A Study on Commercial Bank Sustainable Development Countermeasures in the Face of Real Estate Credit Risk in China

Transcription

1 A Study on Commercial Bank Sustainable Development Countermeasures in the Face of Real Estate Credit Risk in China Ping Guo * School of Economics and Management Shijiazhuang Tiedao University Shijiazhuang050043, China Abstract According to the People's Bank of China's statistics, in the development process of real estate industry, the ratio of bank loans used directly or indirectly into total loans of the real estate industry is more than 55%. In 2013, non-performing loans of real estate credit accounted for 48% of commercial banks' bad loans. From this, the credit risk of commercial banks is mainly from real estate credit, which seriously threats stability and sustainability of commercial banks. This paper selects financial data and non-financial data of 146 real estate listed companies in for samples. It takes "whether net cash flow from operating activities is greater than zero" as the grouping criteria. It goes to establish the real estate credit risk evaluation Logistic model applicable to commercial banks in China, using SPSS21.0 software, then proceeds to validate real estate credit risk assessment model. The output results show that the model has good discriminatory capacity to help commercial banks to: i) make the right judgments based on the probability of default of the real estate business, and ii) to assess and make decisions. Keywords - credit risk; real estate; logistic model; prediction I. INTRODUCTION Real estate credit, as the most important credit business of commercial bank in China, continues to increase with the rapid development of real estate industry. In recent years, affected by the land prices, labor costs, material costs climbing steadily[1-2], the amount of capital required for the real estate industry has been increasing, and commercial banks have been expanding real estate credit[3]. However recently under the influence of national macroeconomic policies[4], house prices of Beijing, Shijiazhuang and other first-tier and second-tier cities have signs of loosening, rational regression of real estate prices severely affected the real estate capital chain, reducing the debt paying ability of real estate companies, resulting in non-performing loan ratio of commercial banks increased, thus making the real estate credit risk of commercial banks tend to rise[5]. Therefore, China's commercial banks need to actively seek countermeasures to strengthen the real estate credit risk management[6], reduce risk and indirectly guide the healthy development of the real estate industry[7-8]. The paper takes financial data and non-financial data of 146 listed real estate companies in samples, and takes "net cash flow from operating activities is greater than zero" as the grouping criteria, which is applicable to commercial banks in China, to build Logistic Model of real estate credit risk evaluation. The results show that the model has better discriminating ability for commercial banks to judge the probability of default of the real estate business, and can help the commercial banks make assessments and decisions[9-10]. II. DEFINITION OF THE RESEARCH OBJECT With the increasing amount of commercial banks real estate credit, the quality of the real estate companies operating conditions will affect the liquidity of bank funds to a certain extent, and the loss caused by the breach of real estate credit companies will eventually happen to be burdened by commercial banks[11]. Therefore, how to assess and to prevent such losses is particularly important for the commercial banks[12-13]. The probability of the real estate companies loans default eventually defined as the study object of the paper[14], and it is represented by P. Take into account the smooth acquisition of sample data and sample data acquired being true, the paper selects the listed real estate companies as study samples. The paper chooses the standard "net cash flow from operating activities is greater than zero" as the division of real estate listed companies, and mainly two reasons: first, considering from the real estate companies, when its net cash flow from operating activities is less than zero, indicating that there is a problem in the enterprises operating, is likely to lead companies into financial crisis, the ability to repay a variety of credit decline, and the likelihood of credit default increases. Second, from the point of view of the bank, according to the five-category classification standard of bank loans, when the borrower's net cash flow from operating activities is negative, the loans will be classified as nonperforming loans, which also explains the occurrence possibility of credit default of the real estate companies is relatively large this time. III. SAMPLES SELECTION Samples in this study were derived from GTA database(csmar), Sharp Thinking database and website of DOI /IJSSST.a ISSN: x online, print

