. Banks were able to build up capital and now have a greater cushion against potential losses. However, results are heterogeneous across the Austrian banking sector. With credit risk costs rising across the board, banks with a larger share of CRE exposures relative to their total loans are especially vulnerable to higher losses. Falling interest rates will compress interest margins, adding to downward pressure on capital.

The stress test underlines the importance of a well-capitalized banking sector. Even though capital ratios have increased significantly in recent years, overall uncertainty remains high. Given the speed of recent interest rate movements and rising credit risk costs, banks might face substantial headwinds in the years to come. Therefore, it is important that Austrian banks be forward looking and prudent with profit distributions. These conclusions are confirmed by a special topic in this report entitled “Results of the first dynamic balance sheet stress test in the ARNIE framework.” It simulates banks’ reactions to the same macroeconomic scenarios used in this stress test and finds that better capitalized banks are able to grow even in the adverse scenario, providing credit to the real economy in times of crisis.

While mortgage lending seems to have bottomed out, corporate lending growth has been slowing down. Demand for corporate loans has been falling since 2022, with a persistent weakness particularly in the demand for long-term loans to finance investments. This subdued demand and the restrictive lending policies of banks mean that corporate investment activity has not been contributing to economic growth in Austria, which is reflected in the current weak economic outlook. In contrast, housing loans have seen a moderate recovery in demand since the first quarter of 2024, starting from a historic low after sharp declines in the previous one and a half years. The moderate rebound is due to improvements in affordability driven by rising real incomes and slightly falling financing costs. That said, the annual growth rate for domestic loans remains low. As of August 2024, loans to Austrian companies grew by 0.7% and loans to households contracted by 1.4% year on year. However, the bottom for the latter seems to have been reached.

This is Chart 3.

The share of lending at variable interest rates continued to decline. In the first half of 2024, new borrowers in Austria sought low interest rate risk. Consequently, the share of variable rate loans in total new loans continued to decline to around 40% for households and three-quarters for companies. The decline was especially pronounced in mortgage lending, where only one in five new loans had a variable interest rate. However, the overall share of variable rate loans in Austria is still above the European average. Supervisors therefore continue to closely monitor further developments.

This is Chart 4.

The deterioration of credit quality continued in the first half of 2024, albeit at a noticeably slower pace. Triggered by major bankruptcies in the construction and CRE sector, Austrian banks’ overall loan quality started to deteriorate in late 2023. This trend continued in the first half of 2024, albeit at a noticeably slower pace. In recent quarters, the decline in banks’ credit quality was more pronounced in Austria than in other European countries.

As of mid-2024, the consolidated nonperforming loan (NPL) ratio of the Austrian banking sector stood at 2.7%, well above the all-time low of around 2% marked two years ago. Banks’ risk provisioning in the first half of 2024 did not, however, keep up with the increase in NPLs. This means that the NPL coverage ratio, i.e. the ratio of loan loss provisions to NPLs, fell to 40%, as old NPLs with higher coverage were written off and the volume of new NPLs still continues to grow. Compared to Austrian banks’ peak NPL ratio of nearly 7% in 2015, however, the current level is still moderate.

A further increase in forbearance points to a continued deterioration in credit quality. Forbearance involves granting concessions to borrowers who are unlikely to repay their loans under the current terms and conditions, with the aim to return borrowers to a sustainable repayment path. It can take the form of refinancing or restructuring the loan or modifying the terms and conditions. Forborne loans are a leading indicator of future credit quality. In Austria, the share of forborne loans in total loans increased from 1.7% at the end of 2022 to 2.3% as of mid-2024.

Borrower-based measures for residential real estate (RRE) lending in Austria (the KIM-V) are effective. 3 Data for the first half of 2024 show a further increase in the share of sustainable loans 4 , from 80% to 84%. Moreover, a difference-in-differences estimation shows that the introduction of the KIM-V is associated with a reduction of the NPL ratio for RRE loans, thus effectively contributing to financial stability (see the special topic in this report entitled “From part of the problem to part of the solution: evaluating the effectiveness of borrower-based measures in Austria”). This contributed to a relatively stable credit quality of RRE lending, where NPL ratios remain at 1.1%. 5 In the interest of administrative simplification, the KIM-V was amended for a second time in July 2024. The indicator-specific exemptions were abolished and only the 20% institution-related exemption remains in place. The key role of the KIM-V has also been emphasized internationally: S&P Global Ratings positively highlighted the regulation in their Banking Industry Country Risk Assessment for Austria 6 and acknowledged that exemptions remained largely unused. The rating agency also confirmed that the decline in new lending was the result of increased financing costs, and not brought about by the KIM-V. This remained true throughout the first half of 2024, as close to EUR 500 million in exemption volume continued to be available. The share of banks that used less than half of their available exemption volume increased from 46% in the second half of 2023 to 62% in the first half of 2024. Given that the KIM-V has its legal sunset date on June 30, 2025, the OeNB is currently evaluating if borrower-based measures remain necessary to address systemic risks from the RRE sector.

CRE loan woes have intensified in the course of 2024. CRE loans have been under scrutiny by Austrian and international supervisors for several years now. Since interest rates started to increase in 2022, the vulnerabilities of this sector’s funding, which rested on increasing real estate values and low interest rates, have come to the fore. The number of defaults of real estate companies 7 has increased, as have nonperforming CRE loans on banks’ balance sheets (see chart 5). CRE loan loss provisions have increased as well, but to a lesser extent. Accordingly, CRE loans’ coverage ratios have decreased, while real estate values – another cushion to protect banks from losses in the event of defaults – have been under pressure as well.

This is Chart 5.

This report features a special topic on systemic risks from CRE loans in Austria. It finds that a further deterioration of the economy and of real estate valuations, as experienced in past crises, could lead to CRE loan losses that are not covered by regulatory (“pillar 1”) and microprudential (“pillar 2”) requirements. The FMSB has concurred with this assessment and found that potential losses from CRE loans, in the event of a further deterioration of the economic environment, can pose an increased risk to financial stability in Austria. After its 42nd meeting, the FMSB therefore recommended that Austria’s Financial Market Authority (FMA) set a sectoral SyRB of initially 1% for risk-weighted exposures to domestic nonfinancial corporations of the ÖNACE 2025 sectors “M.68 real estate activities,” “F.41 construction of buildings” and “F.43 specialised construction activities” as of July 1, 2025. As limited-profit housing associations do not pose a systemic risk due, among other things, to their markedly lower probabilities of default, the FMSB recommended excluding them from the scope of the sectoral SyRB.

This is Chart 6.

The importance of Austrian banks’ foreign business continued to grow. With EUR 530 billion in foreign assets as of June 2024, 43% of Austrian banks’ business was located abroad (see chart 6), mainly within the EU. The most important foreign markets are Czechia, Germany and Slovakia, accounting for 40% of all foreign business. While most banks’ business is done locally, either in Austria or via subsidiaries in host countries, one-fifth of all business occurs across borders.

This is Chart 7.

The total assets of Austrian banking subsidiaries in Central, Eastern and Southeastern Europe (CESEE) surpassed EUR 300 billion in mid-2024, 8 with more than 80% located in EU member states. Six countries continue to be dominant, as Czechia accounts for 38%, Slovakia and Romania make up 15% and 12%, respectively, followed by Hungary, Russia and Croatia with shares of less than 10% each. As shown in chart 7, growth was strong during the COVID-19 pandemic but lost steam over the last two years. One of the reasons was tighter monetary policy, but as inflationary pressures are decreasing in the region and central banks start cutting rates, it will be important to monitor loan growth.

This is Chart 8.

Profits of Austrian banking subsidiaries in CESEE reached a new high of EUR 3.1 billion in the first half of 2024, driven by higher net interest income and a marginal provision release. Net interest income of Austrian subsidiaries in CESEE rose to EUR 4.5 billion (+11% year on year), boosted by moderate asset growth (+3%) and an expansion of the net interest margin to 3.1% (+22 basis points). At the same time, fees and commissions income fell by 13% to EUR 1.8 billion. Consequently, operating income rose by 2% to EUR 6.6 billion. 9 As operating costs declined by 3%, driven by staff cost that fell despite ongoing wage pressures, the subsidiaries’ operating profit reached EUR 3.8 billion (+6% year on year). Risk costs were very benign in the first half of 2024, as EUR 26 million of credit risk provisions were released, compared to a buildup of more than EUR 300 million in the same period of 2023.

Credit quality of Austrian banking subsidiaries in CESEE remains at historically good levels, as reflected in an NPL ratio for total loans of just 1.9%, a 65% coverage of NPLs (both stable year on year) and a noticeable increase in the share of stage 1 loans, i.e. loans with no significant increase in credit risk since initial recognition (see chart 8). All these trends resulted in a half-year profit of EUR 3.1 billion, up 15% from last year. As of mid-2024, the aggregate CET1 ratio of Austrian banks’ CESEE subsidiaries stood at 20% and their loan-to-deposit ratio was 71%. These solid levels reflect past efforts by banks and supervisors to make local banking systems more resilient by increasing the subsidiaries’ risk-bearing capacity and ensuring a balanced refinancing structure. 10

As for macroprudential capital requirements, the SyRB addresses structural systemic risks, such as the domestic banking sector’s specific ownership structures and its high exposure to emerging economies in Europe. 11 Disruptions in the whole or in parts of the Austrian financial system may entail severe negative consequences for the entire financial system and the real economy. In 2024, the SyRB was evaluated according to a biennial assessment plan. As it was found that the major structural systemic risks identified in the previous assessment from 2022 continue to exist, the FMSB recommended keeping SyRB rates unchanged. All previously identified banks will have to maintain a SyRB of 0.5% to 1.0%. Additionally, one institution was identified as a SyRB bank for the first time (with a rate of 0.5%). 12

Recommendations by the OeNB

The profitability and capitalization of the Austrian banking sector remained strong in the first half of 2024. Nevertheless, a geopolitical polycrisis, two years of domestic recession in 2023 and 2024 as well as the forecast rise in the Austrian unemployment rate mark a challenging economic backdrop for financial stability. 13 Rising pressures, including weakening domestic corporate credit quality, are likely to challenge earnings over time, while less restrictive monetary policy will take time to stimulate loan growth. The OeNB recommends that banks further strengthen financial stability by taking the following measures:

  • Continue to safeguard or, where appropriate, further strengthen their capital position by exercising restraint regarding profit distributions.
  • Adhere to sustainable lending standards for residential as well as commercial real estate (CRE) financing and prepare for stricter supervisory requirements for CRE loans.
  • Ensure adequate risk management practices, especially a commensurate coverage of NPLs by risk provisions and a conservative valuation of collateral.
  • Ensure sustainable profitability by maintaining cost discipline while investing in information technologies as well as in protection against cyber risks and the impact of climate change.

