Study On The Loan Allocation And Credit Risk Measurement Of Commercial Bank

The current economy new normal highlights the tremendous pressure on commercial banks brought by the quality of credit assets, and how to strengthen credit risk management is an issue unavoidable in discussion on accelerating business transformation of commercial banks under new situation and an eternal theme explored by commercial banks to build themselves the long-established business. Since 2014, affected by the external economic environment and national industrial policy adjustment and other factors, the procyclical businesses with weak anti-risk ability and deterioration of industry operating conditions have resulted in concentrated outbreak of default risk. Along with economic slowdown, the quality of commercial bank’s credit assets is subject to greater impact, both non-performing loans and NPL ratio show a "rising" trend, and the banking industry enters into the high-risk and low-return development stage. Therefore, combining economy transformation background under the new normal, further discussion of credit risk management of commercial banks has important practical significance.Under the general background of Chinese transformation and the overall framework of the Basel agreement, in this thesis, research focuses on the three stages of risk whole process management, from exploring the ownership source of crdit risk during pre-loan period, to optimize credit risk measurement model during loan period, at last to carry out macro factor measurement of credit risk stress test during post-loan period. It is also the inherent logic idea of this study. As to the risk source analysis during pre-loan period, from the unique prospective of China’s economy transformation, with O-B decomposition method and the quantile decomposition method for the first time, this thesis studies the allocation differences of credit allocation among state and non-state industries at different amnunts and the ownership source of credit risk. The accuracy of measurement of credit risk "expected losses" and "unexpected losses" at a certain confidence level was improved through optimization and improvement of the credit risk measurement models in loans.The "abnormal loss" risk was further measured through macroeconomic factors measurement of credit risk stress testing and by selecting variables with stronger risk transmission in the post-loan management.This thesis is divided into six chapters, and the contents of which read as follows:Chapter Ⅰ, Introduction.This chapter introduces the background, research significance, research ideas and approaches, and structure framework of the thesis.Chapter Ⅱ, literature and theory review. This chapter introduced the theoretical basis of this research literature review through literature review, including basic theory of credit allocation, credit risk connotation and characteristics, and the representative models of modern credit risk researches. Available studies were evaluated, and some current existing research issues were summarized. At the same time, the current situation of credit risk management of commercial banks was briefly analyzed, and the formative stages of non-performing loans and grim situation of risk control were introduced and reviewed, providing necessary background basis for subsequent empirical research and analysis.Chapter III, credit allocation, ownership and credit risk research based on quantile decomposition method. The data from China’s non-financial listed companies in 2002-2012 was taken as the sample to study the allocation situation of different amounts of credits in state-owned and non-state-owned industries and enterprises with the O-B decomposition method and the quantile decomposition method for the first time, as well as the differences in credit risks between them, and the institutional roots of credit risk were analyzed deeply.Chapter IV, Improved credit risk measurement based on Copula function and Monte Carlo simulation. How to measure credit risk of portfolio is one of the difficult problems to be solved in risk management of commercial banks. This chapter improves the Credit Metrics model through Copula function and Monte Carlo simulation. The credit risk measurement of single asset and portfolio asset in Credit Metrics model is discussed, in which the problems need to be addressed is correlation and "fat tail" problem of assets. Combining with Copula function and Monte Carlo simulation, take credit portfolio of real estate, steel, petrochemical three industries as example, the full process of measuring credit risk of portfolio with this method in practice is elaborated.Chapter V, Macro factor measurement of credit risk stress test.Scenario stress test based on macroeconomic factors is to examine the adverse effects of macroeconomic downturn on the quality of commercial bank credit asset. According to different natures of commercial bank credit assets, this chapter divides it into company banking credit assets and retail banking credit assets, to investigate the effect conduction mechanism of macroeconomic factors during credit risk stress test. To study company bank credit assets, this thesis designs the MEF model under the Basel Capital Accord to examine the significance of macroeconomic factors. The retail banking credit assets are mainly researched through default rates of a large number of customers and macroeconomic factors correlation. Finally, the determination of macroeconomic factors in the overall credit assets stress test of commercial banks is comprehensively investigated.Finally, Concluding chapter. The main research conclusions, main innovation and limitations of the full thesis are summarized, and the future research prospect is provided.The main conclusions are as follows:First, the study shows that the state-owned industries and enterprises obtained more credits than non-state-owned industries and enterprises at all amounts, with maximum differences in microcredit, and little differences in medium and large amounts of credits. The credit risks due to ownership factors reduce with increase in the amount of credit. The difference in microcredit stems mainly from ownership, which is the system root cause of credit risk, and the difference in large credits is mainly due to the differences in characteristics of industries. With the market-oriented reforming of the banking sector, the reasonable elements in credit differences from the differences in characteristics of industries gradually increase, and unreasonable elements from ownership discrimination gradually decrease, so the credit risk introduced by ownership factor is gradually lowered.Second, establishment of combination credit risk measurement model has two steps:describing the marginal distribution of single