Managing credit risk is very essential as it provides some knowledge of how much exposure do the banks have with counterparties versus how much counterparties have against the contracts. Assessing credit risk help banks to control the customer defaults and adjusts their capital.
Credit risk can lead to other different risks like country risk, systematic risk; market risk etc. For e.g.: when banks started lending to the subprime borrowers and took the new subprime debts during 2007-2008 financial crises.
Credit rating agencies gave AAA ratings to a lot of mortgaged backed securities. This somehow led the investors to sell the bundled debts to other banks and investors around the world. This led the interest rate to rise and cheap loans to disappear and borrowers started to default on their mortgages. This tells that credit risk measurement plays a vital role in the economy. And because of the 2007-2008 financial crises, the Basel committee implemented Basel III Accord strengthening the risk management of the banking sector. Also, Basel II had pivotal role in 2007-2008 financial crises.
Basel II allowed banks to create own internal rating system to estimate the risk exposure. FIs prioritised the maximisation of banks equity and underrated the credit risk (Benink and Kaufman, 2008). There is always a factor of risk when banks offer different type of financial products such as mortgages, credit cards, loans etc. to the borrower. Credit risk can be defined as a risk due to the borrower’s failure to repay the loan principal and interest. The traditional approach of banks toward credit quality is assessed using the qualitative based system; the basic 5Cs (Capacity, Collateral, Capital, Conditions, and Character) of credit analysis, which purely depends on the loan officer whose judgment, can be highly subjective.
Most financial institutions used to depend on the bank experts or loan officers. Credit risk measurement and modelling have developed significantly in past 20 years. Financial institutions have gradually moved away from the subjective system and are focused on the objective based systems. There are many models that banks use in the process of the loan application. Different banks use different models. Objective based systems like Quantitative Model are used to minimise risk for the shareholders. Credit rating can be an example of credit measurement.
The outcome of credit rating is an indication for banks to make decision for credit scoring. Credit modelling such as credit scoring models is used to evaluate and analyse the creditworthiness of an enterprise. The credit agencies take different key accounting ratios of the borrowers and compare it with the market. Zhang (2016) states that JP Morgan’s Credit Metrics as the most outstanding model since 1990s. This procedure is carried out using the credit rating while estimating the expected probability of default of at least 5 years of data and migration of the credit. However, this model is not suitable to evaluate the future performance as it uses the historical data of a company. For e.g.
a company, which is fairly new, won’t have enough data to estimate the score. The size of the enterprise also plays huge impact on the lending process. Before making a loan to any individual or an enterprise, banks look at the relationship between the bank and the customer and other common actions taken by the customer. Loans to SMEs are differ from loans to large corporations, retail borrowers, sovereign and non-sovereign borrowers.
SMEs plays vital role in boosting the economy of the country. Since 1997 Asian financial crisis, banks in Asia restrained themselves when it comes to lending to SMEs. Also, because of the Basel III capital requirement, start-up or small enterprises face difficulty when borrowing loans from the bank (Taghizadeh-Hesary et al., 2015 cited Hirano 2013, p.19). Because of the strict SME lending from the banks, the Credit Risk Database (CRD) was created in Japan.
This project is fully supported by the government where “52 credit guarantee corporations collected the data of Japanese SMEs”. The data includes financial and non-financial information of SMEs throughout Japan. The purpose of this database is to collect solid data from SMEs and examine it thoroughly, building up the credit scoring models and provide the information on overall situation of the company. This database rates SMEs on basis of their financial and non-financial data and classifies into three groups.
Group 1 contains SMEs that are financially healthy, Group 2 contains medium risk SMEs and Group 3 contains financially risks SMEs. Financial institutions can use this credit rating information in decision-making process. This process benefits both FIs and SMEs because FIs will be more likely get the principal and interest and SMEs can get loans on lower interest. If the performance of the higher risk SMEs advances over time, those SMEs can get lower level of interest. Similarly, Altman et al. (2007) also suggests that using separate credit scoring and rating system for SMEs like CRD will be easier for the banks to carefully analyse and proceed to manage the risk exposures.
Moody’s Analytics states “SMEs is one of the most challenging tasks in banking”. Even though FIs have applied best credit risk model before considering any type of loans or investments in SMEs, there’s always a probability that non-performing loan will occur. Therefore, in order to reduce the risk exposure and default from SMEs, the existing model needs to be developed. Several researchers have applied structural model approach, which was pioneered by Merton (1974). This structural model approach was further implemented in Moody’s KMV model. This is also called KMV-Merton model. Bharath et al. (2004) assess the accuracy and contribution of KMV-Merton model.