“Never waste a good crisis” – The OeNB’s Crisis Simulation Tool

Background

The OeNB has developed a Crisis Simulation Tool 14 that allows supervisors to run macroeconomic crisis scenarios for all Austrian banks and provides timely and accurate information in times of a global polycrisis. Indeed, the COVID-19 pandemic, Russia’s war against Ukraine, the ensuing period of high inflation, the disruption of global supply chains and energy markets, as well as the end of zero interest-rate policies have entailed new and complex challenges for bank supervisors. The volatile business environment and emergence of novel and very different shocks require swift supervisory action and a flexible toolkit that allows for real-time evaluations.

Introducing the tool

Supervisors can use the Crisis Simulation Tool to run both standardized and customized shock scenarios, based on economic sectors and countries, on the profits, capital positions and leverage ratios of individual banks and groups of banks. The tool requires that supervisors take two main decisions: first, whether they want to run a predefined macroeconomic scenario or create a custom scenario; second, whether the chosen scenario should be applied to an individual bank or a group of banks. Both predefined and custom scenarios are based on economic sectors (based on NACE codes 15 ) and countries, which subsequently act as filters on the banks’ credit portfolio. Having picked a scenario and a bank or group of banks, analysts can set the rate of default on the banks’ unsecured credit exposure in the countries and sectors included in the scenario. In addition, supervisors can determine a specific haircut on the existing collateral and allow for mitigating factors such as the bank’s expected profits, retained earnings or hidden reserves. Based on the settings, the Crisis Simulation Tool calculates additional impairments and their effects on profits, capital position and leverage ratio, and flags potential breaches of supervisory capital requirements in real time. For individual banks, the tool provides an in-depth analysis of the scenario’s impact, while at the banking group level, a more abstract, aggregated view is available.

Applications

The Crisis Simulation Tool is utilized both in micro- and macroprudential supervision, and its results have been reported to senior management and external stakeholders. Within microprudential supervision, the tool has been used to swiftly identify vulnerable banks in times of crisis, assess potential breaches of regulatory capital requirements, challenge banks’ assumptions and statements, and complement analytical assessments and reports with macroeconomic shock simulations. The results, including the identification of vulnerable banks with significant commercial real estate exposures and an impact study of the energy crisis on Austrian banks, have regularly been reported and presented to senior management and external stakeholders, such as Austria’s Financial Market Authority (FMA) and the ECB. Moreover, supervisors employed the Crisis Simulation Tool to run reverse scenarios to identify individual danger thresholds and determine specific risk potentials for Austrian SIs (significant institutions) and LSIs (less significant institutions).

The tool has been recently extended to macroprudential applications. It is used there to quickly assess the exposure to and the potential capital losses from macroeconomic shocks or crisis events from an aggregate banking-sector perspective to detect systemic vulnerabilities from banks’ credit exposure and to assess the robustness of current profitability and credit quality trends. After potential systemic risks have been identified in specific sectors, these preliminary results are sometimes utilized as the basis for a more detailed analysis, e.g. in the special topic in this report entitled “Systemic risks from commercial real estate lending of Austrian banks.” In all these applications, the tool’s focus on credit risk shocks rather than a gradual worsening of the macroeconomic environment as well as its time horizon, which is limited to the current year, makes it a complement to, and in no way a substitute for, supervisory stress tests.

Database and technical details

The Crisis Simulation Tool is based on well-established regulatory reporting data and macroeconomic crisis scenarios defined by OeNB economists and macroprudential supervisors. The tool’s most important data sources are the granular credit data and credit risk data, which are aggregated for each bank based on the countries and sectors of the economy included in the chosen scenario. They are complemented by capital adequacy, leverage, profitability and balance sheet data, which are necessary to calculate the additional impairments and subsequently the impact of the chosen scenario on profits, capital position and leverage ratio. All regulatory data are retrieved on a quarterly basis. The predefined macroeconomic scenarios have been designed by OeNB economists and macroprudential supervisors. The tool itself is a web-based R Shiny solution.

Conclusion

Overall, the Crisis Simulation Tool has proven to be a timely and cutting-edge addition to the OeNB’s existing analytical toolkit. Indeed, the Crisis Simulation Tool has filled a void between regular analytical reports and the annual supervisory stress tests by allowing for rapid initial assessments of shock scenarios. For example, the tool has enabled supervisors to swiftly identify vulnerable banks with significant commercial real estate exposures or to conduct an impact study of the energy crisis on Austrian banks. Since it can be easily adapted and extended to cover new emerging crisis scenarios and provides easily accessible real-time evaluations with high user convenience and satisfaction, it enables supervisors to conduct informed assessments and stay ahead of the curve even during the current polycrisis.

2 https://www.fmsg.at/en/publications/warnings-and-recommendations/2024/recommendation-fmsb-4-2024.html

3 KIM-V is the regulation for sustainable loan origination standards for residential real estate financing (in German: Kreditinstitute-Immobilienfinanzierungsmaßnahmen-Verordnung).

4 Sustainable loans are all loans with a debt service-to-income ratio of up to 40%, a maturity of up to 35 years and a loan-to-collateral ratio of up to 90%. Loans that are not clearly assignable are classified as sustainable.

5 The special topic uses the median corrected NPL ratio on an unconsolidated level for significant institutions in Austria to ensure comparability to the control group. The consolidated NPL ratio in Austria stands at 1.4% in mid-2024.

6 S&P Global Ratings (August 2024) Banking Industry Country Risk Assessment: Austria.

7 ÖNACE sectors “F construction” and “L (as of 2025 M) real estate related activities.”

8 A first since UniCredit Bank Austria transferred its CESEE subsidiaries to its Italian parent in 2016.

9 Changes in trading and valuation income cancelled each other out.

10 For more details, refer to box 4 entitled “Success of the Austrian Sustainability Package” in the Financial Stability Report 47.

14 The tool’s conception and implementation were led by Thomas Kögler and Thomas Resch, who are both members of the Expert Pool for Business Model Assessment, ESG and Digitalization within the Off-Site Supervision Division – Less Significant Institutions. They also authored this box. A joint team of off-site supervisors and the Financial Stability and Macroprudential Supervision Division leads the future development of the tool.

15 The tool utilizes NACE (Nomenclature statistique des activités économiques dans la Communauté européenne = Statistical Classification of Economic Activities in the European Community) levels 1 (containing 21 sections of the economy) and 2 (containing 88 subdivisions).

The impact of the digital euro on Austrian banks from a financial stability perspective

Manuel Gruber, Christoph Siebenbrunner, Alexander Trachta, Christian Wipf 16

We study the impact the introduction of the digital euro might have on Austrian banks from a financial stability perspective. The premise is that the digital euro will not bear interest and will be subject to a holding limit. More specifically, we analyze (1) the impact on Austrian banks’ liquidity positions in a liquidity stress scenario and (2) long-run profitability effects on banks’ net interest income and income from payment services. With respect to liquidity risk, we find substantial effects only for extreme scenarios and high holding limits. For instance, at a holding limit of 3,000, the most extreme stress scenario we consider results in outflows of 9.0% of total retail deposits into the digital euro. Besides, 7.4% of banks (accounting for 4.2% of the sample’s total assets) would breach the regulatory liquidity coverage ratio (LCR) threshold of 100%. Smaller banks are disproportionately affected because they have a larger share of retail funding, which leads to higher outflows. The picture is similar with respect to the long-run effects on banks’ net interest income. In a similarly extreme scenario as above, we estimate that the digital euro would cause interest income losses and a drop in the aggregate sample return on equity (RoE) of 51 basis points – the aggregate sample RoE is 14.9% – at a holding limit of 3,000. Smaller banks and less capitalized banks would be affected more strongly. In a more realistic scenario, the effects are substantially lower, with 1.0% of total retail deposits outflowing and the aggregate sample RoE dropping by 5 basis points. Lower holding limits effectively contain adverse outcomes both with respect to interest income losses and liquidity risk. As to the effect on payment services income, it is harder to arrive at reliable estimates given a lack of suitable bank-level data and high uncertainty about the digital euro’s impact on transaction volumes and fees of retail current accounts and about how digital euro transactions and account management will be remunerated. In a tentative estimation, we find an aggregate sample RoE effect of around 26 basis points. By determining the remuneration of the digital euro, the central bank can effectively control the magnitude of this effect. Overall, we conclude that the introduction of a digital euro would not pose a threat to the stability of the Austrian banking system provided the digital euro is subject to a carefully designed holding limit and remuneration model. From a purely financial stability perspective, low holding limits would be preferable to higher ones.

JEL classification: E42, G21

Keywords: central bank digital currencies, digital euro, financial stability, substitution of bank deposits, liquidity, profitability, business models

Central banks worldwide are investigating the issuance of central bank digital currencies (CBDCs). In October 2023, the Eurosystem finalized the two-year investigation phase for a euro area CBDC, i.e. the digital euro. We are now one year into the preparation phase scheduled to last until October 2025. After that, the Governing Council of the European Central Bank (ECB) will decide whether the digital euro project will progress toward potential development and rollout (ECB, 2024). With the digital euro, the Eurosystem aims for the euro to evolve alongside the general public’s digital payment preferences and to facilitate electronic payments in the euro area. The digital euro is also meant to strengthen Europe’s monetary infrastructure and sovereignty by reducing Europe’s dependence on non-European private payment providers that currently dominate the European payments landscape. At the same time, a widely accepted CBDC might generate systemic repercussions for bank intermediation, which might have negative effects on financial stability. If households substitute CBDC for bank deposits, which are a relatively stable and cheap funding source for banks, this might have adverse consequences for banks’ liquidity and profitability. Ultimately, this might also impact the overall resilience of the banking sector and its intermediation function for the real economy.