2 Eastmoney, were selected the primary data of 153 listed real estate companies in 2012 and After removed the sample serious lack of data, and ultimately 292 data groups of 146 real estate listed the company are retained as samples in this study, in which 146 data groups in 2012 are set as model samples, and other 146 data groups in 2013 are set as the test samples. When Logistic regression model is constructed, it still requires two sample groups, in accordance with the principle of "net cash flow from operating activities is greater than zero" the model samples need be into two groups, one group is the normal companies samples (83 normal samples) that the net cash flow from operating activities is greater than zero, and the other group is the default companies samples (63 default samples) that the net cash flow from operating activities is less than zero. The test samples in 2013 are also at the same standards divided into two groups, and to validate the Logistic regression model. IV. ORIGINAL SELECTION OF INDICATORS The selection of indicators plays a decisive role for effectiveness of the model, so to choose indicators that can fully reveal and disclose operating and financial conditions of listed companies[15-16]. The paper ultimately from the financial and non-financial factors selects 18 financial indicators and the 3 non-financial indicators as model index, selection and interpretation of specific indicators are shown in Table I. V. INDICATORS MODE INDICATORS SCREENING A. Normality Test The paper firstly uses SPSS21.0 software, to do normality test for financial 18 variables of selected 146 real estate companies, to see whether the data follow a normal distribution. The method is based on K-S test named by two former Soviet mathematicians Kolmogorov and Smirnov, the method has a higher accuracy ratio than chi-square test. Significance levels of two-sided test of 5 variables, such as return on equity ratio, return on total asset ratio, assetliability ratio, total asset turnover ratio and cash recovery ratio of total asset are greater than 0.05, follow normal distribution. Significant levels of two-sided test of the remaining variables are less than 0.05, which do not follow a normal distribution. Financial index Non-financial index TABLE I ORIGINAL SELECTION OF INDICATORS Classification Indicator Formula Return on equity ratio (X1) Net profit/average net asset Profitability index Solvency index Operating capability index Development capacity index Cash flow index General conditions of companies Return on total asset ratio(x2) Net profit margin on sales ratio(x3) Ratio of profits to cost(x4) Current ratio(x5) Quick ratio(x6) Asset-liability ratio(x7) Equity ratio(x8) Inventory turnover ratio(x9) Receivable turnover ratio(x10) Fixed asset turnover ratio(x11) Total asset turnover ratio(x12) Net asset growth rate(x13) Operating profit growth rate(x14) Capital maintenance and increment ratio( X15) Ratio of cash flow to current liability(x16) Sale cash ratio(x17) Cash recovery ratio of total asset(x18) Enterprise scale(scal) City located(cit) Senior manager s education(edu) (Total profit + Financial expense)/total asset Net profit/sale revenue Total profit/(operating cost + Selling expense + Management expense + Financial expense) Current asset/current liability (Current assets inventory)/current liability Total liability/total asset Total liability/ Equity Operating cost/average inventory balance Operating revenue/average accounts receivable balance Operating revenue/average net fixed asset Operating revenue/average total asset (Ending net asset beginning net asset)/beginning net asset The operating profit this year/the operating profit last year The total shareholder equity at the end of the year/the total shareholder equity at the beginning of the year Net cash flow from operating activity/current liability Cash received from sales of goods or providing services Net cash flow from operating activity/ending total asset Ln(Total asset) First-tier developed city=1, other city=0 Master degree or above 50%=1 Under Maser degree<50%=0 B. Significant Test Because the ability of variables to distinguish between normal group and default group is different, so the paper uses SPSS21.0 software, to do T test for the 5 variables following normal distribution, such as return on equity ratio, return on total asset ratio, asset-liability ratio, total asset turnover ratio and cash recovery ratio of total asset; to do Kruskal Wallis H test for the other 13 variables following non-normal distribution, then to verify the selected 18 financial variables can effectively distinguish the normal group and default groups. According to statistical theory, when a significant level Asymp.Sig. (2-tailed) <0.05, the indicator is considered DOI /IJSSST.a ISSN: x online, print