They found that the KMV-Merton model does not generate appropriate statistic for the probability of default. According to Chen et al. (2010), it is very difficult to use credit modelling developed my western countries like JP Morgan’s Credit Metrics, McKinsey’s Credit Portfolio view due to the insufficient credit data of the enterprise. However, they believe that KMV model is the most appropriate for SMEs in China, as it doesn’t rely on the historical probability credit rate but mostly depends on the enterprise’s asset and calculated Expected Default Frequency. The assumption of KMV model is that an enterprise will only default when that value of its asset is lower then the liabilities. Chen et al. (2010) recommended new model to improve the KMV model with adjustment of different parameters to measure credit risk of SMEs. They confirm that the model is coherent with reality and is flexible to the default point.
Also, the distance of default is very much closer which means that the model is very effective to identify the credit risk in Chinese SMEs. Looking at earlier studies, the most approached parameter in Credit risk modelling is probability of default. In terms of recent works, academics and researchers have been using loss given default and recovery rate. The study by Kosak et al. (2010) focused on determinants loan given default of the SME loans in Slovenia.
The estimation of LGD was carried out by the discounted cash flow. Their findings suggest that the consistency of LGD can be explained by different determinants such as available loan ratings, size of debt, collateral and different industrial sector. They also mention that the investigation that allows evaluating the explanatory determinants can drastically progress the credit risk management process in a bank. Dietsch and Petey (2002) developed a credit risk model with which focused on one factor probit model to analyse the probability of default of SMEs. There study showed that probability of default has positive correlation with the company’s asset. They also implied that there should be size specific model inside the SMEs. Banks cannot estimate small retail loans by using the models that estimate for the large or SME loans.
Retail credit facilitates loans to individual or first time borrowers. Determinants like individual’s income, credit history etc. is acknowledged to manage each loan. The advanced IRB approach implemented by Basel II grants banks to analyse the internal risk elements. Banks are obliged to use the quantitative models to estimate probability default (PD), loss given default (LGD) and exposure at default (EAD). These parameters are related to the corporate models of credit risk. Due to the differences like repayment behaviour, these parameters cannot be used directly to the retail borrower portfolios.
When lending to an individual, banks need to take some risk factors like loss of income, job etc. in consideration (Thomas, 2003). Traditional models like neutral network, internal rating systems and credit scoring models are used to estimate the probability of defaults. Eletter et al. (2010) develop a neutral network based model to evaluate the credit risk of individual borrowers and reduce customers’ defaults in Jordanian banks. They suggest neutral network based model as a successful technology in evaluating individual loans.
However, Handzic et al. (2003) point out some disadvantages of using this type of approach. They point out that the parameters used in this model are not fixed and takes long to find perfect parameters of any neutral network model. This type of model can require loads of training and cases. Likewise, Altman et al. (1994) analysed over 1000 Italian firms to check if neutral network was better than statistical techniques. They concluded that neutral networks were not better than statistical methods and both resulted about same level of correctness.
They suggested using both methods in complex applications where neutral method would be valuable. Nowadays, banks have moved away from neutral network based model. Many FIs in Japan used to use real estate collateral to cover the default. FIs were forced to seek reliable framework of credit risk due to the boom in macroeconomic environment (Oyama, 2005). However, “real estate collateral continues to be the major form of collateral”. When estimating the loss given default (LGD), the creditworthiness of borrower is reflected by the value of collateral. In USA, banks were better at evaluating the borrower’s probability of default compared to Japan where LGD was mostly used.
Almost all FIs use Credit scoring model to estimate the PD of consumer loans. Fair and Isaac Corporation developed credit score model like FICO score in late 1950s. Since then there has been several redevelopments of the model. FICO model has become widely used model to determine consumers’ credit risk. There are an increasing numbers of score attached to each customer. Credit bureaus like Equifix, Experian etc. use this model to assess mortgage score, personal loans and many more. Researchers like Horkko (2010) applied a logistic regression model to develop the numerical credit scoring system using 14,595 observations of customers from Finland.
The main purpose of her finding was to explore “if socio- demographical and behavioural variables have effect on default”. She suggests that the credit-scoring model that she developed can be used in financial institutions to assess the risk. Similarly, Andrade et al. (2007) applied reduction form approach modelling to assess the loss distribution of credit risk portfolios by using Monte Carlo Simulation in Brazil.
