Credit Risk Assessment in the Climate Shadow: Evidence From White and Grey Literature

This study examines the extension of the economic framework to correlations between PD, LGD, and EAD. We build on a framework that has already been used to figure out and adjust the relationships between loan portfolios’ Probability of Default (PD), Loss Given Default (LGD), and Exposure at Default (EAD). Our analysis explores the implications of incorporating these correlations in portfolio losses, arguing that this structure enables institutions to apply forward-looking correlation models to assess the likelihood of obligor credit quality deterioration, commonly referred to as a significant increase in credit risk (SICR). According to International Financial Reporting Standards (IFRS)-9 regulations, the estimation of SICR and forward-looking information should not entail excessive cost or effort. In line with this principle, we contend that only a limited number of inputs are necessary to implement this robust framework, which allows users to evaluate meaningful forward-looking correlations, identify obligors likely to experience SICR, and ultimately measure a more accurate Expected Credit Loss (ECL). The adoption of this approach will allow institutions to better understand their credit risk and better assess their credit risk practices while adhering to regulatory requirements.

Chapter 4 - Credit Risk

The aggregation of individual risks into total risk using a weighting variable multiplied by two ratio variables representing incidence and intensity is an important task for risk professionals. For example, expected loss (EL) of a loan is the product of exposure at default (EAD), probability of default (PD), and loss given default (LGD) of the loan. Simple weighted (by EAD) means of PD and LGD are intuitive summaries however they do not satisfy a reconciliation property whereby their product with the total EAD equals the sum of the individual expected losses. This makes their interpretation problematic, especially when trying to ascertain whether changes in EAD, PD, or LGD are responsible for a change in EL. We propose means for PD and LGD that have the property of reconciling at the aggregate level. Properties of the new means are explored, including how changes in EL can be attributed to changes in EAD, PD, and LGD. Other applications such as insurance where the incidence ratio is utilization rate (UR) and the intensity ratio is an average benefit (AB) are discussed and the generalization to products of more than two ratio variables provided.

E AD and LGD estimates are key inputs in measurement of the expected and unexpected credit losses and, hence, credit risk capital (regulatory as well as economic). These are the second dimensions of Basel II IRB formula. To estimate LGD and EAD under advanced approach, each bank has to rely on its internal data on defaulted accounts. Basel II specifies that “LGD estimates must be grounded in historical recovery rates” – thus excluding any subjective choice of LGD estimates by banks. Building a drawing power and recovery predictor model from loss perspective side is hard because of non-availability of data within the banks in India. The banks also may not know in which format they should record and collect the relevant data for estimation of these two key risk parameters. This section provides a detailed explanation about EAD and LGD concepts, their estimation methodologies and data collection procedure. It also covers various studies related to EAD and LGD, provides interesting statistics about them, and discusses crucial factors that drive these two risks. What is Exposure at Default (EAD)? EAD is the amount of loss that a bank may face due to default. Since default occurs at an unknown future date, this loss is contingent upon the amount to which the bank was exposed to the borrower at the time of default. This is commonly expressed as exposure at default (EAD). In the case of normal term loan, exposure risk can be considered small because of its fixed repayment schedule. This is not true for all other lines of credit (e.g. guarantee, overdraft, letter of credit, etc.). The borrower may draw on these lines of credit within a limit set by the bank as and when borrowing needs arise. Credit line usage has cyclical characteristics, i.e. the use increases in recessions and declines in expansions. The usage rate increases monotonically as the borrower becomes riskier and approaches towards default risk. Banks as a lender need to closely monitor the potential exposure to assess the credit risk more prudently. It is in this sense that the estimation of EAD is absolutely necessary for computation of regulatory as well as economic capital.

Two-stage consumer credit risk modelling using heterogeneous ensemble learning

We develop an empirical framework for the credit risk analysis of a generic portfolio of revolving credit accounts and apply it to analyze a representative panel data set of credit card accounts from a credit bureau. These data cover the period of the most recent deep recession and provide the opportunity to analyze the performance of such a portfolio under significant economic stress conditions. We consider a traditional framework for the analysis of credit risk where the probability of default (PD), loss given default (LGD), and exposure at default (EAD) are explicitly considered. The unsecure and revolving nature of credit card lending is naturally modeled in this framework. Our results indicate that unemployment, and in particular the level and change in unemployment, plays a significant role in the probability of transition across delinquency states in general and the probability of default in particular. The effect is heterogeneous and proportionally has a more significant impact for high credit score and for high-utilization accounts. Our results also indicate that unemployment and a downturn in economic conditions play a quantitatively small, or even irrelevant, role in the changes in account balance associated with changes in an account’s delinquency status, and in the exposure at default specifically. The impact of a downturn in economic conditions and, in particular, changes in unemployment on the recovery rate and loss given default is found to be large. These findings are of particular relevance for the analysis of credit risk regulatory capital under the IRB approach within the Basel II capital accord.