3 significant. From the results of parametric dual independent sample T test and nonparametric Kruskal Wallis H test, it shows that in the profitability indicators return on equity ratio and ratio of profits to cost, in solvency indicators current ratio and asset-liability ratio, in operating capability indicators inventory turnover ratio and fixed asset turnover ratio, total asset turnover ratio, in development capacity indicators of operating profit growth rate, in cash flow indicators cash ratio of cash flow to current liability, Sale cash ratio, cash recovery ratio of total asset, these 11 financial indicators variables can significantly differentiate between normal group and default group. C. Correlation Test Because in the discrimination during Logistic regression, too large multi-collinearity will have a greater impact on estimation of Logistic regression coefficients, standard deviation of Logistic regression coefficients estimated occurs bias, so before the use of Logistic regression model a correlation test need to do for indicators that through the above two-step screening, namely the 11 financial indicators and enterprise scale in the 3 non-financial indicators to exclude the existence of the problem of multi-collinearity among the independent variables. This paper uses SPSS21.0 software using Pearson test method to test correlation between the independent variables. There is a big correlation between return on equity ratio (X1) and ratio of profits to cost (X4), and the correlation coefficient is as high as 0.890, far beyond the theory higher criteria defined 0.7. When making analysis of overall ability of the firm, the return on equity ratio is at the heart of Du- Pont analysis, so return on equity ratio(x1) is more credible than ratio of profits to cost (X4) to reflect the profitability of enterprises, then excluding ratio of profits to cost (X4). Therefore, after screening the final indicators that is able to enter the model are return on equity ratio (X1), current ratio (X5), asset-liability ratio (X7), inventory turnover ratio (X9 ), fixed asset turnover ratio (X11), total asset turnover ratio (X12), operating profit growth rate (X14), ratio of cash flow to current liability (X16), sale cash ratio (X17), cash recovery ratio of total asset (X18) 10 financial indicators and enterprise scale (SCAL), city located (CIT) and senior manager s education (EDU) 3 non-financial indicators. VI. CONSTRUCTION OF LOGISTIC MODEL The paper assumes a binary dependent variable indicates whether the enterprise is defaulted, P denotes the defaulted probability of real estate companies that = 1, then the 1-P says the probability of real estate business not defaulted, namely = 0. The screened out variables of 10 financial indicators and three non-financial indicators are as independent variables. Input the dependent variable and 13 independent variable data into SPSS21.0 software, implementing Binomial Logistic Regression command, using the forward Wald stepwise selection method to determine the final model, in the calculation process, each step retains the Wald statistic variable that meets the set significant level, until no variable meets set significant level. The main results are shown in Table II and Table III and Table IV: Step 1 Step 2 Step 3 Step 4 TABLE II OMNIBUS TESTS OF MODEL COEFFICIENTS 2 df Sig. Step Block Model Step Block Model Step Block Model Step Block Model Table II is according to the three test methods of relative likelihood ratio, namely Step, Block, Model, and shows the chi-square statistic, freedom degrees and significance level in step 1 to step 4. In the step 4 the likelihood ratio chi-square statistic is , and freedom degree is 4, corresponding significance level is 0.000, far less than the set significance level of 0.05, therefore, it is reasonable to believe that the overall Logistic model is significant. TABLE III MODEL SUMMAR Step -2 Log likelihood Cox & Snell R2 Nagelkerke R n Table III, -2 Log likelihood, Cox & Snell R2 and Nagelkerke R2 are the statistic for evaluating goodness of fit of Logistic models, the table reflects the value of each statistic from step 1 to step 4. As can be seen from the table in step 4, -2 Log likelihood value is , and the smaller the value, the better the goodness of fit of the model, the results obtained in the table are rather ideal. In addition, the value of Cox & Snell R2 and Nagelkerke R2 are respectively and 0.578, indicating that the effect of the model fit is better. The final model by forward Wald stepwise selection method through a four-step calculation to be determined, the final model retains the 4 variables, including 3 financial indicators, asset-liability ratio (X7), ratio of cash flow to current liability (X16) and cash recovery ratio of total asset (X18), and 1 non-financial indicator the city located (CIT). The regression coefficients of each indicator and significant test results are shown in Table IV. As can be seen from step 4 of Table 8, the corresponding significance level of the coefficients of ratio of cash flow to current liability (X16), cash recovery ratio of total asset(x18), asset-liability ratio (X7) and city located(cit)are less than 0.05, and the significance level of DOI /IJSSST.a ISSN: x online, print