We study these concerns, using bank-level data from Austrian banks. 17 We first analyze the impact the digital euro would have on Austrian banks’ liquidity risk in a stress scenario in section 1. Then, in section 2, we assess the long-run profitability effects of the digital euro, in particular on banks’ net interest income and income from payment services, abstracting from initial implementation costs. A particular emphasis lies on bank heterogeneity because the impact of the digital euro most likely depends on a bank’s business model, e.g. on its share of financing from retail deposits.

Consistent with the current proposals of the ECB, 18 we model the digital euro as a digital alternative to cash, which does not bear interest and can be held by households only. Households can hold digital euro up to a certain holding limit set by the ECB. Their digital euro accounts are directly linked with their other payment accounts to automatically top up the digital euro account up to the holding limit. This “(reverse) waterfall approach” allows households to transact any amount in digital euro, independent of the holding limit.

In the following sections, we focus on a baseline scenario characterizing the most likely outcome and a maximal scenario that captures the extreme but very unlikely upper bound for outcomes. The baseline scenario is mainly calibrated by using the Deutsche Bundesbank’s Survey on Consumer Expectations of 6,000 households in Germany presented in Bidder et al. (2024). This survey contains information on how households plan to use the digital euro both in normal and in crisis times.

1 Liquidity effects

To assess the financial stability impact of the digital euro on Austrian banks’ liquidity positions, we use a sample of 393 Austrian banks at the unconsolidated level, which reported household salary and pension accounts in the second quarter of 2023. 19 Table 1 presents some descriptive statistics of the sample.

Table 1: Descriptive sample statistics  
Total assets,
EUR billion
Household
sight deposits,
EUR billion
Salary/pension
accounts,
EUR million
Accounts per
bank (median)
Account size
(mean), EUR
727.3 191.9 5.75 5,200 33,400
Source: OeNB.

We consider a systemic liquidity stress scenario, where e.g. due to a sudden loss of confidence in the banking sector, Austrian salary and pension account holders abruptly transfer deposits up to the holding limit from their deposit accounts to a digital euro wallet. We model the deposit outflow for bank i in such a liquidity crisis as follows

〖out〗_(i,crisis)=d€uptake_crisis*#accounts_i*(holding limit-d€holdings) (1)

where d€uptakecrisis is the average ratio of account holders that adopt the digital euro in a liquidity crisis, #accountsi is the number of salary and pension accounts of bank i, holding limit is the holding limit set by the central bank and d€holdings are the average intended digital euro holdings of a digital euro adopter before the crisis. If d€holdings > holding limit, digital euro adopters just hold the holding limit. Note that (1) assumes that the share of account holders that adopt the digital euro is uniformly distributed across banks and all digital euro adopters have enough deposits to withdraw up to the holding limit. We now calibrate the baseline and the maximal scenario as follows (table 2).

Table 2: Calibration of the baseline and the maximal scenario liquidity effects  
Baseline Maximal Description/Source
d€uptake_crisis 0.6 1 Bidder et al. (2024)
holding limit (200, 5,000) (200, 5,000)
d€holdings 1,000 0
Source: OeNB.

In the baseline scenario, d€uptakecrisis is the upper bound of Bidder et al. (2024), who find that in a crisis event 56% of respondents would adopt the digital euro. In the maximal scenario, we assume that all account holders adopt the digital euro. We also assume that digital euro adopters intend to hold EUR 1,000 in digital euro in the baseline scenario before the crisis, while in the maximal scenario they hold no digital euro. We motivate the low intended digital euro holdings with the “reverse waterfall approach” explained above and the unremunerated nature of the digital euro. Note that the maximal scenario means that households hold no digital euro before the crisis but in the crisis they all transfer deposits up to the holding limit into the digital euro.

Table 3: Deposit outflows under liquidity stress: baseline vs. maximal scenario  
Baseline scenario Maximal scenario
Holding limit Deposit outflow,
EUR billion
% of total assets % of deposits Deposit outflow,
EUR billion
% of total assets % of deposits
200 0.0 0.0 0.0 1.1 0.2 0.6
1,000 0.0 0.0 0.0 5.7 0.8 3.0
3,000 6.9 0.9 3.6 17.2 2.4 9.0
5,000 13.8 1.9 7.2 28.7 4.0 15.0
Source: OeNB.

Table 3 shows deposit outflows in the two scenarios. In the baseline scenario, deposit outflows increase linearly from holding limits above 1,000 (since households already hold 1,000 digital euro before the crisis) to EUR 13.8 billion or 7.2% of total sight deposits at a 5,000 holding limit. In the maximal scenario, deposit outflows increase linearly to EUR 28.7 billion or 15% of total sight deposits at a holding limit of 5,000.

Do banks have enough liquid assets to withstand these outflows? To answer this question, we calculate banks’ liquidity coverage ratios (LCR). In other words, we compare the high-quality liquid assets (HQLA) banks hold to their outflows due to the liquidity crisis outi,crisis plus their other net liquidity outflows (NLO). 20

〖LCR〗_i=(HQLA_i)/(NLO_i+0.95*〖out〗_(i,crisis) ) (2)

Chart 1 shows how many banks have an LCR below 100% due to these outflows as well as their share in total assets. In the baseline scenario, we only see material effects for high holding limits. At a holding limit of 5,000, 4.6% of banks (3.6% of the sample’s total assets) have an LCR below 100%. In the maximal scenario, the effects are more substantial. At a holding limit of 3,000, 7.4% of banks (4.2% of the sample’s total assets) have an LCR below 100%; at a holding limit of 5,000, this is the case for 27% (9.5% of the sample’s total assets). The divergence between the share of banks and their share of total assets indicates that it is mostly smaller banks that are affected. The reason is that, compared to large and medium-sized banks, smaller banks tend to have more accounts and more retail financing relative to their assets (see table A1 in the annex). Overall, the effects of the liquidity stress scenario are substantial only for high holding limits. Thus, a careful calibration of the holding limit can contain the adverse effects of such a scenario even in the extreme and very unlikely maximal scenario.

This is Chart 1.

2 Profitability effects

To assess the financial stability impact of the digital euro on Austrian banks’ solvency, this section first analyzes the effects on banks’ net interest income (NII) and then the effects on banks’ net payment services income (NPI). In contrast to the crisis focus in the liquidity part, we now concentrate on the digital euro’s profitability effects in “normal times.” This steady state perspective also abstracts from initial introduction costs.

2.1 Net interest income (NII)

Banks might suffer NII losses due to deposit outflows into the digital euro either because banks must replace deposits with more expensive funding or because they shrink their assets. As shown in (3) we model average deposit outflows of bank i outi,normal as a product of the average share of account holders that adopt the digital euro in normal times d€uptakenormal and the number of accounts of bank i #accountsi, the average intended digital euro holdings of digital euro adopters and the fraction of digital euro holdings which digital euro adopters substitute for sight deposits (and not for cash) deposit_sub:

〖out〗_(i,normal)=d€uptake_normal*#accounts_i*d€holdings*deposi〖t_sub〗_ (3)

The NII loss is then the product of outi,normal and the funding advantage of sight deposits, funding_advantage.

〖NII_loss〗_i=〖out〗_(i,normal)*fundin〖g_advantage〗_ (4)

We calibrate the baseline and the maximal scenario as shown in table 4.

Table 4: Calibration of the baseline and the maximal scenario NII effects  
Baseline Maximal Description/Source
d€uptake_normal 0.5 1 Bidder et al. (2024)
d€holdings 1,000 5,000
deposit_sub 0.64 1 Bidder et al. (2024)
funding_advantage 1.61% 1.61% Average interest spread
between new term deposits
and household sight deposits
in Austria 2003–2008
Source: OeNB.

In the digital euro survey of the Deutsche Bundesbank (Bidder et al., 2024), 46% of households responded they would adopt the digital euro in normal times. We take the upper bound of this and – similar to the liquidity part – assume 100% adoption in the maximal scenario. d€holdings are calibrated as in the liquidity section. However, note that high digital euro holdings map into high outflows here, while in the liquidity section high digital euro holdings implied low outflows. Thus, we choose EUR 5,000 for the maximal scenario here (and as a robustness exercise we also consider 3,000 or 5,000 intended digital euro holdings in the baseline scenario). The substitution parameter is derived as follows: In Bidder et al. (2024), digital euro adopters project to hold 21.1% of their liquid portfolio in digital euro (a share similar to cash) while reducing their deposit share by 13.4 percentage points to this end. Thus, deposit_sub is 0.134/0.21=0.64. Note that NII losses are zero if digital euro holders completely substitute digital euro holdings for cash, i.e. deposit_sub is zero. Finally, the funding advantage banks lose with deposit outflows is calibrated to the period before the very low or negative interest rate period with deposit rates stuck at the zero lower bound. The value is close to Austrian banks’ average net interest margin in 2023 (1.53%).

Table 5 shows the deposit outflows and NII losses in the two scenarios. We also express NII losses relative to tier 1 bank capital, thus capturing the effect on the return on equity (RoE). In the baseline scenario, the aggregate deposit outflows amount to EUR 1.8 billion, which results in an NII loss of EUR 29 million and an RoE effect of 5 basis points, which is very small compared to the aggregate RoE (14.9%) in the sample. 21 In the maximal scenario, the effects are more material. Deposit outflows here exactly correspond to the outflows in the maximal crisis scenario above. This is because, with intended digital euro holdings of 5,000, the holding limit always binds and deposit outflows equal the holding limit, as was the case in the liquidity crisis scenario. Also note that the holding limit effectively contains the more material effects in the maximal scenario. Choosing a 3,000 holding limit instead of a 5,000 holding limit reduces the RoE effect from 86 basis points to 51 basis points.

