CREDIT RISK MODELING – 365 DATA SCIENCE

1. What does the course cover

Course Introduction: Credit Risk Modeling in Python

Nikolai, an expert with a PhD in business and economics, will guide you through credit risk
modeling in Python. He has experience in risk modeling, marketing analytics, and consulting for
financial institutions, particularly in improving credit risk models.

Course Structure:

 Theoretical Basics: Start with understanding credit risk—what it is, why it’s important,
and how financial institutions measure it.
 Data Preprocessing: Learn techniques for cleaning data, handling missing values, and
preparing continuous and discrete data.
 Predicting Default: Explore best practices for predicting the probability of default (PD)
using regulatory rules like Basel II/III. This includes techniques like fine/coarse classing,
weight of evidence, and information value.
 Model Building:
1. PD Model: Probability of default (percentage). Predict likelihood of customer
default.
2. LGD Model: Loss given default (percentage from not negotiated or not
recovered - when there’s some guarantee - after default and closure). Determine
potential loss when a customer defaults.
3. EAD Model: Exposure at default (credit at the moment of the default). Calculate
exposure at the time of default.

These models will help create a scorecard for loan application decisions.

 Expected Loss: Combine the models to calculate the expected loss across the loan
portfolio, an essential metric for bank management.
 Model Maintenance & Deployment: Learn how to update and deploy your models for
practical use.

Course Benefits:

 Hands-On: You’ll apply the concepts through homework assignments and a final major
project where you build a credit risk model from scratch.
 Resources: Downloadable materials, notes, exercise files, and notebooks are provided.
 Community Support: Participate in Q&A for assistance and to engage with the course
community.

This course promises to be a comprehensive and practical journey into credit risk modeling.
2. What is credit risk and why is it important?

Credit Risk Overview

Credit risk occurs when a borrower (debtor) fails to repay a lender (creditor) for goods or
services received. The lender provides credit, often in the form of money, which the borrower
must repay with interest.

Types of Credit:

 Credit Cards: Borrowers spend up to a limit and repay with interest, generating profit
for the bank.
 Home Ownership Loans: Borrowers take loans secured by property; if they default, the
bank can seize and sell the property to recover the debt.
 Asset Financing: Businesses use loans to acquire equipment, paying in installments
instead of upfront costs.

Understanding Credit Risk: Credit risk is the chance that a borrower won’t repay their debt,
resulting in the lender losing the principal and interest, and incurring collection costs. Default
occurs when the borrower can’t meet repayment requirements.

Mitigating Credit Risk:

 Collateral: Lenders may require collateral (e.g., property) to cover the debt.
 Risk-Based Pricing: Higher interest rates are charged to borrowers with higher credit
risk.

Example: The 2008 Financial Crisis Failure to accurately assess borrowers' default probabilities
can have severe consequences, as seen in the 2008 financial crisis. During this time, financial
institutions lent to subprime borrowers with high default risks. When these borrowers
defaulted on their mortgages, mortgage-backed securities lost value, leading to massive losses
for banks like Lehman Brothers, which went bankrupt. This crisis highlighted the critical
importance of managing credit risk in the financial system.

3. Expected loss (EL) and its components: PD, LGD and EAD

Credit Losses and Expected Loss Calculation

Lenders typically expect some level of credit loss over time due to the inherent credit risk with
every borrower. These losses are influenced by:
  Borrower-specific factors
 Economic conditions
 A combination of both

Lenders may also face unexpected losses caused by adverse economic events, though these are
rare.

Expected Credit Loss (EL): This refers to the anticipated loss from lending to a borrower,
calculated using three components:

1. Probability of Default (PD): The likelihood that a borrower will default (fail to repay
their debt).
2. Loss Given Default (LGD): The proportion of the loan that the lender will lose if the
borrower defaults.
3. Exposure at Default (EAD): The total value the lender is exposed to at the time of
default, or the maximum loss.

These components are multiplied together to estimate the Expected Loss (EL).

Example Calculation:

 Loan amount: $400,000 (80% of $500,000 house price)

 Outstanding balance: $360,000
 PD = 25% (1 in 4 homeowners have defaulted in previous years)
 LGD = 5% (bank recovers $342,000, so loss is $18,000)
 EAD = $360,000

The expected loss is calculated as: PD × LGD × EAD = 25% × 5% × $360,000 = $4,500

This is the lender's expected loss for this exposure.

