We partially agree with the risk-based approach adopted by the EBA. There are a lot of aspects that should be considered when introducing ESG components within IRB models, ranging from modeling to aspects more strictly related to the goal to be achieved.
As already explained in this response to the DP, Intesa Sanpaolo is already including certain ESG risks in its internal models, however we believe there are several limitations in pursuing only an approach based on the enhancement of the existing Pillar 1 framework. In fact:
• the one-year time horizon of the PD parameter may not be adequate to reflect environmental risks that for their non-linear nature have a long-term time horizon and that could take decades before materialising. From this point of view introducing this uncertainty within historical samples is very challenging because most of the time the manifestation of a long-term environmental risk can exceed the duration of an economic cycle which is the base for calibrating rating models.
• As of today, historical data about environmental risks are difficult to retrieve. Historical databases are usually not constantly updated and have a low degree of accuracy.
• The rarity/unpredictability of these events and the difficult interpretability of data about customers environmental policy implementation, may not lead to robust results when integrated within the risk differentiation phase with the consequent low level of accuracy and importance within the rating model. This effect, on one side, might not be sufficient to encourage banks to sponsor the transition to a more sustainable business structure and on the other side might not lead to adequate effect on prudential treatment of exposures.
If the EBA objective is to ensure an adequate relevance of ESG features, actions with a more significant effect on own funds requirements should be undertaken. Where to address this intervention should be assessed depending on internal processes frameworks, that could differ from one bank to another. Regarding to this, acting downstream of the rating model could give a higher degree of flexibility relying on process aspects like, for example, override actions on the base on a stand-alone assessment of ESG characteristics (e.g. in the short-term the override might be a good solution to incorporate ESG factors, however if during the validation phase the Competent Authority requests to incorporate them in the estimation or to introduce a margin of conservativism, the bank would not be incentivised to introduce this type of override).
The analysis of ESG risk factors in credit risk models is an important element in the overall risk management framework and their relevance is bound to assume an increasing importance in the years to come. For this reason, at the beginning of 2009, with the introduction of the first FIRB corporate rating models, the evaluation of ESG risk factors was incorporated by Intesa Sanpaolo in the “qualitative” component of the PD models in an expert-based way. In particular, given the absence of readily available ESG data at the time (both at counterparty and aggregated level) that could be used to statistically analyze the robustness of correlation with credit risk, the ESG risks were incorporated in the rating models through a set of specific questions to be answered in the qualitative assessment section.
A practical consequence of the above choice was that all the answers provided by the rating analysts could be stored to accumulate a sufficiently long time series of ESG risk factors assessments for the statistical analysis. Indeed, one of the major challenges in ESG risks analysis is data availability and its quality: it could be difficult to observe an actual correlation with the defaults in the past and the available data could be insufficient to produce robust conclusions. Unexpectedly, it is possible that the most environmentally unsustainable corporates have always been the best payers, given their high profitability. This is also since the ESG risk factors are bound to intensify in the future, so have more forward-looking nature.
In general, among the sources of uncertainty of the ESG risk factors modelling one should mention the following points:
• Obligor-level data availability, reliability, comparability and coverage, especially for the smaller obligors or for the obligors belonging to some specific economic sectors or geographical areas (e.g. emerging markets);
• Absence of a complete and uniformly acknowledged taxonomy necessary to identify which corporates can be defined as “green” and which should be considered “environmentally unsustainable”;
• The level with respect to which the ESG risks should be assessed (e.g., counterparty vs. asset/loan level); moreover, some ESG risks could be more relevant for large corporates with sophisticated business structure covering multiple economic sectors and industries (therefore, the same counterparty could be simultaneously considered “green” and “environmentally unsustainable”, depending on the branch of business under evaluation);
The subsequent releases of corporate rating models, occurred in 2010 (transition to AIRB approach) and in 2014, highlighted a satisfactory predictive ability of the risk drivers aimed at assessing ESG risks: all the previous qualitative questions were confirmed in terms of their statistical significance, based on the time series of observed default rates.
