If you are a data scientist or an analytics professional preparing for a risk analytics interview, it’s crucial to be well-prepared. To help you with that, we have compiled a list of questions covering various aspects of credit risk analytics. This article is divided into three sections:
- 10 Questions on Banking
- 10 Questions on Model Development and Validation
- 10 Questions on Time Series
Let’s dive in and explore the world of credit risk analytics!
10 Questions on Banking
What are the 3 Pillars in Basel Framework?
The Basel Framework consists of three pillars that aim to ensure the stability and soundness of the banking system. These pillars are:
- Minimum Capital Requirement: Calculated based on the risk under various heads. For credit risk, the approaches are Standardized, F-IRB, and A-IRB. For market risk, the approach is VaR. For operational risk, the approaches are Basic Indicator Approach, Standardized Approach, and Internal Measurement Approach.
- Supervisory Review: It is based on the Internal Capital Adequacy Assessment Plan. This pillar gives banks the power to review their risk management system and make necessary improvements.
- Market Discipline: Developing a set of disclosure requirements that require institutions to disclose details such as scope of application, capital, risk exposures, risk assessment process, and capital adequacy of the institution. This pillar enhances transparency and accountability.
What are the approaches for the treatment of impaired provisions?
When it comes to impaired provisions, there are three main approaches:
- Standardized Approach: Regulators prescribe the risk weight. If a loss has occurred, it impacts Tier 1 capital.
- Foundation IRB Approach: Banks estimate the 1-year Probability of Default (PD), while regulators prescribe Loss Given Default (LGD) and Exposure at Default (EAD). If Expected Loss (EL) exceeds the provision, the excess is reduced from capital. If EL is less than the provision, the excess is added to capital.
- Advanced IRB Approach: Banks estimate PD, LGD, and EAD based on their internal models.
What is ICAAP?
ICAAP stands for Internal Capital Adequacy Assessment Plan. It is a process through which banks inform their board of directors about the ongoing assessment of the bank’s risk and how the bank intends to mitigate those risks. The purpose of ICAAP is to assess the current and future capital requirements of the bank.
What is Capital?
Capital serves as a buffer to absorb unexpected losses and fund the ongoing activities of the bank. In the Basel Framework, capital is categorized into two tiers:
- Tier 1 Capital: Also known as core capital, Tier 1 capital includes shareholder’s equity and retained earnings. The minimum Tier 1 capital requirement is 6% of Risk-Weighted Assets (RWAs).
- Tier 2 Capital: Also called supplementary capital, Tier 2 capital includes items such as revaluation reserves, hybrid capital instruments, subordinate term debt, general loan loss reserves, undisclosed reserves, etc. The minimum Tier 1 + Tier 2 capital requirement is 8% of RWAs. The minimum capital adequacy ratio, including the capital conservation buffer, is 10.5% of RWAs.
What are the key Capital Ratios?
There are three key capital ratios used to measure a bank’s capital adequacy:
- Common Equity Tier 1 (CET1) Ratio: It represents the common equity under the revised capital framework divided by the Standardized Approach to RWAs.
- Tier 1 Ratio: It represents the Tier 1 capital under the revised capital framework divided by the Standardized Approach to RWAs.
- Total Capital Ratio: It represents the total capital under the revised capital framework divided by the Standardized Approach to RWAs.
What is expected loss and unexpected loss?
In credit risk analytics, two important concepts are expected loss and unexpected loss:
- Expected Loss (Provision): Expected loss is the sum of the values of all possible losses, each multiplied by the probability of that loss occurring. It is calculated as Expected Loss (EL) = Probability of Default (PD) x Exposure at Default (EAD) x Loss Given Default (LGD).
- Unexpected Loss (Capital): Unexpected loss refers to losses beyond the expected loss. It is calculated as Unexpected Loss (UL) = EAD x SQRT[(PD^2 x σ^2LGD) + (LGD^2 x σ^2PD)]. Banks are required to hold capital to cover unforeseen financial losses.
What is IFRS9?
IFRS9 stands for International Financial Reporting Standard 9. It is an accounting standard issued by the International Accounting Standards Board (IASB) for the recognition of impairments. IFRS9 introduces a forward-looking approach to determine credit impairments.
Under IFRS9, impairments are classified into three stages:
- Stage 1: This stage applies to loans that are originated or existing loans with no significant increase in credit risk. Expected Credit Loss (ECL) resulting from default events in the next 12 months is recognized.
- Stage 2: This stage applies if the credit risk has increased significantly and is not considered low. Lifetime ECL is recognized.