4. Capital adequacy, regulations, and the Basel II accord

Capital Adequacy Regulations and the Basel II Accord

The 2008 financial crisis highlighted the risks of inadequate regulation, leading to widespread
bankruptcies and economic disruptions. To prevent such crises, regulators impose capital
adequacy requirements on banks to ensure financial stability.

Role of Banks:

 Banks are essential to the economy, facilitating savings and credit. If people lose
confidence in banks, it would disrupt the entire system.
  Regulators require banks to maintain enough capital to absorb losses from loan
defaults, whether caused by individual borrower issues or broader economic
downturns.

Capital Requirements:

 Banks must hold capital as a proportion of their assets, often loans.

 The capital requirement accounts for the risk associated with each asset, resulting in
risk-weighted assets.
 The capital adequacy ratio (CAR) compares a bank’s capital to its risk-weighted assets,
ensuring it can absorb potential losses.

Basel II Accord:

 The Basel II accord outlines the capital banks need to hold and how it relates to their
risk-weighted assets.
 Pillar 1 addresses Minimum Capital Requirements to cover credit, operational, and
market risks.
 The accord proposes three approaches for modeling credit risk:
1. Standardized Approach (SA)
2. Foundation Internal Ratings-Based (F-IRB) Approach
3. Advanced Internal Ratings-Based (A-IRB) Approach

These frameworks guide how to calculate credit risk and determine necessary capital to
maintain financial stability.

5. Basel II approaches: SA, F-IRB, and A-IRB

Basel II Approaches to Credit Risk Modeling

The Basel II Accord provides three approaches for banks to calculate credit risk and determine
how much capital they must hold to cover potential losses. These approaches allow banks to
model credit risk with varying degrees of precision, each providing different levels of flexibility.

Overview of the Approaches

1. Standardized Approach (SA):

o This approach is based entirely on external credit ratings from agencies like FICO,
S&P, Moody's, and Fitch.
o The capital that banks must hold is determined as a percentage of their total
exposure, depending on the credit rating of the borrower (e.g., governments,
firms, or retail exposures).
 2. Internal Ratings-Based (IRB) Approaches:
o Foundation IRB Approach:
 Banks can model the Probability of Default (PD) (the likelihood of a
borrower defaulting).
 However, the Loss Given Default (LGD) and Exposure at Default (EAD)
are provided by regulators.
o Advanced IRB Approach:
 Banks can model all three components of expected loss: PD, LGD, and
EAD, using their own internal data and credit risk assessments.

Capital Requirements under the Standardized Approach:

 Sovereign Debt (for governments): Banks hold capital as a percentage of the exposure
to governments based on credit ratings. For example, an AAA-rated government might
require 20% of the exposure to be held as capital, while a lower-rated country (e.g., B+)
might require 100%.
 Firm Debt (for companies): Similar to sovereign debt, banks use credit ratings to
determine the required capital. For example, AAA-rated firms might need only 20% of
the exposure to be held as capital, while lower-rated firms require a higher percentage.
 Retail Exposures: For loans like credit cards and consumer loans, banks are required to
hold capital equivalent to 75% of the exposure.
 Residential Property Loans (Mortgages): Banks must hold 35% of their exposure in
capital for mortgages.

Benefits of the Internal Ratings-Based (IRB) Approaches:

 Precision in Capital Allocation: IRB approaches allow banks to use their internal data to
model credit risk more accurately, reducing the amount of capital needed to cover each
exposure. This more precise calculation enables banks to allocate resources more
efficiently, potentially increasing the volume of loans they can issue and enhancing
profitability.
 Flexibility: Unlike the Standardized Approach, which applies fixed percentages to all
exposures within a category, the IRB approaches allow banks to tailor their capital
calculations to the specific risk associated with each loan or borrower. This flexibility
benefits banks by reducing capital requirements for lower-risk exposures, freeing up
capital for new business.

Transitioning from Standardized to IRB Approaches:

 Banks typically start with the Standardized Approach because it is simpler and relies on
external credit ratings. Over time, banks may transition to the Foundation IRB and
Advanced IRB approaches as they gather more internal data and develop more precise
models. Is that a decision of the bank??
  Why Transition? Holding 75% of retail loans as capital under the Standardized Approach
may be overly conservative. By using internal ratings and more detailed data, banks can
more accurately estimate their capital needs and potentially reduce the capital they
hold, allowing them to offer more loans and generate higher profits.