In this paper, we develop a factor-type latent variable model for portfolio credit risk that accounts for stochastically dependent probability of default (PD), loss given default (LGD) and exposure at default (EAD) at both the systematic and borrower specific levels. By employing a comprehensive simulation study, we set our results in contrast to those obtained using the asymptotic single risk factor (ASRF) model that underlies Basel II and III. Several sensitivity and robustness analyses for different parameter assumptions are conducted to break down our results. As required by the regulator, we show how to map our portfolio credit loss quantiles with correlated PD, LGD and EAD into values for downturn LGD and EAD. Our analyses reveal that stochastically dependent defaults, LGD and EAD increase a credit portfolio’s tail risk significantly. Estimating risk markups separately can therefore lead to substantially underestimating the inherent risk of a portfolio with nondeterministic exposures. Further, we show that relative downturn markups on LGD and EAD depend strongly on asset correlations and, in particular, that credit portfolios with low asset correlations might be prone to an underestimation of additional capital charges for stochastically dependent LGD and EAD. Thus, our results are of economic significance for banks and regulators when setting up minimum capital requirements.

The Basel Accords have created the need to develop and implement models for probability of default (PD), loss given default (LGD) and exposure at default (EAD). Although PD is quite well researched, LGD and EAD lag behind in terms of both theoretical and practical insight. This paper proposes some empirical approaches for EAD/LGD modeling and provides technical insights into their implementation. It is expected that approaches proposed in the paper will be helpful for modelers and risk managers in their risk modeling and management practice.

We develpo an empirical framework for the credit risk analysis of a generic portfolio of revolving credit accounts and apply it to a representative panel data set of credit card accounts. These data cover the period of the most recent recession and provide the opportunity to analyze the performance of such a portfolio under significant economic stress conditions. We consider a traditional framework for the analysis of credit risk where expected loss is represented in terms of the probability of default (PD), loss given default (LGD), and the exposure at default (EAD). The unsecured and revolving nature of credit card lending is naturally modeled in this framework. Our results indicate that unemployment, and in particular the level and change in unemployment, plays a significant role in the probability of transition across delinquency states in general and the probability of default in particular. The effect is heterogeneous and proportionally has a more significant impact for high credit score and for high-utilization accounts. Our results also indicate that unemployment and a downturn in economic conditions play a quantitatively small, or even irrelevant, role in the changes in account balance associated with changes in an account's delinquency status, and in the account balance exposure at default specifically. The impact of downturn economic conditions on the recovery rate and loss given default is found to be large. These findings are of particular relevance for the analysis of credit risk regulatory capital under the IRB approach within the Basel II capital accord.

The credit risk measure, Expected Loss (EL) is defined as the product of the three risk parameters: probability of default (PD), loss given default (LGD) and exposure at default (EAD). EL is central to risk management, profit estimation, calculating regulatory capital requirements and the standard accounting rules for credit (IFRS 9). Although correlations between the three risk parameters is evident, there is limited published work exploring these correlations and their impact on estimating EL accurately and without bias. Often EL is calculated simply assuming independence. In this study, EL is derived from first principles, without assuming independence between the three risk parameters. The main results are, firstly, that correlation between PD and LGD has no consequence on the calculation of EL, if LGD is treated as conditional on default. However, correlation between LGD and EAD does have an impact, requiring an adjustment to enable an accurate and unbiassed estimate. Additionally, there is no selection bias resulting from using LGD and EAD models built conditional on default, when applied across the total credit population. These results are demonstrated through a simulation study and by application to a real credit card data set.