4 constant is also less than 0.05, indicating that the 5 coefficients have passed the significance test, the final Logistic model is as followed: 1 p ( X X X CIT ) 1 e By the above formula the probability whether a real estate enterprise is default can be predicted by commercial banks. TABLE IV LOGISTIC REGRESSION RESULT OF FORWARD WALD STEPWISE SELECTION METHODS Step Indicator B S.E. Wals df Sig. Exp (B) Step 1 Cash recovery ratio of total asset(x18) Constant Ratio of cash flow to current liability (X16) Step 2 Cash recovery ratio of total asset(x18) Constant Ratio of cash flow to current liability (X16) Step 3 Cash recovery ratio of total asset(x18) City located(cit) Constant Ratio of cash flow to current liability (X16) Cash recovery ratio of total asset(x18) Step 4 Asset-liability ratio (X7) City located(cit) Constant a.the cut value is VII. MODEL TEST The paper tests sample data of 146 real estate companies with Logistic model to see whether a violation has occurred, in which the amount of normal real estate companies and real estate companies defaulted is 83 and 63, the test results are shown in Table V. TABLE V CLASSIFICATION TABLE Observed Predicted Percentage Correct Step Overall Percentage 93.8 The above table shows that at the probability of 0.5 cut value, the sample data are tested with the final Logistic model. By comparing predicted results and the actual results, there are 79 companies of 83 real estate companies predicted to be normal, 4 companies are predicted error, and prediction accurate rate was 95.2%. There are 58 companies of 63 real estate companies predicted to be default, 5 companies are predicted error, and prediction accurate rate is 92.1%. The model correctly predicts the total 93.8%, indicating that the model has a very good prediction effect, and is able to provide a strong basis for commercial banks to predict real estate companies credit risks, and has important practical significance. It is noteworthy that the prediction result is base on the probability of cut value 0.5, and the critical value P Logistic model theoretically has no specific standards, i.e., the theoretical probability of cut points can be any value between [0,1]. With the different cut-off point, Logistic model prediction is not the same, and leading to the evaluation of the model is also not the same. Prior to the evaluation of the model, we should define the two types of errors, the first category is wrong to judge the normal real estate companies as default group, the second category is wrong to judge the real estate companies as defaulted group. The paper should be based on actual sample data, to determine the optimal cutoff point. The following step will select 0.3, 0.4, 0.5, 0.6, 0.7 as cut value to calculate the probability of error of the tow categories. By analyzing the optimal cut value to determine the best Logistic models to predict the real estate credit risk for commercial banks, and the results are shown in Table VI. As can be seen from Table VI, with cut value increasing, the probability of first category error decreases, the probability of second category error increases, generally the interval [0.4, 0.6] considered to be more suitable, combined with the results ob-tained in the Table, 0.4 will be selected as the default cut value for model to further examination. VIII. MODEL VERIFICATION In addition to sample data selected of real estate companies in 2012 to test model, the paper also selects sample data of 146 real estate listed companies in 2013 to verify the model, of which 69 companies in normal group, 77 companies in default group. Taking 0.4 as the cut value, putting data of 146 real estate listed companies in 2013 into the Logistic model established, the final prediction results as shown in Table VII. From the results in Table 7, with the Logistic model established to predict the groups in 2013, 61 companies of 69 real estate companies are predicted normal, and the predic- DOI /IJSSST.a ISSN: x online, print