- Stage 3: This stage applies when the credit risk increases to a point where it is considered credit impaired. Lifetime ECL is recognized.
IFRS9 has implications such as earlier recognition of losses, differentiation of exposures based on deterioration, and the requirement of loss forecasting.
What is CECL?
CECL stands for Current Expected Credit Loss. It is an accounting standard issued by the Financial Accounting Standards Board (FASB) for the recognition of impairments. CECL and IFRS9 are similar in principle but have some differences.
The key difference between CECL and IFRS9 is the estimation of lifetime losses:
- CECL: CECL estimates lifetime losses upon the initial recognition of assets.
- IFRS9: IFRS9 requires 12-month ECL for performing loans and lifetime ECL for under-performing or non-performing loans.
What is CCAR?
CCAR stands for Comprehensive Capital Analysis and Review. It is an assessment conducted by the Federal Reserve to ensure that large systematically important banking institutions have a forward-looking, institution-specific, risk-tailored capital planning process. The objective of CCAR is to assure that banks have sufficient funds to remain solvent during times of economic and financial distress.
The CCAR process involves stress testing and sensitivity analysis to evaluate the resilience of banks under various scenarios. The Federal Reserve provides supervisory scenarios, and banks also create internal scenarios known as Bank Holding Company (BHC) scenarios.
What is Stress Testing and Sensitivity Analysis?
Stress testing and sensitivity analysis are essential components of the CCAR process. They involve assessing the impact of adverse economic conditions on the financial stability of banks. Stress testing involves subjecting banks’ balance sheets to various hypothetical scenarios, such as economic downturns or market shocks, to evaluate their resilience and ability to withstand adverse conditions.
There are three scenarios provided by the Federal Reserve (supervisory scenarios): Base, Adverse, and Severely Adverse. These scenarios forecast the performance of banks’ balance sheets and financial indicators over a period of nine quarters.
Sensitivity analysis, on the other hand, focuses on measuring the sensitivity of banks’ default exposures to changes in key risk factors. It calculates sensitivity ratios such as the adverse sensitivity and severely adverse sensitivity, which compare default exposures under different scenarios to the baseline scenario.
Both stress testing and sensitivity analysis play a crucial role in identifying potential vulnerabilities in banks’ capital adequacy and risk management practices.
10 Questions on Model Development and Validation
What is Probability of Default (PD)?
Probability of Default (PD) is a key concept in credit risk modeling. It represents the average number of obligors that default in a particular rating grade within a year. PD is estimated using statistical techniques such as logistic regression models. In these models, the outcome variable is dichotomous, indicating whether an obligor will default or not based on various independent variables.
Some of the dependent variables commonly used in PD estimation include current non-payment, historical non-payment, percentage of payment, credit limit utilization, maturity, and other relevant factors.
What is Exposure at Default (EAD)?
Exposure at Default (EAD) is an estimate of the outstanding amount a bank is exposed to in case an obligor defaults. EAD is highly relevant in revolving credit facilities such as credit cards or lines of credit. It focuses on metrics that measure the increase in balances between a reference time and the date of default.
EAD is estimated through methodologies such as the Exposure at Default Formula (EADF), which calculates EAD as the balance at default divided by the balance at the reference date.
What is Loss Given Default (LGD)?
Loss Given Default (LGD) is the percentage of exposure that a bank might lose if an obligor defaults. LGD depends on the characteristics of the loan and the collateral held by the bank.
For example, in the case of mortgages, the collateral (property) determines the potential recovery in the event of default. In contrast, for credit cards, where there is no specific collateral, the recovery is based on factors such as the cash flow generated post-default.
LGD is empirically derived and can be calculated as 1 minus the sum of payments for a specific period divided by the maximum of balances at different time intervals.
What is the difference between Through the Cycle (TtC) and Point in Time (PiT) PD?
Through the Cycle (TtC) PD and Point in Time (PiT) PD are two different approaches to modeling and estimating credit risk:
- Through the Cycle (TtC) PD: TtC PD takes a longer-term perspective into consideration and aims to capture the average default behavior of obligors over economic cycles. It is more stable and less prone to short-term fluctuations in economic conditions.
- Point in Time (PiT) PD: PiT PD focuses on capturing the current credit risk of obligors based on their specific characteristics and the prevailing macroeconomic conditions. It aligns with recent macroeconomic scenarios and provides a more granular assessment of credit risk.
Both TtC PD and PiT PD have their respective advantages and applications in credit risk modeling. The choice of approach depends on the specific requirements and objectives of the modeling exercise.
What is Information Value (IV)?