Table 5: Deposit outflows and NII losses: baseline vs. maximal scenario  
Baseline scenario Maximal scenario
Holding limit,
EUR
Deposit outflow,
EUR billion
NII loss,
EUR million
RoE effect,
basis points
Deposit outflow,
EUR billion
NII loss,
EUR million
RoE effect,
basis points
200 0.4 6 1 1.1 19 3
1,000 1.8 29 5 5.7 93 17
3,000 1.8 29 5 17.2 278 51
5,000 1.8 29 5 28.7 463 86
Source: OeNB.

Chart 2 identifies banks that are particularly affected by NII losses. In the baseline scenario, no bank has an RoE effect above 100 basis points. We only have significant effects in the maximal scenario, where the share of banks with an RoE effect above 100 basis points increases approximately linearly from a holding limit around 1,500. At a 3,000 holding limit, 23.1% of banks (22.6% of the sample’s total assets) have an RoE effect above 100 basis points, while at a 5,000 holding limit, 59.3% of banks (37.3% of the sample’s total assets) have an RoE effect above 100 basis points. As in the liquidity section, the NII losses rather affect small banks but not as strongly as was the case with effects on the LCR above.

This is Chart 2.

2.2 Net payment services income (NPI)

Due to data limitations 22 and the high uncertainty about the digital euro’s impact on banks’ NPI, we restrict the analysis to the Austrian banking sector as a whole in this section. We model aggregate NPI as a linear function of the number of transactions T and the return per transaction with sight deposits RD, i.e. NPI=T*RD. After the introduction of the digital euro, NPI changes to NPI’=(1–x)T*RD’+xT*Rd€ where x is the share of transactions transferred into the digital euro and Rd€ is the return per transaction with digital euro. Note that this NPI definition assumes that the income from digital euro transactions remains within the banking sector. Thus, the NPI loss is calculated as follows:

(5)

where RD’/RD and Rd€/RD denote the change in the return of transactions with sight deposits and digital euro relative to the current return. Rd€/RD can thus be interpreted as a parameter measuring how the central bank remunerates digital euro transactions. We calibrate the baseline and the maximal scenario as shown in table 6.

Table 6: Calibration of the baseline and the maximal scenario NPI effects  
Baseline Maximal Description
NPI 1.1 1.1 Aggregate NPI of sample banks in EUR billion
x 0.16 0.32 16% (32%) of transactions move to the digital euro
R’D/RD 0.95 0.90 Profitability of transactions with sight deposits decreases by 5% (10%)
Source: OeNB.

To estimate the NPI, we proceed as follows: We use the average end-of-year 2020–2023 NPI of the banks for which NPI data are available (EUR 2.16 billion) and set this number in relation to the average end-of-year sum of sight deposits from households, nonfinancial firms and the government from 2020 to 2023. 23 This yields 0.57, which means that Austrian banks on average earn 0.57 cent per euro of (transactional) sight deposits. Assuming this also holds for the household deposits of the banks in our sample, we arrive at an aggregate NPI of EUR 1.1 billion. To define how many transactions move into the digital euro (x), we follow up on the reasoning in the NII section. There, we assumed in the baseline scenario that 50% of account holders adopt the digital euro. We further assume that the share of deposit substitutions we assumed there (64%) also holds for transactions and, finally, we assume that holders of a digital euro wallet split their transactions 50-50 between sight deposits and digital euro, which yields 0.16. For the maximal scenario, we double this to 32% outflows. Finally, the decrease in the profitability of transactions with sight deposits of 5% (10%) reflects the idea that the introduction of the digital euro also increases competition in the conventional NPI business with sight deposits and puts these margins under pressure.

Table 7: NPI losses: baseline vs. maximal scenario  
Baseline scenario Maximal scenario
Digital euro
remuneration
NPI loss,
EUR million
RoE effect,
basis points
NPI loss,
EUR million
RoE effect,
basis points
0.47 140 26 262 49
0.28 173 32 329 61
Source: OeNB.

Chart 3 shows the NPI losses for different digital euro remuneration levels Rd€/RD. Values like 0.5 mean that transactions in digital euro are remunerated at 50% of the current return on sight deposits. In table 7 we provide reasonable parameters for the remuneration of digital euro transactions, following the idea that the remuneration of digital euro transactions should target the return of the most efficient providers of transactions. 24 This yields NPI losses ranging from EUR 140 to 173 million (EUR 262 to 329 million) and RoE effects of 26 to 32 basis points (49 to 61 basis points) in the baseline (maximal) scenario. As with holding limits, carefully calibrating the digital euro remuneration prevents extreme profitability effects for the banking sector.

This is Chart 3.

References

Auer, S., N. Brazoli, G. Ferrerim, A. Ilari, F. Palazzo and E. Rainone. 2024. CBDC and the banking system. Banca d’Italia Occasional paper 829.

Bellina, M. and L. Cales. 2023. Bank profitability and central bank digital currency. JRC Working Papers in Economics and Finance 2023/06.

Bidder, R., T. Jackson and M. Rottner. 2024. CBDC and banks: Disintermediating fast and slow. Deutsche Bundesbank Discussion Paper 15/2024.

ECB. 2024. Progress on the preparation phase of a digital euro. June.

Meller, B. and O. Soons. 2023. Know your (holding) limits: CBDC, financial stability and central bank reliance. ECB Occasional Paper Series No 326.

Annex

Table A1: Sample statistics for small, medium-sized and large banks  
Small banks Medium-sized banks Large banks
Share of banks, % 79.9 16.5 3.6
Share of total assets, % 15.6 25.0 59.4
Share of accounts, % 27.6 27.4 45.0
Share of deposits, % 22.3 27.9 49.8
Accounts per total assets,
EUR million
14.0 8.7 6.0
Share of household deposits
in total assets, %
37.9 29.6 22.1
Source: OeNB.
Note: Small banks are defined as banks with total assets up to EUR 1 billion, medium-sized
banks’ total assets range from EUR 1 billion to EUR 10 billion, while large banks’
total assets amount to more than EUR 10 billion.

16 Oesterreichische Nationalbank, Supervision Policy, Regulation and Strategy Division, christoph.siebenbrunner3@oenb.at; Financial Stability and Macroprudential Supervision Division, alexander.trachta@oenb.at, christian.wipf@oenb.at. Opinions expressed by the authors of studies do not necessarily reflect the official viewpoint of the Oesterreichische Nationalbank or the Eurosystem. We would like to thank Markus Schwaiger (OeNB) for his feedback and guidance.

17 Similar papers include Auer et al. (2024), Bellina and Cales (2023), Bidder et al. (2024) and Meller and Soons (2023).

19 The total bank sample at the unconsolidated level consists of 452 banks, of which 59 report no salary or pension accounts. As the reporting standards for 49 banks of the Sparkassen sector have changed, the most recent data for these banks are from the fourth quarter of 2022. The study sample covers 74% of the total sample in terms of total assets and 95% of household sight deposits.

20 We subtract 5% of outflows from the other liquidity outflows in the denominator because the outflows from these retail deposits are already contained in the outflows due to the crisis. 5% is the usual outflow rate applied to retail deposits in the calculation of the LCR. The HQLA and the liquidity outflows of the Raiffeisen banks in Lower Austria, Vienna, Upper Austria, Styria and Vorarlberg as well as the Volksbanken banks are aggregated to a sector level due to special liquidity reporting requirements that apply to these banks. In the baseline scenario, we also adjust HQLAi for pre-crisis outflows per bank into the digital euro outi,normal derived in section 2.1 on net interest rate income losses below.

21 For intended digital euro holdings of 3,000 (5,000), the RoE effects are only slightly higher at 16 (28) basis points.

22 NPI data for Austrian banks are only available for a different bank sample.

23 The idea here is that sight deposits from these agents are mainly used for transactional purposes, while the sight deposits of financial firms and central banks do not serve this function.

24 The first (second) value, 0.47 (0.28), relates the NPI return of Austrian banks from sight deposits derived above (0.57) to the 25th (10th) percentile of the same NPI return of a sample of European SSM banks, 0.27 (0.16). The 25th percentile corresponds to the NPI return of Dutch banks in the SSM sample (0.28), where the payment system is often considered one of the most efficient and innovative in Europe.

Systemic risks from commercial real estate lending of Austrian banks

Marcel Barmeier, David Liebeg 25 , Sebastian Rötzer 26 , 27

The commercial real estate (CRE) market in Austria – and many other countries – has been under stress at least since interest rate increases began in 2022. Consequently, the evaluation of financial stability risks in the CRE segment is of high relevance for supervisory authorities and policymakers. This study contributes to this goal by providing an integrated approach to gauge systemic risks associated with CRE financing. Combining macroeconomic information with data on the loan, firm and bank level, we estimate the effect of adverse macroeconomic conditions on CRE loan portfolios of Austrian banks. We find that in an adverse scenario, nonperforming loan (NPL) ratios could increase to levels seen in international historical crises and a sizable share of bank capital could be depleted. Thus, we conclude that CRE loans, in the event of a further deterioration of the economic environment, pose an increased risk to financial stability in Austria. This is in line with the assessment that Austria’s Financial Market Stability Board (FMSB) made in its 41st meeting.

JEL classification: G01, G21, G28, R33

Keywords: commercial real estate, systemic risk, financial stability

Commercial real estate (CRE) funding has been at the forefront of worries of financial journalists, analysts, policymakers and investors since the start of interest rate increases in 2022 and even earlier. In the macroprudential sphere, the European Systemic Risk Board (ESRB) has issued a warning on vulnerabilities in CRE markets in Europe, following its recommendations on closing data gaps in 2016 and 2019 (ESRB, 2016, 2019, 2023). Austria’s Financial Market Stability Board (FMSB) has regularly warned about risks of CRE funding in Austria, as have the Financial Market Authority (FMA) and the OeNB nationally, as well as the ESRB, the European Central Bank (ECB) and the International Monetary Fund (IMF) internationally. In its 41st meeting, the FMSB concluded that “potential losses from commercial real estate loans, in the event of a further deterioration of the economic environment, can pose an increased risk to financial stability in Austria.” This paper constitutes a follow-up to the work of Liebeg and Liegler (2022).