Given the ever-rising worldwide awareness of the importance of ESG risks evaluation in general and more specifically of climate change risks, the currently used corporate rating models (authorized by the Regulator in April 2017) were further enhanced with regards to the ESG risks evaluation. Some new questions were added to the qualitative questionnaire to be compiled by the analyst, while further social and environmental risk drivers were included for the first time into the quantitative section of the model:
• regarding the new questions, the presence of business interruption or credit payment insurance against negative events, that could hinder the obligor’s solvency and the continuity of business activity enriches the qualitative questionnaire to be compiled by a rating analyst; in particular, the presence of an insurance against catastrophic events (wildfire, earthquakes, floods etc.) could be considered as a mitigating factor protecting the obligor against physical risks;
• as for the quantitative section, the presence/absence of certain certificates (quality, environmental, occupational health, information security etc.) was included as an independent risk driver into the regression model, aimed at predicting the default target variable.
Following the publication in March 2018 of the EU’s Action plan on sustainable finance (as part of a strategy to integrate environmental, social, and governance features into its financial policy framework and mobilise finance to support sustainable growth), the necessity of ESG factors evaluation begins to be incorporated into the regulatory framework.
As the ESG factors evaluation should be included in the decision-making and risk management framework of credit institutions, in collaboration with Moody’s Analytics (MA) ISP analyzed the relationship between the ESG scores / elementary risk drivers provided by Moody’s Vigeo Eiris and its corporate rating systems: the results showed that the ESG factors were correlated with the internal probability of default. Moreover, the relevance of specific drivers depended on the company’s dimension (e.g. governance features were more relevant for larger firms, while the environment / social “strategy” was more important for small-medium size firms) and economic sector (e.g. some sectors were more sensitive to environmental evaluation).
Starting from the evidence produced by the PoC, ISP conducted the analysis of the ESG drivers significance with respect to the default event (as a target variable) in order to develop a specific module to be included in the corporate model change framework, both at obligor level through a “bottom-up” approach (large corporate groups segment) and at sector/geographical area level through the “top-down” approach (corporate groups segment). This choice was related to data availability at counterparty level, which as expected was increasing along with the firms’ dimension.
The results of the above analysis can be summarised as follows:
• large corporate groups (at obligor level): the results of the statistical significance and discriminatory power analysis of the ESG factors with respect to the default event was demonstrated to be quite satisfactory, thus allowing for the definition of a specific module to be integrated in the LCG rating system;
• corporate groups (at sector/macro area level): given a scarce availability of data, a “top-down approach”, starting from the average scores computed on the large corporate development sample, was applied in order to analyse the significance of ESG risks for the corporate groups portfolio; given that in some circumstances counterintuitive results were obtained, mainly due to the representativeness challenges given by the fact that the average values were estimated on the sample with available ESG data (large corporates), which is different from the one which these averages were then applied to (SMEs), it was deemed more appropriate not to include the automatic ESG score evaluation based on average values in the CG rating system.
As already mentioned, it’s worth to note that in the corporate groups rating model the ESG aspects are covered by the evaluation of certificates and by means of some questions included in the qualitative questionnaire. Moreover, a new module aimed at investigating the exposure of the territory at municipality/regional level to catastrophic risks for Italian counterparties has been included in the model design. The percentage of the geographical area at high CAT risk is calculated to obtain a continuous risk level indicator for different natural calamities (e.g. flood, earthquakes, fire). The module impact on the company’s credit standing can only neutral (no risk) or negative. Notice, that the negative impact is neutralized in the case of presence of insurance coverage against catastrophic risks investigated through a specific answer to a question of the qualitative questionnaire.
As for the future developments, the Intesa Sanpaolo Group is going to further reinforce the risk management practices on ESG and climate related issues, in line with the European Commission’s Action Plan on Financing Sustainable Growth, as well as with EBA’s initial policy recommendations. Using the accumulated historical experience, the Group is committed to enhance climate risk awareness and to guarantee sound risk management practices on climate related issues in order to reduce the potential impact of climate change implications and to protect its reputation, its business and its investors. The climate change risk analysis is being incorporated in its risk management framework, with particular reference to credit risk and reputational risk, while the climate change sensitive exposures monitoring of credit portfolio is being included in the RAF framework, in order to integrate the long-term time horizon dimension and ESG analysis in the business relationships with corporates.
This would also allow to play an important role to raise awareness amongst corporates about long-term challenges and incentivize them to increase the shift towards sustainable business models.
As already explained in Q6 we believe that introducing environmental risks components within the risk differentiation phase would bring no substantial effects on prudential treatment of exposures and drivers on Social risk are not available. An exception can be made for bigger counterparties, where the attention to ESG aspects is more felt also thanks to higher resources that can be addressed to this issue and that can have a direct impact both on the financial statement and on the ordinary economic activity.