asset, describing the relationship of assets by Copula function. Credit risk of single asset is described by its marginal distribution. This requires that the net present values of single asset under all credit rating states are firstly obtained by discount pricing.Then the threshold of credit assets is calculated with the model of return on assets based on the credit ratingtransition probability matrix of assets, and the mapping relationship between credit rating scenarios and scenarios of return on assets, thus the discrete credit risk values of credit assets under various ratings are transformed into continuous marginal distribution of credit risk. Taking credit assets of real estate, steel and petrochemical three industries as example, researches in this thesis have found that the marginal distribution of its credit risk shows non-normal distribution characteristics.Credit risk of portfolio can be described by Copula function connecting the credit assets, and the correlation coefficients among them can be obtained through parameter estimation of Copula function by IFM. Different Copula functions have differente ffects on the fitting of credit risk of real portfolio, and the goodness-of-fit of Copula functions can be judgedin accordance with Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and the log-likelihood (LL) index. Take credit assets of the real estate, steel and petrochemical industries as example, researches have shown that the fitting results of t-Copula function describing the tail symmetry correlation is approximate to the real joint distribution of credit risk of portfolio. Finally, the mean, standard deviation and percentile of portfolio values under different distributions were simulated with Monte Carlo method, and the difference between the mean and percentile was calculated, thus getting the risk value VaR of credit risk of portfolio under various assumed states.Third, when determining the macroeconomic factor of company banking credit assets stress test, the macroeconomic factors were selected as a stress index, NPL ratio and capital adequacy ratio as a stress bearing index, and operating income as a risk driver. Through the establishment of the three-factor linear regression model, the combined effects of macroeconomic factors on the revenues of the industries were calculated.Then, according to articulation between revenues and other financial statement accounts, the simulation report and simulation statement rating under stress scenarios was established to calculate rating migration, PD variation, LGD variation and five-grade migration, and ultimately the defect rate variation was obtained.When determining the macroeconomic factor of retail bankingcredit assets stress test, considering the fact that the status of the assets of commercial banks’ retail customers is difficult to obtain, and the correlation between retail customers’ personal income and macroeconomic is not as close as that of company banking customers, we consider the direct relationship between the macroeconomic factors and default rates, other than through other intermediate media.The main improvements and innovations of the thesis are in the following three aspects:First, the data from China’s non-financial listed companies in 2002-2012 was taken as the sample to study the allocation situation of different amounts of credit in state-owned and non-state-owned industries and enterprises, as well as the differences in credit risks between them, and in-depth analysis was carried out with the O-B decomposition and quantile decomposition method for the first time. The main contributions of this thesis are as follows:First, study the allocation status of credit between state-owned and non-state-owned economy in the overall distribution according to different amounts and analyze the credit risk at any amount with the quantile decomposition method for the first time, revealing the credit differences between the two sectors of economy and the change of credit risk with the amount, providing more comprehensive and more detailed information about overall distribution of credit, making up the deficiency that existing literatures’ conclusions are ambiguous and not targeted, and providing a basis for proper development of credit policy. Second, the root causes of credit difference between state-owned and non-state-owned economy are decomposed in theory, which consist mainly of differences in characteristics and ownership discrimination produced due to internal and external configuration methods of resources essentially. Ownership discrimination is one macroeconomic factor causing systemic banking credit risk, which provides a theoretical explanation for deep understanding of the system cause of credit risk. Third, study the credit allocation and credit risk between state-owned and non-state-owned economy from the industry perspective for the first time, and enrich and expand the research in this domain.Second, how to measure the credit risk of portfolio of all assets is one of the difficult problems to be solved by commercial banks in risk management. For a long time, two problems have not been satisfactorily solved, one is correlation measurement and the other is "fat tailed" problem of risk loss. This thesis promotes researches on portfolio credit risk measurement by a step from theory to practice using Copula function, gives a set of algorithms combining theoretical and practical data with Monte Carlo model, and provides a new research tool for credit risk measurement of portfolio, with certain innovation and practicality. This thesis also, taking credit portfolio of real estate, steel and petrochemicals three industries as example, elaborates the full process of measuring credit risk of portfolio with this approach in practice.Third, the single-factor model and CPV model are representative models in the existing theoretical models to measure credit risk based on macroeconomic factors, but they are restricted by many factors and therefore are used relatively less by domestic commercial banks in practice. Currently, existing researches in China focus mainly on direct measurement of the impact of macroeconomic downturn on non-performing loans using a relatively simple model, without fully considering the requirements of Basel agreement. Based on above researches, this paper emphatically investigated the determination of macroeconomic factors in stress testing mainly based on the contents of Basel agreement and the stress testing practices of domestic commercial banks. The MEF model was designed to determine the impact of macroeconomic factors on credit risk, and significant effect was realized through empirical test. It has great guiding significance in practice for commercial banks to carry out stress testing, which is rare in existing literature.