Machine learning (ML) could strengthen banks’ resilience through improved credit risk screening and ultimately benefit financial stability. Yet, ML adoption in banking remains limited, with simpler linear models still predominating. We argue that the emergence of highly flexible ML models has created a new challenge for forecasting tasks: ‘model multiplicity’—where equally accurate ML models at the aggregate level produce divergent individual-level predictions (‘predictive multiplicity’) or differ in their decision surfaces (‘procedural multiplicity’). These issues raise fundamental questions: Why should an individual or firm be subject to an adverse credit risk model outcome when there is an equally accurate model that treats them more favorably? Using the world’s largest loss database of corporate defaults, we examine these two phenomena in recovery rate ( RR ) modeling and propose heterogeneous ML ensembles as a natural solution. By combining predictions and decision surfaces from multiple well-performing ML models, ensembles mitigate risks associated with predictive multiplicity by ensuring that borrowers are not subject to the fluctuations of a single model, and reduce procedural multiplicity by providing a robust measure of features that ultimately improve out-of-sample RR predictions. By addressing the ‘multiplicity of good models’ problem, our study emphasizes the importance of model stability and provides new insights for the future development of ML models.

This study constructs a financial stress index for Bulgaria, the Czech Republic, Hungary, Poland, and Russia and examines the relationship between financial stress and economic activity. The financial stress index incorporates banking sector fragility, time varying stock market return volatility, sovereign debt spreads, an exchange market pressure index, and trade credit. These variables seem to capture key aspects of financial stress in sample countries as the index peaks at known financial crises in these countries. We then examine the relationship between financial stress and economic activity. Impulse response functions based on bivariate VARs show a significant relationship between financial stress and some measures of economic activity. Overall, the constructed financial stress index provides valuable information on the state of the economy and economic activity.

Regulatory forbearance in times of corporate distress has been a common practice in many countries to achieve bank stability, particularly so in the absence of a unified bankruptcy code, yet very little is known in the context of emerging market economies. Exploiting variation of membership across banks in a corporate debt restructuring programme (CDR) sponsored by the central bank in India, this paper finds that the banks that made use of regulatory forbearance (RF) on the restructured corporate loans could increase their stability significantly due to the extension of low provisioning on restructured loans. However, the positive effect of RF diminishes at higher levels of market power, highlighting that member banks with higher market power tend to originate riskier assets (as reflected in their risk-weighted assets) under the auspices of this programme. Our results remain robust to different estimators (including propensity score matching), ownership structure, and alternative measures of bank stability.

We analyze the relationship between the stance of monetary policy and the implicit risk aversion in European Stock market prices in an international open-economy framework. We use a structural vector autoregression (SVAR) model that incorporates the effect of a factor that reflects the global monetary policy stance. We use shocks in the US Fed monetary policy stance as a proxy of this global factor. Our results indicate mixed evidence depending on whether simultaneity between domestic monetary policy stance and the stock market behavior is taken into full account. When this simultaneity is not allowed we confirm previous evidence found in the literature, extended to the international field: a lax monetary policy, both domestic and global, decreases risk aversion. However, when we take this into account, results indicate that a lax monetary policy increase in the short-run the risk aversion of the domestic representative investor.

We exploit staggered real estate transaction tax (RETT) hikes across German states to identify the effect on the growth rates of regional house prices and outstanding mortgage loans by all local German banks. The results show that a RETT hike by one percentage point reduces regional house prices by 3%–4%. Furthermore, IV-regressions yield that a 1 percentage point drop in regional house prices induced by a RETT increase leads to a 0.3% decline in regional mortgage lending, particularly among low-capitalized banks in rural regions.

We investigate households’ bank switching exploiting a unique representative dataset from Bank of Italy Survey on Household Income and Wealth that follows the households and their bank(s) over time. First, we document that bank switching is quite prevalent in our sample, with almost a quarter of households changing their main bank in a biannual horizon. Next, we relate the decision to switch to the features and dynamics of household bank relationship, and to the characteristics of both households and banks. In line with the switching cost theory, we find that using more than a single bank, as well as the intensity (number of services used), and the scope (bank services used) of the relationship with the main bank play a role in shaping the households’ decision to switch. Moreover, bank switching is strongly and positively correlated with both taking out and having paid off a mortgage.

We investigate whether and how political systems affect the financial soundness of conventional and Islamic banks. Using factors extracted from principal component analysis, we find that Islamic banks underperform their conventional counterparts in more democratic political systems but outperform them in hybrid and Sharia’a-based legal systems. The findings reflect the challenges Islamic banks face in Western countries in terms of perception, financial infrastructure, and regulatory constraints while mirroring the recognition of their specificities and their cultural and religious compliance with Sharia’a law in Muslim countries. The findings are robust to a battery of alternative estimation techniques and methods of correcting standard errors.

We develop an empirically based simulation study to test two types of policies designed to control systemic risk: preventive policies targeting capital requirements and mitigation policies targeting default resolution. We find that capital buffers reduce both the number of defaults and the resulting losses. The loss reduction benefit increases as the magnitude of adverse shocks becomes higher. We find that a simple branch-breakup resolution strategy reduces the loss borne by the Federal Deposit Insurance Corporation (FDIC). The mitigation effect becomes higher as the fraction of assets resolved through auctions and auction competitiveness increase.

We investigate whether or not monetary aggregates are important in determining output. In addition to the official Simple Sum measure of money, we employ the sophisticated weighted Divisia aggregate. We also investigate whether or not the influence of money on output is time varying using data-driven procedures to identify breaks in the data and conduct estimations for the different segments defined by these breaks. We find that structural breaks do exist in some of the variables under investigation and these do influence the relationship between monetary aggregates and output. However, the official Simple Sum aggregate appears to be more affected by the breaks than the theoretically superior Divisia aggregate. In particular, our results show that in some segments of our data, the Simple Sum aggregate does not influence output significantly whereas the Divisia aggregate maintains a significant relationship with output in all segments. We conclude that Divisia money is still influencing output in spite of the diminished role played in monetary policy. Our investigation also suggests that the recovery from the financial crisis using quantitative easing would have been faster if money was not being hoarded.

We examine the impact of the ECB’s asset purchase programmes on euro area non-financial firms’ cost of borrowing and their choice between corporate bonds and syndicated loans. Our findings indicate that the Corporate Sector Purchase Programme (CSPP) reduced spreads for both eligible and non-eligible corporate bonds, and that ECB purchases of covered bonds positively affected corporate bond spreads. The CSPP also compressed spreads across all syndicated loans, irrespective of eligibility. We find evidence supporting a “cost of borrowing channel” for covered bonds under the first programme and asset-backed securities, indicating that syndicated loan spreads reflect banks’ borrowing costs in the bond market. Additionally, our results reveal that the CSPP significantly influenced firms’ debt financing choices, with these effects being more pronounced for non-switchers.