We start our paper with a short introduction on insights from past CRE crises, then present the data available for our systemic risk assessment, describe our empirical strategy, lay out the scenarios we use and the results we obtain, and finally conclude.

1 Insights from CRE crises of the past

CRE crises occur with some regularity and are usually part of a wider real estate crisis or downturn in the economy as the CRE segment is strongly interconnected with both the real economy and the financial system (ESRB, 2023). A crisis in the CRE segment has both been a trigger (Deghi et al., 2021) and a consequence (Davis and Zhu, 2011) of a wider economic downturn. For our assessment, we have analyzed 12 CRE-specific crises since the early 1980s from the literature, some of which transcended national borders and affected a wider region.

Crowe et al. (2013), while concluding that CRE crises are difficult to pin down due to their interweaving with residential real estate (RRE) crises, find that CRE played an important role in the savings and loans crisis in the US in the early 1980s, in Japan from the late 1980s onward, in the Nordic crisis and in Australia in the early 1990s, in the Asian crisis in the late 1990s, and in Ireland and Spain in the late 2000s. Davis and Zhu (2011) employ a similar sample but add France, the UK, and the United States in the late 1980s and early 1990s to the list of property crisis episodes. Ellis and Naughtin (2010) add Australia, the UK and the US to the list of countries with CRE crises during the global financial crisis of the late 2000s. Herring and Wachter (1999) offer deeper insights into the CRE crises of the early 1980s in the US, the early 1990s in Japan and Sweden, and the late 1990s in Thailand. The Danish Systemic Risk Council (2023) finds that in the crisis of the late 2000s, lending to the corporate sector “real estate activities” has given rise to significant impairment charges for credit institutions.

While international CRE price data are comparably more widely available, data for CRE credit risk indicators, such as nonperforming loans (NPLs) and loan loss provisions from historical CRE crises, are scarce. In their overview, Deghi et al. (2021) show that CRE prices tend to reach their troughs roughly two years after their peaks, dropping by 20% to 56% in that time span. Some price drops last longer and are deeper – Ireland saw a decline by about two-thirds over five years. Ellis and Naughtin (2010) demonstrate that CRE price drops are usually larger than RRE price drops. As for risk indicators, the US Federal Deposit Insurance Corporation (FDIC) offers time series data on loan portfolio performance that are available for all FDIC-insured institutions from 1984 onward and for various aggregates of real estate loans. 28 Noncurrent rates 29 of US real estate loans increased from 0.9% in Q1 2007 to 7.6% in Q1 2010. In construction and development loans, there was an increase from 1% to 16.8% in the same time span. Banco de Espana’s BIEST 30 dataset offers time series data on real estate loan quality from 1992 onward. The NPL ratio in Spain increased from 0.7% in Q4 2007 to 34.3% in Q4 2013 for loans to the construction sector. In loans to the real estate activities sector, the ratio rose from 0.7% to 38% in the same period of time. The loan loss provisions ratio for loans to the construction sector increased from 0.3% to 18% during this six-year period. Finally, Danmarks Nationalbank has provided us with data on loan loss provisions for real estate activities loans: They grew from 0.5% in 2007 to 9.4% in 2013.

2 Data

This is Chart 1.

Defining CRE loans from a risk perspective is a challenging task. Depending on the scope, data sources and the applied perspective, the size of banks’ CRE exposure may vary significantly. For our systemic risk assessment, we use a sectoral perspective that includes all loans to domestic nonfinancial corporations in the “real estate activities” (ÖNACE 2008 sector L.68), “construction of buildings” (ÖNACE 2008 sector F.41) and “specialised construction activities” (ÖNACE 2008 sector F.43) sectors. 31 The interconnectedness of these sectors is shown by the high correlation of NPL ratios as well as the mutual reliance on intermediate consumption of goods from the sectors L “real estate activities” and F “construction” in the gross value added in the Austrian economy. 32 Furthermore, this definition of CRE loans allows us to focus on a homogenous group of corporations and thereby follow a targeted approach in modeling sensitivities of corporates to macroeconomic shocks. Recently, the European Central Bank (Ryan et al., 2023; ECB, 2024) and Danmarks Nationalbank (Danish Systemic Risk Council, 2023) have used a similar approach to assess risks from CRE financing. Following this definition, the total loans associated with CRE amount to EUR 127 billion as of December 2023, of which the majority (EUR 102 billion, 80%) fall under the “real estate activities” sector (see chart 1).

This is Chart 2.

One particularity of the Austrian CRE segment is the importance of limited-profit housing associations (GBVs). 33 GBVs account for 22% of domestic CRE loans, and the largest share is in the “real estate activities” sector. The risk structure of GBVs differs significantly from profit-oriented debtors (non-GBVs). As of year-end 2023, the NPL ratio stood at 3.7% for profit-oriented debtors, whereas it was 0% for GBVs. 34 Furthermore, we also use information from the Bureau van Dijk SABINA database to investigate the balance sheet structure of CRE companies. As of year-end 2021, CRE companies had on average a lower capitalization than companies from other sectors irrespective of profit orientation (see chart 2). 35 However, compared to profit-oriented debtors in the CRE sector, the number of GBVs with negative equity is much lower and close to zero. Moreover, with respect to the availability of liquid assets, we find that GBVs generally have stronger liquidity positions (cash and bank balances) than non-GBVs. Since better capitalization and liquidity as well as legal provisions effectively shield GBVs from market stress, we treat them separately in our assessment of systemic risks in the CRE segment.

3 Empirical strategy

Our systemic risk assessment follows a two-step approach. First, we estimate the sensitivity of individual borrowers’ probability of financial distress to a selection of macroeconomic drivers, which we then use to project expected probabilities of stress for a given set of macroeconomic scenarios. Second, we employ the projected probabilities and a dataset of lender-borrower interlinkages to conduct a simulation exercise to map simulated balance sheet stress of borrowers into banks’ portfolio losses. Our simulation approach follows the methodologies in Harrison and Mathew (2008) and Górnicka and Valderrama (2020), adapted to the distinct requirements of CRE lending, and allows us to track default probabilities (PDs) and losses given default (LGDs) for individual banks as well as the aggregate banking system. 36 Figure 1 illustrates the two main steps and their respective substeps, with a more detailed description given below.

3.1 Step I – Projecting CRE companies’ probability of financial distress

Earlier studies on systemic risks from real estate financing have primarily focused on households. However, CRE companies are fundamentally different: Their debt servicing capacity depends, at least to some degree, on price developments in real estate markets. That is, both the first and second lines of defense of the banking system’s loan book depend on the same macroeconomic drivers. Therefore, and due to the gap in the existing literature, the modeling of borrowers’ probability of financial distress conditional on macroeconomic states is a key contribution of our study.

This is Figure 1.

We use a two-step approach to map changes in macroeconomic state variables into shifts in probabilities of firm balance sheet stress. First, using data from the Banque de France’s BACH database, we estimate the impact of changes in macroeconomic variables on within-country aggregate revenue changes in the construction and real estate sectors. In the second step, following the micro-simulation approach of Ryan et al. (2023), we employ Bureau van Dijk SABINA data on firm-level balance sheets in the construction and real estate sectors to assess liquidity and solvency stress for a large number of historically plausible shocks to revenues and interest rates. In determining the thresholds for a corporation to fail, we rely on Guth et al. (2020) and Puhr and Schneider (2021): A firm is insolvent when either its cash and bank reserves are below –10% (e.g. bank lines are overdrawn) or its equity ratio is below –30%. 37 Taken together, the combined sensitivities obtained from these two exercises allow us to project any macroeconomic scenario, be it historical or hypothetical, into a shift of borrowers’ probabilities of solvency or liquidity stress. 38 To ensure that our projection method returns only real probabilities, we constrain output values of stressed borrower PDs, i.e. base PDs from the lender-borrower level data plus shifts, to the interval of [0.05%, 100%].

Taken together, let βNACE denote a vector of sectoral sensitivities to the vector of macroeconomic changes, ∆Z, including the particular element ∆NFCrate for the change in corporate lending rates. Let αS and αR be the sensitivities obtained from the micro-simulation in the second step and let PSSi and ShareVariablei be the ex ante probability of stressed sales by firm i and its share of variable rate loans. Then the projected probabilities of CRE companies’ financial distress in our model are given by

where the lower bound of projected stress probabilities of 5 basis points is aligned with the lower bound guidance in the capital requirement regulation (CRR) version III.

3.2 Step II – Simulation of banks’ portfolio losses from real estate financing

Using our projection method for borrowers’ probabilities of financial distress and the macroeconomic scenarios detailed below, we conduct a simulation exercise to gauge the conditional loss distributions of Austrian banks’ CRE credit portfolios. To this end, we construct a large sample of lender-borrower relationships where we track the total exposure amount, the loan’s conditional net present value (NPV), the risk premium and the available loan collateral in the form of residential and commercial real estate as well as other, non-real estate, assets. For each out of a sample of S = 2000 simulations, we draw a vector of financial distress indicator variables for the population of borrowers, where an outcome of 1 indicates financial distress and the probability of such an event is governed by the projection method detailed in step I above.

For each borrower in distress, the process of distress resolution, as illustrated in step II of figure 1, may lead to economic default if the proceeds from a collateral sale cannot cover the cost of debt following Harrison and Mathew (2008) and Górnicka and Valderrama (2020). Any resulting losses are then collected at the bank portfolio level. This approach allows us to track measures like probability of economic default, loss given default and NPL ratios not only on the individual bank-level but also for within-bank subportfolios, such as lending to each of the individual economic sectors considered, as well as to distinguish between profit-oriented debtors and GBVs.