Information Value (IV) is a widely used concept in variable selection during model development. It plays a crucial role in credit risk modeling, particularly in the credit card industry.
IV measures the strength of the relationship between a predictor variable and the target variable (default or non-default). It is calculated as the sum of the differences between the distribution of defaults and non-defaults multiplied by the Weight of Evidence (WoE).
WoE is the logarithm of the ratio of the distribution of non-defaults to defaults for a specific category or bin of a predictor variable. The higher the IV, the greater the explanatory power of the variable in predicting the target variable.
What is Population Stability Index (PSI)?
Population Stability Index (PSI) is a metric used to monitor the stability of a population or data distribution over time. It is particularly relevant in credit risk modeling when assessing the performance and stability of predictive models.
PSI is calculated by comparing the differences between the actual and estimated distributions of a specific variable. It measures the extent to which changes in economic conditions or internal policy changes affect the population or data distribution.
A PSI below 0.1 indicates no significant change, between 0.1 and 0.25 suggests the need for close monitoring, and above 0.25 indicates the need to redevelop the model to account for changes in the underlying population.
What is a Confusion Matrix?
A Confusion Matrix is a square matrix that summarizes the performance of a classification model. It is commonly used in binary classification problems, where the outcome can be either positive or negative.
The Confusion Matrix consists of four cells:
- True Positive (TP): The model correctly predicted the positive cases (e.g., default) as positive.
- True Negative (TN): The model correctly predicted the negative cases (e.g., non-default) as negative.
- False Positive (FP): The model incorrectly predicted the negative cases as positive (Type I error).
- False Negative (FN): The model incorrectly predicted the positive cases as negative (Type II error).
The Confusion Matrix provides valuable insights into the performance metrics such as accuracy, precision, recall (sensitivity), and specificity, which measure the model’s effectiveness in correctly predicting the positive and negative cases.
What is Concordance?
Concordance is a concept used in the evaluation of predictive models, particularly in credit risk modeling. It measures the rank ordering of the predicted probabilities.
A pair of observations is considered concordant if the observation with the desired outcome (e.g., event) has a higher predicted probability than the observation without the desired outcome (e.g., non-event). A pair is discordant if the opposite is true, and a pair is tied if both observations have the same predicted probability.
Concordance assesses the ability of a model to correctly rank the likelihood of events or outcomes. It is particularly useful in applications such as credit scoring, where the ranking of individuals based on their creditworthiness is critical.
What is a Gain and Lift Chart?
A Gain and Lift Chart is a visual representation of the performance of a predictive model. It is used to evaluate the rank ordering of predictions and assess the model’s effectiveness in identifying positive cases (e.g., defaults).
The Gain Chart shows the percentage of positive cases (events) covered at each decile level. It helps to identify the proportion of targets captured by the model at different levels of prediction.
The Lift Chart, also known as the Lift Curve, measures the ratio of the model’s performance to random expectation. It is calculated by dividing the gain percentage at each decile level by the random expectation percentage.
Gain and Lift Charts provide insights into the model’s ability to prioritize the most significant cases and identify high-risk individuals or events.
What is KS, AUROC, and Gini?
KS (Kolmogorov-Smirnov), AUROC (Area Under the ReceiverOperating Characteristic curve), and Gini coefficient are commonly used metrics to evaluate the performance of classification models, including credit risk models.
- KS (Kolmogorov-Smirnov): The KS statistic measures the separation power of a classification model. It calculates the maximum absolute difference between the cumulative distribution functions of the positive and negative cases. A higher KS value indicates better discrimination power of the model. Typically, a KS value above 30 is considered excellent.
- AUROC (Area Under the Receiver Operating Characteristic curve): AUROC is a fundamental tool for evaluating the performance of diagnostic tests, including credit risk models. The ROC curve plots the true positive rate (sensitivity) against the false positive rate (1-specificity) at various classification thresholds. The AUROC represents the area under this curve. An ideal model will have an AUROC close to 1, indicating high discriminative power.
- Gini coefficient: The Gini coefficient is derived from the AUROC and provides a single-value summary of the model’s discrimination power. It measures the ratio between the area between the ROC curve and the diagonal line (random model) and the area above the diagonal. The Gini coefficient can range from 0 to 1, with 1 indicating perfect discrimination and 0 indicating no discrimination.
These metrics help assess the effectiveness of credit risk models in differentiating between positive and negative cases, and provide valuable insights for model selection and performance evaluation.
By thoroughly understanding the concepts and answering these questions, you will be well-prepared for a risk analytics interview. Remember to adapt your responses to your specific experience and expertise. Good luck!