Do Banks Practice What They Preach? Brown Lending and Environmental Disclosure in the Euro Area

The latest report of the Global Sustainable Investment Alliance (2021) discloses that sustainable investments have reached an impressive USD 35 trillion globally (in North America, Europe, Japan and Australasia). As a result, banks are increasingly keen on attracting these types of investments.

ESG rating agencies rely heavily on corporate disclosures within their methodologies to compute ESG ratings. See for instance the methodology employed by MSCI (https://​www.​msci.​com/​documents/​1296102/​21901542/​MSCI+ESG+Ratings​+Brochure-cbr-en.​pdf) and by Standard & Poor’s (https://​www.​spglobal.​com/​esg/​documents/​sp-global-esg-scores-methodology-2022.​pdf).

For further information on the use of expert-validated disclosure indexes and measures of disclosure tone, see Altunbas et al. (2022) and Loughran and McDonald (2011).

This third pillar requires a set of disclosure requirements that “allow market participants to assess key pieces of information on the scope of application, capital, risk exposures, risk assessment processes, and hence the capital adequacy of the institution.” (Basel Committee on Banking Supervision (2004), pp. 187).

See https://​www.​ngfs.​net/​en.

Central banks worldwide have started to address climate-related issues, particularly related to financial risks and financial stability (Campiglio et al. 2018). An important example is the ECB that is exploring ways to fight pollution and climate change, including taking actions in areas such as banking supervision and financial stability (see https://​www.​ecb.​europa.​eu/​ecb/​orga/​climate/​html/​index.​en.​html)

Although banks can still be in a safe position by lending to brown borrowers if they correctly charge higher interest rates or impose more restrictive contractual terms to account for higher levels of ‘brown risk’, the methodological approaches to estimate it are not yet corroborated nor there is a common methodological framework widely employed in the industry, mainly because they are new emerging risks that are still characterized by higher levels of uncertainty compared to traditional financial risks and, consequently there is a high degree of disagreement in their computation (Berg et al. 2022). In this regard, the literature has shown that banks generally underestimate carbon risk, potentially resulting in a less stable financial position for environmentally irresponsible banks (Ehlers et al. 2022).

While we cannot completely rule out the hypothesis that polluting brown borrowers might decide to borrow less from environmentally concerned banks, fearing that lending from such banks could be tightened in the near future, we use multiple bank relationships and introduce borrower fixed effects to account for firm credit demand to address this concern.

We are grateful to the climate stress-test experts in the Stress Test Modelling Division of the ECB for the validation of the dictionary.

Detailed documentation about AnaCredit can be found here: https://​www.​ecb.​europa.​eu/​stats/​money_​credit_​banking/​anacredit/​html/​index.​en.​html

Based on the 1997 Kyoto Protocol, seven GHG were considered: (a) carbon dioxide; (b) methane; (c) nitrous oxide; (d) hydrofluorocarbons; (e) per-fluorocarbons; (f) Sulphur hexafluoride; and (g) nitrogen trifluoride.

Given the skewness of the distribution, there might be some doubts that the results of our analysis might be driven by extreme cases. To rule out this hypothesis, we have checked that (i) those banks with the highest levels of disclosure do not have a rather negative tone and that (ii) those banks with the lowest levels of disclosure do not have a rather positive tone. Specifically, amongst the top-five disclosers, none of them is amongst the top-five negative tone users, and amongst the bottom-five disclosers, none of them is amongst the top-five positive tone users. This finding confirms that the results of our analysis are not driven by extreme cases of word distributions.

As a robustness test, we re-ran our baseline regressions by using the same deflator (total assets) for all independent variables. The (unreported) results are qualitatively unchanged and available from the authors on request.

In addition, the analysis conducted in paragraph 4.2 shows that, although in general we reject the ‘environmental window dressing’ hypothesis, we do find evidence of ‘environmental window dressing’ behaviour by those banks adopting a positive tone in their disclosures.

An F-test for joint significance confirms that the difference between the two point estimates (≈ 7.5%) is statistically different from zero.