Despite this, we think that currently the introduction of these risks should be conducted through more direct actions even if the adoption of adjustment factors of risk weights for green and/or environmentally harmful exposures represents an extreme solution. Initially, intervention can be undertaken at process level by providing elements supporting for instance the override process or other actions like a differentiation of risk pricing on the base of an expert assessment.
The ESG risk factors, as already mentioned in Q9, can affect the performance both at obligor level and at sector / geographic area level. In the first case, we are talking about the risks that could have an impact on the issuer, not on the entire market. These risks are linked to certain factors, such as the issuer's governance, its regulatory compliance and the reputation of the brand, and the obligor-level information set necessary for their evaluation is generally available for larger obligors. In the second case, however, there could exist some broader issues affecting the entire sector or territory which the issuer belongs to and that may be linked to regulatory factors, technological changes or exposure to natural catastrophes. In this case, the necessary aggregated data could be available also for smaller obligors and provided by public sources.
Therefore, the lack of homogeneous and structured data at obligor level makes it difficult to carry out an ESG risk assessment using a "bottom-up" approach, i.e. starting from the specific characteristics of a certain counterparty. Such an approach would produce more robust ESG risk estimates but is hardly applicable for smaller businesses. An alternative way is to use external data available from public sources, which however exist only at geographical area or sector level, and therefore can only be used for "top-down" assessments. Nevertheless, such an approach does not allow to evaluate the actual management of ESG risks for a specific counterparty, given that all companies belonging to a certain cluster are attributed the same average level of risk. This could penalize the obligors which, although belonging to sectors or geographical areas deemed risky, seek to reduce their negative impact on the environment, the economy or the community. Furthermore, the estimation of the average risk for a certain subset of counterparties is generally carried out on a sample of observations for which a precise assessment has been conducted and the data to be analyzed is available. If this estimation sample were not representative of the reference population, the results would be inaccurate.
In general, in order to evaluate the correlation between a given risk factor and the risk of default, a time series of historically registered data is needed. When available, a traditional estimation approach can be applied to study the discriminatory power of ESG risk factors. This means that one should analyze different sets of information collected at a given point in time (T) and connect them to the observed default frequency over the next period (usually equal to 12 months, between T and T+1). The data collected by means of qualitative questionnaire can be used for this purpose. Rating analysts can be tasked with compiling the questionnaire during the rating attribution process, based on their expert judgment and specific guidelines. The latter are necessary to make the answers based on specialist evaluation homogeneous across different assessments.
Given the stored data, the analysis of default rates behavior for each answer to the relevant question, as well as the univariate analysis aimed at establishing the discriminatory power of risk drivers, could give a first insight on the correlation of the ESG risks with the credit risk of a certain counterparty. Subsequently, the ESG data can be included as independent risk drivers into the regression analysis model (e.g. logistic regression), aimed at predicting the default target variable.
It’s worth to underline that in general the possibility to include some risk factors into the rating model through a quantitative approach is highly dependent on the availability of data and of a representative sample: these are essential for the estimation of the relationship of hypothetical risk drivers with a given target variable, as well as for the analysis of the correlation intensity in terms of discriminatory power. A qualitative approach can be used instead as a data collection instrument, which is fundamental during the transition towards the quantitative approach. Moreover, the collected data can be also used for an ex-post validation of the judgmentally assigned factor weights, eventually adjusting the initial assumptions. Another useful tool could be the override process applied by the analysts in the rating attribution process, generally used to account for the information not already taken into consideration by the rating model or weighted by the model in a way that is not representative of the risk profile of the obligor.
The question regards four different points, that we address below in the order that we understand to be logically sequential.
• New Risk Factor: as clearly pointed out in the EBA paper: “while environmental risks may not lead to the introduction of new risk factors per se, they may affect the magnitude of their shocks. In other words, the presence of environmental risks may lead to a ‘classical’ risk factor (e.g. an equity price, or an exchange rate) being more volatile than historically observed”, it’s very difficult to disentangle the ESG Risk component embedded into market prices. As a consequence, independently of the framework adopted within the standardize approach to capitalize such risk, it would result very difficult, or even impossible, to identify the appropriate financial instruments or calibrate the shocks/stresses to observable market prices.