They suggest this model as a flexible model that use different statistical distributions to model credit loss in various parts of portfolio. The similarities between these proposed models are that they both were concerned in socio-demographical and behaviour variables, which are especially dedicated to consumer loans. These models can be used in underdeveloped countries where default risk is very high. However, using the approach by Andrade et al. (2007) in large portfolios is very limited because of the simulation of joint defaults, it takes a lot of time and computational power. There are many modern methods that banks use to model and measure credit risk.
Measurement like option theoretic structural approach is inspired by Merton (1974). Models like KMV ‘s portfolio manager, Moody’s Risk Calc use this approach. The suitable method to assess credit risk is Credit Risk plus which use reduced form approach (Allen et al., 2004). This model requires no credit spread data; just minimal data of mean loss. Crouhy et al. (2000) describe this model as a very easy model to implement.
The assumption of this model is either a consumer is in default or not in default. However, some background factors like future interest rates, which are represented by the volatility of default rates, can lead to exclusion of the factors. According to Altman et al. (2007) there should always be separate models for both SMEs and large corporations. Using same model might cause chaos on both corporate portfolios. Financial institutions have been using the traditional way i.e.
internal ratings to measure credit risk for large corporations. In a survey of 30 financial institutions done by The Bank for International Settlements (2000), found that about 96% of large corporation were assessed using internal rating system. Treacy et al. (2000) carried out survey of 50 largest banks in the USA and found that different banks use distinctive internal rating models.
They also discovered that the qualitative factors play a vital role in regulating the ratings of SME borrowers. However, this finding is not applicable to large corporations. For the large firms the loan officer especially set the ratings using credit-scoring models. Altman (1968) Z score model is a traditional model use to provide default risk estimation. This model is mostly used for corporate borrowers like large corporations.
It is based on five financial ratios as to when a corporation may go to bankruptcy. If the Z score is of higher value then the default risk is low. If the Z score is low or negative then there is evidence that the firm might default. However, this model does not take consideration of many factors like business cycle effects. Also, the historical data used in this model might not hold sufficient information to estimate the risk. Due to these complications, there has been development of new techniques: RORAC models, KMV’s credit monitor model.
RAROC model was developed in late 1990s. Almost all FIs have developed this model to estimate lending procedure. Bankers Trust (acquired by Deutsche Bank in 1998 pioneered the RAROC (risk-adjusted return on capital) (Saunders and Cornett, 2011).
There are only 3 popular organisations of credit rating agency: Fitch, Moody’s and S;P. Credit rating agency is a high-priced business. Financial institutions have become excessively dependent on the judgement by the agencies. Sometimes financial institutions blindly accept the ratings from agency, which lead to the default on loans by borrowers. Basel II Accord allowed financial institutions to develop their own credit risk model to assess the probability of risk. Internal system helps bank to assess risks they acquired through lending.
Due to the financial crisis, the number of bankrupt borrowers increased. Assessing credit risk became an important subject matter. Banks started to seek reliable framework of credit risk. Since then most of the banks have there own method of estimating default risk through IRB approach.
One of the motives behind banks’ to develop own model is to decrease the managerial work. This makes the lending decision efficient. Cost efficiency can also be a factor of banks’ inclination. There has been a rapid development of credit risk measurement and modelling since the 2007-2008 financial crisis. There are many applications of model created for assessing the risk management of firms or retail borrowers. The risk models are created based on assumptions and theories so financial institutions should treat risk measures with caution even though it is generated by highly regard models. Increasing number of competition and macroeconomic environment are forcing banks to develop effective credit management process. The traditional credit risk measurements are totally different than the modern ones.
Every country has distinctive way of assessing the credit risk. Banks in China prefer KMV model to assess risk of SMEs whereas CRD database is popular in Japan. The value of consumer’s asset cannot be identified, as marked to market for retail borrowers so cannot be considered as a reliant predictive.
Hence, Merton-like model cannot be applied when assessing credit risk of retail borrowers. However, banks need to provide a model with clarification that includes elements of consumer lending. Financial institutions are sensitive when it comes to lending to large corporations. They carry out every procedure to assess the probability of default of large corporations. This might be because they are investing huge sum of money.
Consumers and small businesses are moving away from traditional lending and moving towards Fintech lending. Some lenders have developed their own credit risk assessment using big data on their algorithms. This new evaluation of credit risk can give consumers and small business an opportunity to enhance credit access.
