In our first blog we looked at Taking Loss Given Default Estimation to the Next Level: An Aspiration for All Creditors, Not Just Banks(read full blog). We now look at how our credit models have evolved over time to streamline analysis and meet the needs of a larger set of users.
Ascertaining the creditworthiness of borrowers is a critical factor for the success of debt capital markets and bank lending. This is particularly the case in today’s ultra-low-yield credit environment, which requires the accurate and efficient analysis of large numbers of counterparties to identify appropriate opportunities and risks.
The name S&P Global, formerly Standard & Poor’s, is synonymous with credit ratings, which are one important component of global credit market analysis. Credit ratings, as well as associated historical default rates, are used by different organisations for diverse purposes (e.g., listed entities reporting under IFRS9 or insurance companies’ investment decisions). Built upon a solid foundation of credit ratings produced by S&P Global Ratings, the S&P Global Market Intelligence Credit Cycle Projection Overlay (or CCPO) is designed to estimate probabilities of default (PDs) for entities in different sectors and geographies.
Complementing the credit picture, the S&P Global Market Intelligence Loss Given Default (LGD) Scorecards estimate the proportion of debt which will not be recovered post default. Given LGD is estimated on an exposure basis (on each bond or loan), the number of LGD estimates for a portfolio of loans or bonds is typically a multiple of the number of entities or issuers. It is often the need for efficiency in assessing large portfolios of bonds or loans that contributes to entities taking analytical “shortcuts” when estimating LGD. Analytical shortcuts may include the use of average historical losses as estimates of future losses or quantitative models built and tested on relatively small samples, which often do not perform well.
Automated credit risk analysis, decisioning and surveillance
It is worth stressing that automation should not entail the total exclusion of human intelligence from the credit analysis and surveillance process, but rather enable analysts to spend more time delving deep into potential issues, supported through a largely automated process.
The principal components of credit risk analysis are borrower/issuer data, credit ratings (and related PDs) and LGD. Thus, the partial/full automation of these components would need to be considered in order to streamline the credit risk analysis and surveillance process.
For data, our clients rely on our award-winning platform of fundamental company data. The platform captures hundreds of data points for more than 62,000 public and 10 million private companies. It also captures more than 57,000 bank and 1,200 insurance company data. You can learn more here >.
For the PD analysis, a common approach involves assigning a credit rating to each borrower or bond issuer, and then associating a PD with a rating grade. The rating grade is usually based on credit rating agency default rates or internal default data on borrower defaults at banks. Many non-bank investors typically use public ratings for this purpose.
As is the case in recent years, many investors have ventured into unrated (i.e., not publicly rated) debt looking for greater yields. This has led to an increased demand from bank and non-bank investors for credit assessment Scorecards, which are analytically aligned to criteria used to assign public ratings. For over 20 years, our clients have benefited from the structured approach offered by the sector-specific S&P Global Market Intelligence Credit Assessment Scorecards (“Scorecards”). The analytical consistency between issuer scores produced by our Scorecards for unrated issuers and ratings for publicly-rated issuers is particularly important for the purpose of increasing efficiency of analysis, whilst maintaining a high level of analytical rigour.
Given a portfolio of issuers that are either scored using our Scorecards or are rated publicly, the attention of investors can then shift to assigning PDs both to through-the-cycle (TTC) and point-in-time (PIT)). The CCPO is an Excel®-based framework that supports the estimation of both TTC and PIT PDs for issuers in different sectors and geographies. The estimation process uses a combination of six factors:
- Two country-level macroeconomic factors (e.g., GDP growth).
- One country/sector-specific market factor (e.g., equity price changes).
- Two commodity-related factors.
- One portfolio-level factor (e.g., momentum of changes in credit ratings).
These factors are pre-filled for 100+ countries and updated quarterly by S&P Global Market Intelligence analysts, enabling users to automatically estimate PIT and TTC PD term structures for entire portfolios in an efficient and robust manner. All that is required from the user is the credit rating, country of domicile and sector of operations for each issuer (e.g., BBB+/Germany/Bank). The CCPO has been in use for over five years at banking and non-banking clients (e.g., insurers) for use cases including regulatory and economic capital estimation, IFRS9 credit loss estimation and risk-adjusted pricing.
The final piece involves the LGD analysis of all loans or bonds. As eluded to earlier, this is the most challenging piece, as the number of loans or bonds are large and potentially from multiple asset classes or sectors. Despite these difficulties, for the past 20 years, our clients have managed to employ our LGD Scorecards for facility- or bond-specific LGD estimation.
The LGD Scorecards are a family of cutting-edge, sector-specific LGD models that estimate LGD based on issuer/borrower and bond/facility specifics. These Scorecards have been tested on over 2,000 historical losses from around the world and have managed to predict the “correct” loss range for over 1,300 bonds and loans, with an average difference (predicted versus actual) of only 0.9%.
Pre-2017, our LGD Scorecards were mainly used by banks in conjunction with borrower and facility inputs, sourced internally within the bank. The LGD Scorecards were very much reliant on input data, thus making the use of the Scorecards far more challenging for non-banking institutions. The latter typically lacked the resources to internally provide all needed risk factor inputs. As of January 2018, we have managed to link our most popular LGD Scorecards to S&P Global Market Intelligence fundamental databases, unleashing the full potential of our Scorecards for the entire market (not just banks). The estimation of LGD for over 130+ sovereigns, and for the majority of rated and listed companies, is automated, with no risk factor inputs required from the user for our Sovereign, Banks, Insurance and Corporate LGD Scorecards.
The employment of the CCPO and LGD Scorecards can help investors/lenders streamline the credit risk analysis and surveillance process, whilst reflecting decades of credit experience from S&P Global. For senior executives and managers, the time and resources saved from the use of these solutions can be deployed to higher-value activities. Resource savings aside, the benefits from a more informed and timely decision-making process are immeasurable.
Flexible and transparent delivery
The Credit Assessment Scorecards, CCPO and LGD Scorecards are delivered in Excel and can either be used within that format or “coded” into internal credit risk software. The delivery in Excel ensures that the methodology is housed in a glass-box, with weights, combination protocols and benchmarks all visible, enabling a transfer of intellectual property. This fulfils a key requirement from internal/external audit and regulatory bodies, whilst supporting the internal ingestion of cutting-edge credit risk methodologies.