Loss given default (LGD) is another of the key metrics used in quantitative risk analysis. It is defined as the percentage risk of exposure that is not expected to be recovered in the event of default.
BBVA basically uses two approaches to estimate LGD. The most usual is that known as “workout LGD”, in which estimates are based on the historical information observed in the entity, by discounting the flows that are recorded throughout the recovery process of the contracts in default at a certain time.
In portfolios with a low rate of defaults (low default portfolio, or LDP), there is insufficient historical experience to make a reliable estimate using the Workout LGD method, so external sources of information have to be combined with internal data to obtain a representative rate of loss given default.
In general, the portfolios managed by the Bank show bimodal behavior in terms of loss given default: in other words, following default, the recovery of debt is in some cases total and in others, nil. Chart 11 shows the distribution of the LGD of mortgage portfolios in BBVA Spain, filtering the most recent defaults to prevent the short amount of time to recover from slanting the results.
The LGD estimates are carried out by segmenting operations according to different factors that are relevant for its calculation, such as the default period, seasoning, the loan to value ratio, type of customer, score, etc. The factors may be different according to the portfolio being analyzed. Some of these are illustrated below with examples.
- For contracts already in default, an important factor is the time since the default on the contract. The longer the contract has been in default, the lower the recovery of the outstanding debt pending. For the purposes of calculating the expected loss and economic capital, contracts not in default are also imputed a LGD comparable to contracts that have just defaulted (Chart 12).
- The seasoning is the period elapsed from the origination of the contract to the default. This factor is also relevant, as there is an inverse relationship between LGD and seasoning (Chart 13).
- Loan-to-value ratio: Internal analyses show that LGD depends on the relationship between the amount of the loan and the value of the collateral (loan-to-value, LTV), a characteristic feature of the mortgage portfolio (Chart 14).
- Customer type: In the specific case of products for companies the customer type has proven to be a relevant factor. Therefore, a LGD estimate has been obtained for each size of company: corporations, large companies, SMEs, etc.
- Score: The credit rating of contracts may be used to estimate LGD, as there is a positive correlation between score and LGD (Chart 15).
- Exposure at default (EAD): The exposure of contracts at the time of the default is positively related to LGD (Chart 16).
Progress in building LGD scorings and ratings is becoming increasingly important for adapting LGD estimations to economic and social changes. These estimates allow new factors to be included without losing the robustness of the information and obtain models that are more sensitive to improvements or deteriorations in the portfolio. BBVA has already begun to work on incorporating these modifications in the internal models used.
In the BBVA Group, different LGDs are attributed to the outstanding portfolio (performing and non-performing), according to combinations of all the significant factors, depending on the features of each product and/or customer.
This can be seen in Chart 14, where LGD is explained according to the seasoning of the contract and its loan-to-value (LTV) ratio.
Finally, it is important to mention that LGD varies with the economic cycle. Hence, two concepts can be defined: long-run LGD (LRLGD), and LGD at the worst moment in the cycle, or Downturn LGD (DLGD).
LRLGD represents the average long-term LGD corresponding to an acyclical scenario that is independent of the time of estimation. This scenario should be applied when calculating expected losses. DLGD represents the LGD at the worst time of the economic cycle, so it should be used to calculate economic capital, because the aim of EC is to cover possible losses incurred over and above those expected.
All estimates of loss given default (LGD, LRLGD and DLGD) are performed for each portfolio, taking into account all the factors mentioned above. However, no LRLGD or DLGD estimates are made in portfolios in which the loss given default is not significantly sensitive to the cycle, as recovery processes cover extended periods of time in which the isolated situations of the economic cycle are mitigated.
In addition to being a basic input for quantifying expected losses and capital, LGD estimates have other uses for internal management. For example, LGD is an essential factor for appropriate price discrimination. Similarly, it can be useful for determining the approximate value of a non-performing portfolio in the hypothetical event of outsourcing its recovery, or prioritizing potential recovery actions.