The subject of this paper is the analysis of the application of banking internal credit risk measurement models for the purpose of calculating the minimum regulatory capital. The Basel Committee established proposals for an internal rating-based approach (IRB approach-internal rating-based) to capital requirements for credit risk. Such an approach, which relies on the bank's internal assessment of counterparties and exposures, can ensure two key objectives: the first is additional risk sensitivity, in which capital requirements based on internal ratings can be much more sensitive on the drivers of credit risk of economic losses in the banking portfolio; the second is incentive compatibility, where the appropriate structure of the RBI approach can provide a framework that encourages banks to continue to improve their internal risk management practices. The internal ranking approach aims to improve the safety and soundness of the financial system. The paper defines the terminology and classification of the rating system. The probability of default (PD-probability at default) and the other two risk components LGD (loss given default) and EAD (exposure at default) are key input parameters for the calculation of regulatory capital. The rating system is, therefore, a significant driver of risk management and financial performance measurement. To be in a position to demonstrate to supervisors that an internal rating system should be used for the purposes of determining minimum regulatory capital requirements, banks must first demonstrate that the rating system is an integral part of their ongoing operations and risk management culture.

Benchmarking regression algorithms for loss given default modeling

Managing the credit risk inherent to a corporate credit line is similar to that of a term loan, but with one key difference. For both instruments, the bank should know the borrower's probability of default (PD) and the facility's loss given default (LGD). However, since a credit line allows the borrowers to draw down the committed funds according to their own needs, the bank must also have a measure of the line's exposure at default (EAD). Our study, which is based on a census of all corporate lending within Spain over the last 20 years, provides the most comprehensive overview of corporate credit line use and EAD calculations to date. Our analysis shows that defaulting firms have significantly higher credit line usage rates and EAD values up to five years prior to their actual default. Furthermore, we find that there are important variations in EAD values due to credit line size, collateralization, and maturity. While our results are derived from data for a single country, they should provide useful benchmarks for further academic, business and policy research into this underdeveloped area of credit risk management.

The Basel Accord emphasizes the necessity of employing internal data models to manage key credit risk components, including Probability of Default (PD), Loss Given Default (LGD), and Exposure At Default (EAD). Among these, internal datasets are critical for estimating PD, a fundamental measure of borrower creditworthiness. Nevertheless, practical application often faces challenges due to incomplete datasets, which can skew analyses and undermine the accuracy of credit scoring models. Traditional approaches to addressing missing data, such as sample deletion or mean imputation, are widely used; however, they often prove insufficient for accurate prediction. Consequently, imputation methods are typically favored over deletion, as they allow for the full utilization of available data. Recent advancements have introduced more sophisticated techniques, such as Generative Adversarial Imputation Networks (GAIN), which utilize a generative adversarial network to model data distributions and impute missing values with greater precision than conventional methods. Building on these developments, this study proposes a novel imputation approach, SMART (Structured Missingness Analysis and Reconstruction Technique) for credit scoring datasets. SMART consists of two primary stages: first, it normalizes and denoises the dataset using randomized Singular Value Decomposition (rSVD), followed by the implementation of GAIN to impute missing values. Experimental results demonstrate that SMART significantly outperforms existing state-of-the-art methods, particularly in high missing data contexts (20%, 50%, and 80%), with improvements in imputation accuracy of 7.04%, 6.34%, and 13.38%, respectively. In conclusion, SMART represents a substantial advancement in handling incomplete credit scoring datasets, leading to more precise PD estimation and enhancing the robustness of credit risk management models.

There has been increasing support in the empirical literature that both the probability of default (PD) and the loss given default (LGD) are correlated and driven by macroeconomic variables. Paradoxically, there has been very little effort from the theoretical literature to develop credit risk models that would include this possibility. The goals of this paper are: first, to develop the theoretical reduced-form framework needed to handle stochastic correlation of recovery and intensity, proposing a new class of models; second, to understand under what conditions would our class of models reflect empirically observed features and, finally, to use concrete model from our class to study the impact of this correlation in credit risk term structures. We show that, in our class of models, it is possible to model directly empirically observed features. For instance, we can define default intensity and losses given default to be higher during economic depression periods - the well-know credit risk business cycle effect. Using the concrete model we show that in reduced-form models different assumptions - concerning default intensities, distribution of losses given default, and specifically their correlation - have a significant impact on the shape of credit spread term structures, and consequently on pricing of credit products as well as credit risk assessment in general. Finally, we propose a way to calibrate this class of models to market data, and illustrate the technique using our concrete example using US market data on corporate yields.