5 tion accurate rate was 88.4%, 72 companies of 77 real estate companies are predicted default, and the prediction accurate rate was 93.5%. The total prediction accurate rate is 91.1%, again indicating a high predictive effect of the model. Cut Value Observed Test group Normal(83) Default(63) TABLE VI THE IMPACT ON PREDICTED RESULT WITH DIFFERENT CUT VALUE Percentage correction% TABLE VII CLASSIFICATION TABLE 0 1 Predicted Percentage Correct Overall 91.1 Percentage IX. CONCLUSIONS Taking "Net cash flow from operating activities is greater than zero" as the standard, divide the sample data and test data into normal group and default group in empirical research, the result shows that financial indicators, i.e. assetliability ratio (X7), ratio of cash flow to current liability (X16), and cash recovery ratio of total asset(x18), and nonfinancial indicator city located(cit)have a greater impact on the real estate credit risk of commercial banks, which agreed with the actual concerning direction of the real estate credit risk management of commercial banks, and therefore the model have a strong practical significance for commercial banks. According to the model results, commercial banks should be more concerned about the city where the real estate companies located, debt-paying ability and cash flow information. The higher the ratio of cash flow to current liability (X16) and cash recovery ratio of total asset(x18), at the same time the lower the asset-liability ratio (X7), the default probability of the real estate companies is lower, then the real estate credit risk of commercial banks is smaller. On the contrary, the risk is greater. ACKNOWLEDGMENT The paper is part of the research results of " Intelligent prediction method and empirical study of financial volatility ", which is one of the humanities and social sciences youth fund project of Ministry of Education (No.11JC790048); and is part of the research results of " Study on county function reconstruction and construction layout planning in Hebei province", which is one of the Hebei science and technology plan projects(no d); and received the fund Error probability of first category% Error probability of second category% Total error probability% of school of economics and management in Shijiazhuang Tiedao University. REFERENCES [1] Gao Qiang, Real estate credit risk and prevention of commercial bank, Money China (Academic),vol.1, pp. 80, [2] ang Wensheng, Zhao ang, Review on real estate credit risk system in commercial banks, China Price, vol. 5, pp , [3] Li Li, Fu Bingtao, Analysis of influence of the real estate market regulation to commercial banks,chinese Finance,vol. 24, pp , [4] Zhong Feng, Research on real estate credit risk of Chinese commercial banks - a case study of Guangzhou branch of A bank, Times Finance, vol.1, pp , [5] Wang Cheng-Hao,Lai Kin-Keung,Dong Ji-Chang, Reviewing prevention of the credit risk of real estates, Xitong Gongcheng Lilun yu Shijian/System Engineering Theory and Practice, vol.30, pp [6] Zhao Jinlu, Zhuang uan,zhao Lei, Risk of Commercial Bank Loans in Post-crisis Era, Banker, vol.10, pp , [7] Luo Wei, Study on credit risk stress test of real estate loans for commercial bank - based on empirical test of vector error correction model, Financial Theory and Practice, vol.10, pp , [8] Wen Shuhui, He uhua, real estate development loan risk and prevention of commercial banks,academic Exploration, vol.8, pp , [9] ang anzhong, Achievement of commercial banks sustainable and healthy development in post-crisis era, China Credit Card, vol.3, pp , [10] He Guo-sheng, LI Dong, On measuring and comparing the sustainable development of China' s listed commercial banks. Journal of Northeastern Universi ty (Social Science), vol.14, no. 1, pp , [11] Wang Cheng-Hao,Lai Kin-Keung,Dong Ji-Chang, Reviewing prevention of the credit risk of real estates, Xitong Gongcheng Lilun yu Shijian/System Engineering Theory and Practice, vol.30, pp , [12] Hao Jiangxia, Analysis of policy impacts of current real estate regulation on commercial banks, China Real Estate, vol.12, pp , [13] Cao ang 2014 Research on real estate financial risk state of commercial banks Modern Business, vol. 15, no. 21, pp , [14] Wang Huijuan, Analysis on credit risk management of real estates of bank, WIT Transactions on Information and Communication Technologies, vol.49, pp , DOI /IJSSST.a ISSN: x online, print

6 [15] Xie Chi,Wang Peng,ang Jiaojiao,Wang Gangjin, 2013 Measurement of the credit risk contagion effects between the sectors in the tertiary industry based on DCC-MSV-KMV model, Hunan Daxue Xuebao/Journal of Hunan University Natural Sciences, vol.40, no. 5, pp , [16] Wang Ping, Xu linhan, 2013 Relationships between real estate and financial services of commercial banks and development strategies Business Economy, vol.4, no. 3, pp. 4-5, DOI /IJSSST.a ISSN: x online, print