In the implementation of our simulation methodology, we place particular emphasis on the consistency of the macroeconomic channels between the projection method in step I and the resolution of credit risk in step II. That is, the same shocks that drive up distress probabilities of firms in the broader real estate sector also reduce collateral value and increase ex ante loan NPVs. This is to maintain the intuitive perspective on how this sector accumulates a systemic risk to financial stability: It is the twofold dependency of debt servicing capacity and collateral value on real estate prices and interest rates that makes the real estate sector, and thereby its lenders, particularly vulnerable to adverse shifts in the macroeconomic environment.

4 Macroeconomic scenarios

Table 1: Macroeconomic scenarios: cumulative change over three years  
Baseline Adverse
%
Real GDP +4.1 –5.0
Risk free and
corporate rate
Unchanged
from YE 2023
Unchanged
from YE 2023
CRE prices +2.4 –28.4
RRE prices +4.1 –33.2
Source: OeNB, ESRB, authors’ compilation.

For our systemic risk assessment, we need macroeconomic scenarios for the paths of domestic GDP and property prices (both residential and commercial) as well as that of interest rates. Our macroeconomic scenarios for GDP and real estate prices 39 draw from the OeNB 2024 banking stress test exercise (see box 1 “Results of the OeNB’s 2024 solvency stress test” in this report’s “Recent developments and macroprudential policy update”), while we keep the risk free and corporate interest rates at the level as of year-end 2023.

Following the usual approach in stress testing, we use a three-year horizon, which is also backed by the experiences made in historic CRE crises that roughly lasted from two to six years peak to trough. In line with the static balance sheet assumption, we assume that the banks’ balance sheets do not change over time. The adverse scenario of our stress simulation assumes a severe stagflation: a period of negative GDP growth associated with elevated levels of inflation that hinders central banks from lowering their policy rates and a materialization of accumulated risks in the real estate sector that leads to an extended fall in property prices.

Cumulative GDP growth from 2024 to 2026 is 4.1% in the baseline scenario and –5.0% in the adverse scenario. CRE property prices rise by 2.4% in the baseline scenario and fall by 28.4% in the adverse scenario; RRE property prices increase by 4.1% and decline by 33.2%, respectively. 40

5 Results

This is Chart 3.

Applying the discussed methodology and the baseline as well as the adverse macroeconomic scenario to the Austrian banking system according to year-end 2023 data, we observe that CRE financing poses a heightened systemic risk to financial stability. We come to this conclusion by investigating the changes in the estimated PDs and LGDs as well as the impact on the NPL ratio and capitalization of Austrian banks.

In the adverse macroeconomic scenario, the estimated PDs and LGDs for CRE loans increase significantly. While the PD is 0.3% is the baseline scenario, it increases to 16.4% in the adverse scenario. LGDs triple from 12% in the baseline scenario to 35.3% in the adverse scenario (see chart 3).

Consequently, the increase in PDs also leads to a higher share of NPLs. We find that our estimated NPL ratios in the adverse scenario are in the range of historical CRE crises: During those crisis periods, the NPL ratio stood at 8%–17% in the United States 41 and 34%–38% in Spain. 42

Increases in PDs and LGDs subsequently also increase banks’ expected credit losses. In total, bank losses are estimated to amount to 13% of total CET1 capital in the adverse scenario. One third of capital headroom, i.e. the difference between CET1 capital and the required CET1 capital for fulfilling the overall capital demand (OCD), 43 could be depleted in the adverse scenario. To put the number in another perspective, estimated bank losses could be larger than the average annual bank profits between 2019 and 2023 (see chart 4). Virtually all losses (98%) stem from profit-oriented debtors (non-GBVs), thus confirming the risk-mitigating character of limited-profit housing associations in Austria. This observation is supported by the distribution of losses across banks. Banks with a high share of GBV financing have lower average losses.

This is Chart 4.

In a systemwide CRE crisis, various factors that are not incorporated in our model would influence the severity of the crisis. While banks’ operating profits could mitigate some effects, various aspects could amplify the impact of the crisis. These include an increase in bank funding costs, interbank contagion effects as well as negative spillovers to other industries. In our view, these observations confirm the systemic nature of risks associated with CRE financing.

6 Conclusions

Losses from commercial real estate (CRE) loans, in the event of a further deterioration of the economic environment, pose an increased risk to financial stability in Austria. We come to this conclusion by extending the approaches used by, among others, the European Central Bank and the International Monetary Fund to identify systemic risks in real estate markets. We show that in the event of an adverse macroeconomic development, a sizeable number of loans could become nonperforming and a significant share of available capital in the Austrian banking sector could be depleted. Our results therefore support the view of Austria’s Financial Market Stability Board (FMSB) that macroprudential measures are necessary to address systemic risks stemming from CRE loans. Accordingly, the FMSB recommended that the Financial Market Authority set a sectoral systemic risk buffer of initially 1% of risk-weighted CRE assets. 44

References

Crowe, C., G. Dell’Ariccia, D. Igan and P. Rabanal. 2013. How to deal with real estate booms: Lessons from country experiences. In: Journal of Financial Stability 9(3). 300–319.

Danish Systemic Risk Council (Det Systemiske Risikoråd). 2023. Activation of a sector-specific systemic risk buffer for corporate exposures to real estate companies. Recommendation.

Davis, E. P. and H. Zhu. 2011. Bank lending and commercial property cycles: Some cross-country evidence. In: Journal of International Money and Finance 30(1). 1–21.

Davydenko, S. A. 2007. When do firms default? A study of the default boundary. Mimeo.

Deghi, A., J. Mok and T. Tsuruga. 2021. Commercial Real Estate and Macrofinancial Stability During COVID-19. IMF Working Paper 2021/264.

Ellis, L. and C. Naughtin. 2010. Commercial Property and Financial Stability – An International Perspective. In: Reserve Bank of Australia Bulletin. June. 25–30.

ECB. 2024. Macroprudential Report – Technical Annex. June.

ESRB. 2016. Recommendation of the European Systemic Risk Board of 31 October 2016 on closing real estate data gaps (ESRB/2016/4).

ESRB. 2019. Recommendation of the European Systemic Risk Board of 21 March 2019 amending Recommendation ESRB/2016/14 on closing real estate data gaps (ESRB/2019/3).

ESRB. 2023. Vulnerabilities in the EEA commercial real estate sector. January.

ESRB. 2024. Adverse scenario for the 2024 European Insurance and Occupational Pensions Authority’s insurance sector stress test exercise.

Górnicka, L. and L. Valderrama. 2020. Stress Testing and Calibration of Macroprudential Policy Tools. IMF Working Paper 2020/165.

Guth, M., C. Lipp, C. Puhr and M. Schneider. 2020. Modeling the COVID-19 effects on the Austrian economy and banking system. In: Financial Stability Report 40. OeNB. 63–86.

Harrison, I. and C. Mathew. 2008. Project TUI: A structural approach to the understanding and measurement of residential mortgage lending risk. Reserve Bank of New Zealand.

Herring, R. J. and S. Wachter. 1999. Real Estate Booms and Banking Busts: An International Perspective. The Wharton School – Financial Institutions Centre Paper Nr. 99-27.

Liebeg, D. and M. Liegler. 2022. Systemic risks of commercial real estate funding in Austria. In: Financial Stability Report 44. OeNB. 45–67.

Puhr, C. and M. Schneider. 2021. Have mitigating measures helped prevent insolvencies in Austria amid the COVID-19 pandemic? In: Monetary Policy & The Economy Q4/20–Q1/21. OeNB. 77–110

Ryan, E., B. Jarmulska, G. De Nora, A. Fontana, A. Horan, J. H. Lang, M. Lo Duca, C. Moldovan and M. Rusnák. 2023. Real estate markets in an environment of high financing costs. In: ECB Financial Stability Review. November 2023.

25 Oesterreichische Nationalbank, Financial Stability and Macroprudential Supervision Division, marcel.barmeier@oenb.at, david.liebeg@oenb.at.

26 Financial Market Authority, Integrated Financial Markets Division, sebastian.roetzer@fma.gv.at.

27 Opinions expressed by the authors of studies do not necessarily reflect the official viewpoint of the Oesterreichische Nationalbank or the Eurosystem. The authors would like to thank Christian Wipf, Stefan W. Schmitz, Clemens Bonner and Markus Schwaiger (all OeNB) for helpful comments and valuable suggestions.

28 https://www.fdic.gov/analysis/quarterly-banking-profile/index.html

29 According to the FDIC, noncurrent loans are those that are 90 or more days past due or are on nonaccrual status. The noncurrent rate is the sum of noncurrent loans to total loans.

30 https://app.bde.es/bie_www/faces/bie_wwwias/jsp/op/Home/pHome.jsp

31 Note that for the monitoring of CRE-related risks, we focus on domestic and foreign loans to legal persons (including nonfinancial corporations) in the sectors L.68, F.41 and F.43 as well as loans that fund the constructing, developing or acquiring of (commercial or residential) properties. For an overview of the exposure and risk structure based on this definition, see Financial Stability Report 47.

32 Statistics Austria: i nput-output statistics .

33 In German: gemeinnützige Bauvereinigungen.

34 Note that the NPL ratio is based on data from AnaCredit and uses a sectoral perspective to define CRE. This differs from the CRE definition based on FINREP data.

35 Unfortunately, balance sheet information in Bureau van Dijk SABINA database is available with a significant time lag. Thus, we rely on data as of year-end 2021.

36 CRE lending differs from RRE insofar as both lenders and borrowers are heterogeneous and the latter may have multiple loans outstanding. Therefore, we depart from the representative bank approach employed in prior literature and instead simulate and resolve borrower defaults for the entire network of CRE loans in the Austrian banking system.

37 The overindebtedness threshold is justified by cross-country empirical studies that show that the equity ratio commonly associated with insolvency ranges from –30% to –35% (see Davydenko, 2007). The foundation for the illiquidity threshold is weaker. As by Puhr and Schneider (2021), we use a negative liquidity threshold to account for the firms’ possibility to rely on undrawn credit lines from banks.