• Increase risk weight: despite this is the simplest approach, following the discussion above it would be very difficult or even impossible to calibrate an increase to take into account an ESG component. This uncertainty could also lead to unintended increase of capital charges across all institutions. Furthermore, this approach would entail double counting of ESG risk
• New bucket: since ESG is a macro risk that affect all risk classes, defining a specific bucket is theoretically possible but not feasible in practice. For instance, which products should be classified into this additional ESG bucket? Beyond commodity and credit instruments, concerning other risk classes the classification may be more ambiguous. Furthermore, this approach would entail double counting of ESG risk as above.
• RRAO: this framework is intended to address risks not fully covered by the SBA and default risk frameworks, but which are already identified by the risk management. This seems not to be the case of ESG risk, according to the discussion above.
Since the FRTB Framework requires the alignment between Front Office (FO) and Risk Management prices, at the moment it seems not possible to define a new risk factor which is not priced by FO systems and directly linked to the current asset classes. Risk Management may model curves, surfaces, vertices for each asset class, but may not unbundle a cross risk factor like ESG without seriously affecting the PLA (Profit and Loss Attribution). Otherwise, considering ESG risk as a non-modellable risk factor would result in an unintended penalization of ESG compliant instruments.
Outside the IMA model a stress approach would be possible, at least for a limited number of instruments. For example, for credit instruments a stress on PDs could be developed in order to capture the different default probability for companies/issues related to ESG risk.
As well described into the EBA Discussion Paper, environmental risk is a macro risk factor that is already incorporated in market price, thus it’s already capitalized (e.g. taking two issuances, one Green, the other one not green, they will have different market spread). We do not see much room for incorporating ESG into Pillar 1 trough Market Risk RWAs. However, it could be developed a model for weighting the ESG component trough sophisticated econometric/historical models as a managerial tool for monitoring the institution investments.
It would be appropriate to clarify that environmental, social and governance phenomena do not represent new forms of risk, but risk factors that increase the probability of occurrence and / or the severity of already existing risk categories (e.g., Credit, Market, Liquidity, Operational and Reputational).
Therefore, the existing regulatory event types taxonomy should be considerate as adequate, and at this stage an environmental or ESG flag will be enough to enable a more effective risk event reporting and analysis. The same goal could be met by means of a more granular internal taxonomy, for example by a better characterization of the cause dimension (e.g., weather effects). It would be however desirable to encourage Operational Risk consortia and associations to develop an industry standard and a guidance describing the application of the aforementioned flag.
A further issue to be clarified is the differentiation between normal weather events and climate risk phenomena.
It is important to note that the regulatory framework in the context of the Basel III accord relies solely on a size-based metric (Business Indicator) especially if the Internal Loss Multiplier (ILM) will be set equal to 1. Such approach doesn’t seem capable at all to capture the increasing severity and frequency of physical environmental events. Furthermore, also considering the ILM contribution it won’t be possible to extrapolate the portion connected to environmental issues as the ILM is based on aggregate historical losses. Therefore, it implies the such an approach is not forward-looking and it is not very risk-sensitive (it estimates the historical average rather than the proper risk profile).
Therefore, it is sensible to include a forward-looking approach under Pillar 2 framework which could be based on risk-sensitive models.
We agree with the proposal to keep the risk factor on strategic and environmental risks under Pillar 2. This would allow banks to use specific a risk-sensitive framework to assess properly the risk factors.
It is crucial that any framework treats ESG as a risk theme (i.e., similar to conduct risks) rather than an isolated new risk type. It is therefore desirable to set industry standards and relevant guidance to allow a real level playing-field of ESG management among firms.
Intesa Sanpaolo welcomes the EBA discussion paper on the role of ESG risks in prudential framework for credit institutions and investment firms aimed at establishing whether and how environmental risks are to be incorporated into the Pillar 1 prudential framework. Some of the most important aspects to be considered concern the extent to which the ESG risk drivers are already captured by the current prudential framework in order to avoid double counting, to ensure the framework’s consistency and robustness. At the same time, Intesa Sanpaolo (ISP) agrees that environmental risks, related to a more general climate change risk, are more forward looking, meaning that they are going to result in higher and potentially more extreme losses in the future with respect to their impact observed in the past. In addition, social risk is also connected to the environmental-climate risk, but including this impact in the framework would be challenging. Generally speaking, the focus of the DP is mainly on environmental risk and only in part on social risk, however we highlight the need for a comprehensive assessment, also at a later stage, on all ESG risks.