Conclusion:

The Basel II Accord offers a flexible framework for banks to manage credit risk and determine
how much capital they need to hold. By choosing the appropriate approach based on the level
of available data and internal modeling capabilities, banks can ensure they maintain enough
capital to absorb losses while optimizing their ability to lend and grow their business. The
transition from the Standardized Approach to the IRB approaches allows banks to use internal
data for more precise risk management and capital allocation.

6. Different facility types (asset classes) and credit risk modeling approaches

Different Facility Types and Credit Risk Modeling Approaches

In this lesson, we explore how different facility types (products like loans or credit cards) and
borrower types (individuals, corporations, countries) influence the way credit risk is modeled
and how capital requirements are calculated under the Basel II framework.

Facility Types and Borrower Types:

 Facility Types: These refer to the various products that a bank offers, such as consumer
loans, mortgages, and credit cards. These products differ in terms of capital
requirements and the type of information that banks use to assess credit risk.
 Borrower Types: This refers to the type of borrower that applies for credit, such as
individuals, small and medium enterprises (SMEs), or corporations.

Capital Requirements Based on Facility Types:

Under the Basel II standardized approach, different borrower types and facility types are
treated differently:

 For example, retail loans like consumer loans may have a capital requirement of 75% of
the exposure, while mortgages have a lower capital requirement of 35%.
 Corporate loans or sovereign debt have different capital weights based on the
borrower’s credit rating.

Credit Risk Modeling Approaches:

1. Internal Ratings-Based (IRB) Approaches:

 o Under the IRB approaches, banks calculate the Probability of Default (PD), Loss
Given Default (LGD), and Exposure at Default (EAD) using internal data, which
provides a more accurate and tailored risk assessment.

o The Foundation IRB approach allows banks to model PD, but regulators provide
LGD and EAD.
o The Advanced IRB approach allows banks to model all three components (PD,
LGD, and EAD).
2. Standardized Approach:
o In the Standardized Approach, external credit ratings are used to assess credit
risk and calculate capital requirements.

The Role of Data in Credit Risk Modeling:

 Data Availability: The type of borrower or facility influences the availability of data. For
example, there is often more data available for individual loans (retail) than for
corporate loans due to the large number of individual borrowers.
o Retail Loans: For individual borrowers, banks have access to demographic
information (age, sex, income, education) and product-specific data (loan term,
interest rate, credit limit, etc.).
o Corporate Loans: For companies, banks use financial statements, ratios (e.g.,
return on equity, current ratio), and other qualitative data like industry and
company size to assess risk.

Building Credit Risk Models:

1. PD Modeling:
o For modeling Probability of Default (PD), binomial logistic regression is
commonly used. This method predicts the likelihood that a borrower will default,
based on the available data.
2. LGD and EAD Modeling:
o For modeling Loss Given Default (LGD) and Exposure at Default (EAD), beta
regression is typically employed. Beta regression is suitable for modeling
variables that are constrained between 0 and 1, such as recovery rates or loan-
to-value ratios.
3. Types of Models:
o Application Models: These are used to estimate the creditworthiness of a
borrower at the time of application. They may include application-specific
characteristics like the borrower’s financial situation, credit history, and the loan
terms.
o Behavioral Models: These models predict the future behavior of a borrower
based on both application data and observed behavior after the loan has been
granted, such as default rates or payment history.
 4. Simplified Presentation for Front Office Users:
o Statistical models can be presented in a simplified way using scorecards, which
assign scores based on the output of a PD model. Scorecards are user-friendly
tools that make it easier for non-technical staff (e.g., credit agents) to
understand and use complex credit risk models.

Risk-Based Pricing:

 Banks use credit risk models to perform risk-based pricing, where the interest rate
charged to a borrower is determined by their credit risk. Riskier borrowers will face
higher interest rates to compensate for the increased risk of default.

Tools and Software for Credit Risk Modeling:

 Common tools used for statistical modeling include SPSS, SAS, and R. However, in this
course, we will focus on Python, a popular and growing programming environment for
data analysis, which is widely used by data scientists and financial analysts for building
predictive models.

Conclusion:

This lesson emphasized how the facility type and borrower type affect credit risk modeling and
capital requirement calculations. It also introduced the basic approaches to modeling credit
risk, including binomial logistic regression for PD, and beta regression for LGD and EAD.
Furthermore, we discussed the importance of data in credit risk modeling and the tools banks
use to calculate and manage credit risk.