38 A firm experiencing solvency and/or liquidity stress will attempt to rebalance its accounts by selling assets. If the proceeds from a sale are deemed insufficient to cover the firm’s needs, it will default on its obligations, thereby appearing as default in the lending bank’s loan books.

39 The underlying scenarios are based on the methodology of the European Central Bank that takes overvaluations in real estate markets into account.

40 Note that the ESRB (2024) scenario allows for more severe assumptions about risk-free rates (+119 basis points for ten-year Austrian government bond yields) and corporate shocks (+288 basis points for BBB rated exposures and +435 basis points for BB rated exposures) as well as property prices (–33.1% for RRE and –43.6% for CRE over a three-year horizon). In the global financial crisis, Ireland experienced an RRE price drop of 34.2% and a CRE price drop of 56.3% within two years (Ellis and Naughtin, 2010).

41 Total real estate loans and loans for construction and development loans, respectively.

42 Loans to the NACE sectors “construction” and “real estate-related activities,” respectively.

43 OCD is calculated as the sum of the overall capital requirement (Total SREP capital requirement + combined buffer requirements) and the Pillar 2 guidance.

44 For details, refer to the FMSB website .

From part of the problem to part of the solution: evaluating the effectiveness of borrower-based measures in Austria

Marcel Barmeier, Selin Johanna Scheuerer 45

Evaluating macroprudential policies is key to ensuring that measures are implemented effectively. Borrower-based measures were introduced in Austria in August 2022 via the so-called KIM-V regulation that defines sustainable lending standards for residential real estate (RRE) financing. In our evaluation, we provide evidence on how effective these measures have been so far in addressing systemic risks in Austria’s RRE sector. Based on data for lending standards, we find that the KIM-V has halved the share of new lending with a debt service-to-income ratio (DSTI) above 40%. In addition, by applying estimations in a difference-in-differences setting, we find that the ratio of nonperforming loans (NPLs) of RRE loans has decreased by up to 0.5 percentage points since mid-2022. Our findings support the literature, which shows that borrower-based measures effectively reduce systemic risks in the housing sector.

JEL classification: G21, G28, R31

Keywords: borrower-based measures, KIM-V, financial stability, residential real estate

From 2016 onward, the Austrian Financial Market Stability Board repeatedly highlighted the importance of adhering to sustainable standards in real estate lending. When these recommendations did not have the intended effects and systemic risks from residential real estate (RRE) financing kept building up, it was decided to implement legally binding borrower-based measures (BBMs) in August 2022, known as the KIM-V regulation (“Kreditinstitute-Immobilienfinanzierungsmaßnahmen-Verordnung”). The KIM-V sets standards for credit institutions’ new lending by limiting (1) the loan-to-collateral ratio (LTC) to 90%, (2) the debt service-to-income ratio (DSTI) to 40% and (3) the maturity to 35 years. 46 Given that BBMs have been in place for over two years now, the question arises how effective they have been so far in mitigating risks to financial stability. To provide an answer, we draw on national and international bank-level data in a difference-in-differences framework. The study is structured as follows: In section 1, we discuss the link between BBMs and financial stability, section 2 includes a short literature review, section 3 provides an empirical approach to estimating the effectiveness of the KIM-V and section 4 concludes.

1 Borrower-based measures and financial stability

Poor lending standards in RRE financing increase the likelihood and severity of disruptions to financial stability, i.e. systemic banking crises, as underlined e.g. by Aikman et al. (2021) and Muellbauer (2022). The quality of new loans trickles down to the quality of a bank’s RRE portfolio, which constitutes a significant share of banks’ domestic credit exposure; in Austria approximately 30%. 47 Given that mortgages play such an important role for banks, housing market turmoil and banking crises often go hand in hand (Jordà et al., 2016). Two-thirds of 46 systemic banking crises for which house price data are available were preceded by housing boom-bust cycles (Crowe et al., 2013; Roy, 2022). Systemic banking crises imply high social and economic costs: the public sector on average pays 6.7% of GDP to fight such crises, public debt rises by 21% of GDP and output losses are roughly 35% of GDP (Laeven and Valencia, 2018). 48 To reduce the risks of a real estate-related banking crisis, BBMs became the most commonly used macroprudential policy tool in Europe: 22 out of 30 countries in the European Economic Area deploy BBMs. The most common BBMs are income- and collateral-based measures. Income-based measures, such as limits to the DSTI or the debt-to-income ratio (DTI), aim at increasing household resilience to income and interest rate shocks. In times of crisis, borrowers have more income at their disposal to cover their regular expenses, which lowers the household’s probability of default (PD). Collateral-based measures, like the LTC, aim at improving lender resilience during real estate downturns by requiring higher down payments. If a household defaults on its debt, the bank’s loss given default (LGD) is reduced (Lo Duca et al., 2023).

2 Literature review

A major challenge in evaluating BBMs is how to define the target variable for measuring effectiveness. Financial stability is difficult to define in an implementable way (BIS, 2023). Thus, policymakers commonly target specific intermediate objectives, which can be broken down into (1) maintaining borrower resilience, (2) maintaining lender resilience, (3) dampening the housing credit cycle and (4) promoting sustainable house price growth (BIS, 2023). Since many authorities mandated with assessing systemic risks, including the Oesterreichische Nationalbank, focus on the first two objectives, the following literature review covers the impact of BBMs on borrower and lender resilience.

To measure borrower resilience, the target variable is often a single credit risk indicator, e.g. the PD, which is regressed on loan and borrower characteristics. Examples include de Haan and Mastrogiacomo (2020), who find that in Denmark limits to the loan-to-value ratio (LTV) and DSTI reduce the probability of nonperformance of loans, which encompasses arrears, foreclosures and defaults. If the LTV (DSTI) is 10 percentage points higher, the probability of nonperformance of loans increases by 0.19 (0.75) percentage points. Galán and Lamas (2019) corroborate the main insights for Spain, emphasizing that income-based measures are more robust determinants of mortgage default than LTV limits. Nier et al. (2019) show for Romania that if the DSTI limit of 40% had been implemented earlier, the PD would have been lowered by approximately 23% in comparison to the case without BBMs. Catapeno et al. (2021) rely on an agent-based model to assess the effectiveness of potential BBMs in Italy. They acknowledge that BBMs reduce the probability of mortgage default but find negligible effects for the Italian market. The TUI 49 -model developed by Górnicka and Valderrama (2020) is another method to estimate effects on credit risk indicators. It was for instance applied to Switzerland (Maslova et al., 2022) and Austria (Górnicka and Valderrama, 2020) to measure the effectiveness of various theoretical DSTI, DTI and LTV limits. For Austria, the PD decreased from 3.9% to 2.2% in an adverse macroeconomic scenario thanks to a DSTI limit of 40% combined with an LTV limit of 80%.

With respect to lender resilience, the literature directs attention to risk measures on the level of individual institutions. Gross and Población (2017) developed a structural micro-macro model which combines household information of the Household Finance and Consumption Survey with macroeconomic and bank-level data. Household resilience is indicated by PD, LGD as well as the expected loss. Any change in these variables subsequently affects banks’ capital position via the mortgage portfolios. The model has been applied in a cross-country context, e.g. by Giannoulakis et al. (2023) or Ampudia et al. (2021), but also for individual countries, e.g. Slovakia (Jurča et al., 2020). Giannoulakis et al. (2023) find that the median capital ratio across countries implementing BBMs increases by up to 1 percentage point compared to no policy intervention. Some researchers construct their own bank-level risk measures to evaluate the impact of BBMs. The target variables are typically based on data from stock markets as well as banks’ financial statements. Meuleman and Vander Vennet (2020) distinguish between individual bank risk and risk from the linkage with the financial system. They find that BBMs are most effective in lowering banks’ individual risk. In other words, a unit increase (tightening) of their self-constructed index for BBMs on average reduces risk by 4.2 percentage points. They do not find a statistically significant effect on the linkage component. Belkhir et al. (2023) add that DSTI and LTV limits are effective in reducing banks’ expected capital shortage in a crisis – but only in combination with an inflation-targeting regime. Altunbas et al. (2018) find that asset class measures, which encompass DSTI, LTV and credit growth limits and limits on the exposure to the housing sector, reduce (increase) the expected default frequency 50 for the average bank by 0.15 (0.66) percentage points when tightened (eased).

To summarize, the literature finds support for the effectiveness of BBMs in addressing systemic risks, measured by indicators evaluating borrower and lender resilience. This gives authorities well-founded arguments to apply BBMs. However, as national specificities play an important role for the effectiveness of BBMs, national characteristics should be considered.

3 Effectiveness of BBMs in Austria

To add to the understanding of BBMs in Austria, we evaluate their effectiveness in a two-step approach. First, we present descriptive statistics for the development of lending standards and the NPL ratio for RRE financing. Second, we use these data in a difference-in-differences setting to estimate the effect of the introduction of the KIM-V on the NPL ratio. Regarding the target variable, we contribute to the literature on the effectiveness of BBMs with respect to borrower resilience. 51 De Haan and Mastrogiacomo (2020) as well as Galán and Lamas (2019) are the papers which bear the most resemblance to ours.

3.1 Data

To conduct our analysis, we compare Austrian and German bank-level data on lending standards and the quality of the RRE loan portfolio, i.e. the NPL ratio.

This is Chart 1.
This is Chart 2.

In Austria, data on lending standards, i.e. on DSTI, DTI, LTV, LTC and maturity, are available from 2011 onward. From 2011 until 2020, banks reported their lending standards as part of the “Hypothekarkreditumfrage” (HKU) 52 . Starting from 2020, reporting standards were amended and reporting via “VERA H – Private Wohnimmobilienfinanzierung” (VERA-H) 53 related to RRE lending became legally binding. For Germany, we rely on data that are provided by a loan brokerage platform to the Deutsche Bundesbank (Ausschuss für Finanzstabilität, 2024). As the DSTI is the most relevant indicator for debtors’ ability to repay their loans, we discuss its development in more detail. 54

Chart 1 shows the average volume-weighted DSTI for new lending in Germany and Austria. Although interest rates increased gradually from July 2022, the average DSTI for new lending in Austria remained below 30%, whereas in Germany the DSTI peaked at 33.5% in the first half of 2022. While the average DSTI in Austria also increased slightly between the first half of 2022 and the second half of 2023, the reduction in the share of loans with a DSTI above 40% dampened the overall increase in the DSTI (chart 2). In the first half of 2022, 16% of the new lending volume was issued with a DSTI above 40%; in the second half of 2023, this percentage dropped to 8%. The improvement of the DSTI and other lending standards (see the annex) is a first indication of the effectiveness of the KIM-V.