As such, Intesa Sanpaolo deems the subject under discussion to be of great relevance and has attentively followed both regulatory and methodological developments regarding the evaluation of the ESG risks in its IRB models. In fact, the first evaluation of ESG risk factors was incorporated by Intesa Sanpaolo in the “qualitative” component of the PD models back in 2009, with the introduction of the first foundation internal ratings-based approach (FIRB) for corporate rating models.
More recently, in 2021, a model change request was submitted to the Supervisory Authorities for the validation and authorization of new rating models within the corporate regulatory segment which further strengthens the analysis of ESG factors. For large corporate counterparties operating at international level, the model design includes an ad hoc internally developed obligor-level ESG score, while for the Italian Corporate SMEs the qualitative component of the rating model also includes a top-down assessment of aspects linked to catastrophic events characterizing specific geographic areas.
As for the future developments, the Intesa Sanpaolo Group is going to further reinforce the risk management practices on ESG and climate related issues, in line with the European Commission’s Action Plan and EBA’s initial policy recommendations. The Group is committed to enhance climate risk awareness and sound risk management practices on climate related issues in order to reduce the potential impact of climate change implications and to protect its reputation, its business, its clients and its investors.
The climate change risk analysis is being incorporated in ISP risk management framework, with particular reference to credit risk and reputational risk, while the climate change sensitive exposures monitoring of credit portfolio is being included in the RAF (risk appetite framework), in order to integrate the long-term time horizon dimension and ESG analysis in the business relationships with corporates. This would also allow to play an important role in raising awareness amongst corporates on long-term challenges and incentivize them to increase the shift towards sustainable business models.
Intesa Sanpaolo would like to anticipate hereby the most important and impacting issues of the discussion paper regarding the context of the IRB framework, while more details will be contained in the specific answers to the consultation. Intesa Sanpaolo would like to flag:
• that it would be appropriate to clarify that environmental, social and governance phenomena do not represent new forms of risk, but risk factors that increase the probability of occurrence and / or the severity of already existing risk categories (eg. Credit, Market, Liquidity, Operational and Reputational)
• the necessity to avoid possible overlaps with the existing legislative framework in order not to excessively increase regulatory costs and validation process complexity;
• the importance to define a commonly used taxonomy regarding ESG risks with the objective to produce comparable results and to ensure a level playing field among institutions;
• with regards to the physical and transition risks of the climate change, there is a high level of uncertainty of the time horizon related to the negative events occurrence and intensification, as well as to the regulatory framework strengthening aimed at risk mitigation;
• Social and Governance risk should also be analyzed and taken into account;
• given a certain level of correlation between the credit risk and ESG risk factors, the only inclusion of these variables in the risk differentiation phase of the IRB models could be not sufficient as a significant incentive for the economy to transition to a more sustainable framework, given the backward looking time series-based nature of the model development process aimed at estimating a probability of default over a one-year time interval, as opposed to the forward-looking nature of the climate change;
• the availability and coverage of historical data, characterized by accuracy, completeness and reliability could influence the portfolio-specific results obtained by inclusion of ESG risk drivers into IRB models, while the climate risks reduction should be managed in a homogenous way for all economic players; for this reason, a uniform regulatory adjustment factor could be an additional incentive to transition to a more sustainable business structure.
• finally, given the importance of the climate change risks and the necessity to lead the market participants to a more sustainable business structure, the current challenges related to the ESG risk drivers’ assessment in the IRB models should not be further complicated by margins of conservatism requests, thus discouraging banks from considering these risks when evaluating the obligor’s credit risk profile.
Overall, in ISP view, the Pillar 1 framework could be insufficient to deal with environmental risks, as highlighted in the previous points.
The scenario and sensitivity analyses, among other instruments, could be additional incentivizing tools, by inserting qualitative and quantitative assessments into granting processes able to address the long-term strategies of the institutions. Intesa Sanpaolo is currently working on sectoral limits on Large Exposures in specific sectors where the CO2 emissions are the most relevant, and is implementing climate scenarios for the stress testing framework. As one issue institutions are facing is comparability, ISP is working on specific scenarios defined by the Network for Greening the Financial System (NGFS)-phase 2.