Stress Testing European Banks’ Transition Risk and Lending Behavior

How the climate stress test changes loan pricing and allocation

Non-Technical Summary

Banks play a central role in the transition to a low-carbon economy. They finance investment in cleaner technologies, but they are also exposed to losses if climate policies, technological change, or shifts in consumer preferences reduce the value of carbon-intensive activities. Supervisors have therefore introduced climate stress tests to assess banks’ preparedness, improve data and modelling, and identify vulnerabilities. Although these exercises were presented as learning tools rather than mechanisms that directly affect capital requirements, the information they generate and the accompanying supervisory scrutiny may nonetheless influence lending decisions.

The paper studies the ECB’s 2021-22 climate stress test and asks whether participating banks changed the way they price and allocate new loans according to borrowers’ greenhouse-gas intensity. The analysis combines more than two million observations on newly originated loans from the Eurosystem’s AnaCredit register with a firm-level emissions-intensity measure. It compares significant banks that participated in the ECB exercise with smaller non-participating institutions, before and after the stress test, while accounting for the earlier French pilot exercise and controlling for differences across banks, firms, loan products, and time.

The headline result is that the average sensitivity of loan spreads to carbon intensity changes very little following the stress test. Taken in isolation, this near-zero effect could suggest that the exercise had little influence. However, the decomposition developed in the paper reveals a different picture. Banks subject to the stress test do not systematically reprice continuing bank-firm relationships. Instead, they adjust primarily through two channels that work in opposite but economically consistent directions. First, they reallocate new lending within their existing customer base away from more carbon-intensive firms. Second, they apply more carbon-sensitive pricing to relationships entering or leaving their portfolios. For new borrowers, a one-standard-deviation increase in carbon intensity is associated with loan spreads that are approximately 10 basis points higher. These effects are stronger among the highest-emitting firms.

The figure illustrates why the overall coefficient is close to zero. The negative contribution from reallocating credit among incumbent borrowers is largely offset by the positive contribution from entry-and-exit pricing, while repricing within continuing relationships and changes in borrower composition play only a minor role. Specifications that account for the different timing of the French and ECB exercises yield more moderate estimates but continue to identify entry-and-exit pricing as the main adjustment channel. Among significant banks, which were required to participate in the stress test, involvement in the most demanding bottom-up module does not generate a noticeably stronger response than participation in the lighter modules.

Overall, the results suggest that climate stress tests can influence bank behaviour even without an explicit ‘brown’ capital surcharge. Their main effect is not a uniform increase in borrowing costs for all carbon-intensive firms, but rather a gradual reallocation of lending portfolios combined with more risk-sensitive pricing at the margins of bank-firm relationships. This pattern is consistent with an information-and-scrutiny channel: the exercise makes transition risk more salient and encourages banks to reconsider which firms they finance. The findings should nevertheless be interpreted with caution because firms’ emissions are estimated rather than directly observed. Although other climate-related supervisory measures were introduced over the same period, the analysis carefully controls for these concurrent policies. A wide range of robustness checks, based on state-of-the-art econometric method, confirms the main results.

For supervisors, the results imply that stress tests should provide sufficiently granular information to influence both loan pricing and portfolio allocation, and that their effects should be assessed using more than average loan spreads alone. The paper provides a clear empirical framework for assessing the impact of prudential regulation at the granular level.

Keywords: Banking, Climate Stress Tests, Transition Risk, Credit Register, Loan Pricing, Carbon Intensity
Codes JEL : C23, E51, E58, G21, G28, G32, Q54