Determining Estimated Loss in loans based on probability of default and loss rate

Today's financial market witnesses the strong development of financial and non-financial institutions, accompanied by the urgent need for credit risk management. Accurate measurement of this risk plays a pivotal role in ensuring safety and stability for the operations of lending institutions.

To build modern credit risk measurement and assessment tools according to the Internal Rating Based Approach (IRB – Internal Rating Based Approach), banks choose to implement the Basel Accord on “international capital standards”. In which, credit risk measurement plays an important role in estimating losses and the required reserve capital.

In addition to self-quantifying risk, organizations can apply two main methods to manage credit risk: hedging and trading risk. The development of the credit derivatives market, along with the need for effective risk management, has promoted new methods to estimate credit risk.

Most risk hedging models include the 2 most basic components to determine credit risk: Probability of Default (PD - Probability of Default) and Loss rate estimated (LGD - Loss Given Default).

Today's Viet Hustler article will focus on the most basic theoretical aspect in calculating Loss due to default and the most common methods in credit risk assessment.

Determining estimated losses of financial institutions in cases of bad debts

According to Basel II requirements, banks must build measurement tools such as:

• PD – Probability of Default: used to measure the customer's ability to not repay debt within a period (usually one year);

• LGD – Loss Given Default: the proportion of capital lost on total outstanding debt if the customer defaults;

• EAD – Exposure at Default: total outstanding debt at the time the customer fails to repay.

Through the above variables, the bank will determine estimated loss (EL: Expected Loss), via the formula:

We will examine in turn the 3 indicators that constitute the above formula

1. PD - Probability of Default

PD – Probability of Default used to measure the customer's ability to not repay debt within a period usually one year.

PD is calculated based on the customer's credit history data for at least 5 years prior (Basel II requirement): including repaid debts, on-schedule debts, and unrecoverable debts

This data is divided into the following 3 groups:

• Group financial ratios: for example

  • evaluations by credit rating agencies on customers (for corporate customers)…

  • or previous personal credit ratings by credit institutions (individual customers…)

• Group of qualitative non-financial data: management level, ability to research and develop new products, data on industry growth potential…

• Warning data: for example, phenomena signaling the inability to repay the bank such as deposit balance, overdraft limit,.. and financial data such as income, assets…

Many studies will focus on models to determine the probability of default of an organization / individual based on the above data.

2. EAD - Exposure at Default: Total outstanding debt of the customer at the time of default

EAD is calculated based on 2 approaches:

a. F-IRB - Foundation Internal Rating Based evaluation approach

• The EAD calculation method under F-IRB is stipulated by market regulators.

  • Basic factors included in EAD calculation: underlying assets, expected future recovery value, credit type and details in the commitment.

• For on-balance sheet transactions, EAD = actual loan amount.

  • Assets and deposits at the bank itself can be deducted.

• For off-balance sheet items, consider 2 main factors:

  • (1) Debt arising from transactions with uncertain future drawdowns, such as commitments and revolving credits, and

  • (2) Debt arising from OTC foreign exchange, interest rate and equity option contracts.

b. Advanced Internal Ratings-Based approach (A-IRB)

• Banks are free to decide how to calculate EAD for each loan as appropriate.

  • => Banks have more flexibility. For example: focus on more serious risks, or use detailed data on transactions and borrowers…

• However, to use their own EAD calculation method, the bank must demonstrate to the regulatory authority the ability to meet the requirements for reliability and accuracy of these estimates.

3. LGD - Loss Given Default: Estimated loss rate

Estimated loss rate (LGD - Loss Given Default) is the proportion of capital that may be lost (on total outstanding debt) if the customer fails to repay at a certain point in time.

• In reality, LGD is not simply based on loan principal loss (principle) but also other losses arising when the customer fails to repay, for example:

  • Overdue interest not paid (interest)

  • Administrative costs incurred such as: handling collateral assets, legal services...

  • However, these costs are relatively small compared to the principal loss, so for simplification, they can be omitted.

Previously, studies often focused on modeling and estimating PD, with the assumption that LGD is constant.

However, this is completely wrong because LGD can change if the value of collateral changes over time.

• Sometimes the value of collateral increases due to firm-specific or macroeconomic factors. For example:

  • Estimated loss given default (LGD) of real estate purchase loans fell below the 0% threshold during the period before and during Covid due to the significant increase in real estate value (collateral assets)…

    • Partly due to the generous QE policy of the Fed, causing all asset values to increase.

  • However, by the end of 2022-early 2023, LGD unexpectedly increased and nearly reached 80% due to falling house prices from rising interest rates.

• LGD is also based on the initial value of collateral assets at contract signing:

  • Large banks are more prone to loss risk because their loans tend to be less collateralized.

  • Average LGD of small banks fluctuates from 30%, of large banks from 50%

• Therefore, accurately predicting LGD is extremely important for calculating credit risk and implementing appropriate management measures.

Conversely to the estimated loss rate, recovery rate (RR - Recovery Rate) is the proportion of loan capital still recoverable if the borrower defaults (formula below):

• Therefore, in theory, initially, LGD and RR both lie in the range [0,1].

  • However, in reality, LGD of a loan can exceed or fall below the above range at different times: when the collateral value changes!

• Example of RR:

  • Recovery rate (RR) after companies declared bankruptcy (2008 period) was lower than at the start of the crisis (2007).

Three main methods to calculate LGD

1. Market LGD - Market-based loss proportion

• This method is used when credits can be bought and sold on the market.

• Banks can determine the loss proportion of a loan based on the price of that loan on the secondary market after (or near the time) the loans are classified as non-performing. For example:

  • Credit Default Swap contracts on the market,

  • Yields of bonds of defaulted companies (e.g., the case of Evergrande in China)…

2. Workout LGD - Loss proportion based on the process of handling credits that may not be repaid if the borrower defaults.

• Banks estimate future recoverable cash flows (e.g., from collateral assets, from the enterprise's net assets…) and discount them to present value.

• This method requires the ability to accurately forecast the received cash flows (cashflow), expected recovery time period, and discount rate.

3. Implied Market LGD: This method is based on the value of risky bonds equivalent on the market.

FYI. There are many models used to calculate LGD. The LossCals model introduced by Moody's in 2005 is the most popular calculation method, based on applying a multivariate linear regression model:

• The variables used in the regression function include: secured risk factors, industry risk factors, macroeconomic factors, and transition risk factors.

Conclusion

In recent years, methods and models for forecasting and calculating Credit Risk have seen significant developments and improvements. Many new models and methods are used to estimate and evaluate Probability of Default (PD) and Loss Given Default (LGD), to determine potential losses and latent loss levels.

The main factors affecting LGD calculation that need to be considered include: debt seniority (seniority of debt), the impact of business cycles on debt recovery ability, and the impact of industry on debt recovery ability. Each financial or non-financial organization will have its own selection of models and LGD measurement methods depending on the nature of the loans.

The above article by Viet Hustler mainly provides readers with basic knowledge for reading and understanding data on PD or LGD to evaluate credit risks in the market, thereby making correct investment decisions.

References:

1. Misankova, M., Spuchľakova, E., & Frajtova – Michalikova, K. (2015). Determination of Default Probability by Loss Given Default. Procedia Economics and Finance, 26, 411-417. ISSN 2212-5671.

2. Corporate Finance Institute. (n.d.). Exposure at Default (EAD). Retrieved from https://corporatefinanceinstitute.com/resources/commercial-lending/exposure-at-default-ead/