This is Chart 3

To gauge the loan quality of banks’ RRE portfolio, we consider the NPL ratio for RRE loans. The NPL ratio is corrected for loans that are past due more than one year. 55 Chart 3 shows the development of the median corrected NPL ratio on an unconsolidated level for significant institutions from Germany and Austria since June 2020. 56 While in Austria and Germany the NPL ratio remained relatively constant up to the introduction of the KIM-V, the NPL ratio increased in Germany from mid-2022 onward, namely from 0.3% to 0.5% in March 2024. In Austria, the NPL ratio stood at 1.1% in June 2022 and March 2024.

When evaluating the effect of the KIM-V on the NPL ratio, we need to consider that improved lending standards do not immediately reduce defaults in the stock. 57 Thus, the direct increase in the NPL ratio recorded by German vs. Austrian banks should be considered as part of general fluctuations. Only the persistent increase in the NPL ratio in Germany relative to Austria might be attributable to the KIM-V. Since other confounding factors might have played a role, we continue our analysis with an econometric approach to estimate the causal relationship between the introduction of the KIM-V in Austria and the evolution of nonperforming loans.

3.2 Empirical strategy

Estimating the causal effect of the introduction of BBMs in Austria is challenging. Ideally, we would randomly allocate banks to a group that has to fulfill the requirements for new lending according to the KIM-V (treatment group) and a group that does not have to fulfill the requirements (control group). However, as the KIM-V targets all banks in Austria, we need to find other methods for estimating the impact. Thus, we rely on a difference-in-differences approach, where we compare the NPL ratio of banks that are treated by the KIM-V and banks that are not treated by it.

This is Chart 4.

Since the assignment of banks into a treatment or control group is crucial, we rely on two alternative approaches. First, as the KIM-V was introduced only in Austria, we draw on bank-level data from Germany to build a control group (baseline specification). 58 As the banking sectors in Austria and Germany are alike (e.g. high degree of bank competition, large number of banks), German banks are most suitable to serve as a control group when we estimate the effects of the implementation of the KIM-V in Austria.

Second, we classify Austrian banks into a treatment and a control group based on their standards for new lending prior to the introduction of the KIM-V (robustness specification). 59 Chart 4 shows the distribution of Austrian banks with respect to their share of new lending with a DSTI above 40% in 2020 and 2021, i.e. before the KIM-V was introduced. The control group comprises banks that had a below-median share of new lending with a DSTI above 40%, while banks with an above-median share of new lending with a DSTI above 40% make up the treatment group. 60 Given the numerous exemptions to the KIM-V, banks with a low share of new lending with a DSTI above 40% did not need to change their lending standards significantly once BBMs were introduced. 61

The econometric validity of the difference-in-differences approach rests on critical assumptions. Most importantly, the method assumes that the NPL ratios of treated and non-treated banks have parallel trends in the absence of the KIM-V (“parallel trends assumption”). While this is generally not testable, the pre-KIM-V trends provide an indication. Chart 3 shows for the baseline specification that the evolution of the median NPL ratios were fairly parallel for banks from Austria and Germany before the treatment. In June 2020 and June 2022, the median NPL ratios corresponded for banks in both countries, with the German NPL ratio standing at 0.3% and the Austrian one at 1.1%. As a further assumption for the difference-in-differences approach, the composition of the control and the treatment group should not change over time (“time-invariant composition assumption”). This assumption would be violated in the baseline specification if banks endogenously changed their headquarters between Austria and Germany in response to the introduction of the KIM-V. However, this has not been observed in the Austrian and German banking markets.

To estimate the effects of BBMs on the credit quality in Austria, the following two-way fixed effects model in its baseline specification will be estimated: 62

〖NPL ratio〗_(i,j,t)  =β_1  〖BBM〗_i× 〖Time〗_t+ γX_(i,t-1)+δ_t+ η_j+ ε_(i,j,t)    (1) (1)

where NPL ratioi,j,t is the corrected NPL ratio for RRE loans 63 of bank i in country j at time t, BBMi is a dummy variable that is 1 if the bank is in the treatment group and 0 otherwise, Timet is a dummy variable that is 0 before the introduction of the KIM-V and 1 afterward and Xi,t–1 refers to lagged control variables on the bank level. As suggested by Manz (2019), we include the common equity tier 1 (CET1) ratio, the return on assets (ROA) ratio and the overall NPL ratio in the estimation. 64 Bank variables are lagged by one quarter to control for potential endogeneity between control variables and the NPL ratio. δ_t and η_j are time- and country-fixed effects, respectively. Ideally, we would also control for heterogeneity on the bank level by applying bank-fixed effects. In addition, we would control for bank-specific reactions to changes in macroeconomic variables (e.g. interest rate) via bank-time fixed effects. However, given the small sample size, either is infeasible. 65 The estimated coefficient of interest is β ̂_1.β ̂_1<0 would indicate that the introduction of the KIM-V in Austria reduced the NPL ratio compared to the case where no BBMs were in place.

3.3 Results

Estimation results for evaluating the effectiveness of BBMs with respect to their impact on NPLs are shown in table 1. As discussed in section 3.2, two alternative empirical strategies are executed with respect to assigning banks to a treatment and a control group.

Columns (1) and (2) show that the introduction of BBMs in Austria is associated with a 0.5-percentage-point decrease in the NPL ratio of Austrian banks compared to German banks. With respect to the robustness specification, we find that the KIM-V reduced the NPL ratio of Austrian banks that were relatively more exposed to the regulation by 0.1 percentage points compared to Austrian banks that were relatively less exposed (columns (3) and (4)). The results are confirmed when we consider bank control variables.

The results need to be interpreted with caution. While we are confident that the KIM-V reduced the NPL ratio for RRE loans, the magnitude is of greater uncertainty. This is shown by the relatively large difference between the estimated coefficients in the baseline and robustness specifications, which indicates the importance of choosing an appropriate control group. Furthermore, a reduction of the NPL ratio in the range of 0.1 to 0.5 percentage points may appear small. However, given that the KIM-V has only addressed a portion of the RRE loan volume currently outstanding 66 , this would translate into a significantly lower NPL ratio for loans granted since August 2022. When we factor in the unfavorable macroeconomic developments since then (e.g. rising interest rates), it could seem unrealistic for RRE loans granted since mid-2022 to have a very low NPL ratio.

Table 1: Estimation results for the effectiveness of the KIM-V  
Baseline specification Robustness specification
Dependent variable NPL ratio NPL ratio
(1) (2) (3) (4)
BBM x Time –0.0046** –0.0045** –0.0008** –0.0013*
(0.00002) (0.00006) (0.00001) (0.00010)
Bank controls No Yes No Yes
Time-fixed effects Yes Yes Yes Yes
Country-/group-fixed
effects
Yes Yes Yes Yes
Observations 1,472 1,371 1,222 1,083
R2 0.05263 0.17945 0.04423 0.31703
Source: OeNB.
Note: Clustered standard errors are in parentheses. Significance codes: *** = 0.01, ** = 0.05, * = 0.1.

4 Concluding remarks

Borrower-based measures in Austria have been effective. Combining evidence from descriptive statistics on the development of lending standards with an empirical approach to estimate the effect on borrower resilience, we find that the BBMs have reduced systemic risks in the residential real estate market. Standards for new RRE loans have improved significantly since the KIM-V was introduced in mid-2022, while NPL ratios for RRE lending have remained relatively stable. Deploying a difference-in-differences approach to empirically evaluate the effectiveness of BBMs, we find that the introduction of the KIM-V reduced the NPL ratio of Austrian banks by up to 0.5 percentage points compared to a control group.

Given that BBMs such as the KIM-V address only the new lending volume, it can take many years for their full effect to unfold with respect to borrower and lender resilience. Many member countries in the Single Supervisory Mechanism therefore regard BBMs as a structural measure in the nature of a backstop (Lang et al., 2022).

References

Aikman, D., R. Kelly, F. McCann and F. Yao. 2021. The macroeconomic channels of macroprudential mortgage policies. Central Bank of Ireland. Financial Stability Notes 2021(11).

Altunbas, Y., M. Binici and L. Gambacorta. 2018. Macroprudential policy and bank risk. In: Journal of International Money and Finance 81. 203–220.

Ampudia, M., M. Lo Duca, M. Farkas, G. Pérez-Quirós, M. Pirovano, G. Rünstler and E. Tereanu. 2021. On the effectiveness of macroprudential policy. ECB Working Paper Series 2559/2021.

Ausschuss für Finanzstabilität. 2024. Elfter Bericht an den Deutschen Bundestag zur Finanzstabilität in Deutschland.

Belkhir, M., S. Ben Naceur, B. Candelon, W. Gyu Choi and F. Mugrabi. 2023. Macroprudential policy and bank systemic risk: does inflation targeting matter? IMF Working Paper Series 23/119.

BIS. 2023. Macroprudential policies to mitigate housing market risks. CGFS Papers No 69. December.

Catapeno, G., F. Franceschi, M. Loberto and V. Michelangeli. 2021. Macroprudential Policy Analysis via an Agent Based Model of the Real Estate Sector. Bank of Italy. Termi di Discussione (Working Papers) 1338.

Crowe, C., G. Dell’Ariccia, D. Igan and P. Rabanal. 2013. How to Deal with Real Estate Booms: Lessons from Country Experiences. In: Journal of Financial Stability 9(3). 300–319.

De Haan, L. and M. Mastrogiacomo. 2020. Loan to Value Caps and Government-Backed Mortgage Insurance: Loan-L