Assessing the Probability of Bankruptcy with the Merton

Assessing the Probability of Bankruptcy with the Merton Approach Relative to Accounting Variables –Evidence from Taiwan and China Mei-Ying Liu Soochow University Dachen Chang Soochow University Hsien-Feng Lee National Taiwan University 1

Purpose Using data for both Taiwan and China, this study compares the default prediction power performance of the market-based model relative to the accouting-based model. 2

Literature 1/2 In theory, the market-based model is better to immediately reflect the firm’s operating conditions based on the stock price, and thus Merton’s marketbased model should be superior to the accountingbased model. 3

Literature 2/2 1. Empirical results support the market-based model(Hillegeist et al. (2004), Vassalou and Xing (2004), Benos and Papanastasopoulos (2005) 2. In Taiwan, empirical results do not support (Chen, Wang and Hsu (2004), Lin (2005), Su and Lin (2006), Lo and Lin (2006) 4

Argument However, the effectiveness of the accounting-based model to evaluate the firm’s credit risk depends on the quality of the information contained in the financial statements, whereas the market-based model is determined by the efficiency of the financial markets. It should be an empirical question. 5

Sample Selection and Data Sources 1/2 Data Sources: Taiwan Economic Journal (TEJ) database Period: 1998~2007 Sample: all listed companies in Taiwan and China exchanges (Shanghai and Shenzhen) 6

Sample Selection and Data Sources 2/2 Taiwan: 4, 951 annual observations with 90 firms defaulting China: 10, 525 annual observations with 319 of firms defaulting. 7

Market-Based Credit Risk Assessment Model Black-Scholes option pricing formula (Merton 1974): probability of default: 8

Accounting-Based Credit Risk Assessment Model Logit model(defaulting firm Y= 1, non-defaulting firm Y = 0. ) default probability: Include 31 financial variables belong to five financial ratio indicators- profitability, the growth trend, debit management, asset management and liquidity. 9

Hybrid Credit Risk Assessment Model integrate the information provided by the accountingbased model and the market-based model To the accounting-based model we add the PD variable included in Merton’s model, and use the Logit model to build the hybrid credit risk model. 10

Test of Discriminatory Power of the Models Kolmogorov–Smirnov test( K–S test) where F(x), S(x) is the cumulative probabilities at each different stage of the defaulting and nondefaulting firms respectively. The K-S value is the largest cumulative probability deviation. K-S value lies in between [0, 1], and the larger the value, the better is its ability to discriminate. (Mays (1998)) 11

Hypothesis 1/2 Hypothesis 1: The performance of the market-based model is better than the accounting-based model. Hypothesis 2: The performance of the hybrid model is better than both two models in terms of evaluating credit risk. 12

Hypothesis 2/2 Hypothesis 3: For the accounting-based model, the performance of Taiwan is superior to that of China. Hypothesis 4: For the market-based model, the performance of Taiwan is superior to that of China. 13

Table 1 Results of the K-S Test Empirical Results Model K-S value Taiwan China Accounting-based Model 0. 6558 0. 6281 Market-based model 0. 3578 0. 2084 Hybrid model 0. 6630 0. 6244 14

Empirical Results 1/2 For both Taiwan and China, the discriminatory power of the accounting-based model is markedly superior to that of the market-based model, Hypothesis 1 is not established. Taiwan’s hybrid model slightly outperforms the other two models. The finding supports Hypothesis 2. However, the hybrid model is inferior to the accounting-based model for China. The finding does not support Hypothesis 2. 15

Empirical Results 2/2 The discriminatory power of the accounting-based model for Taiwan being slightly greater than that of China. Thus, Hypothesis 3 is established. The discriminatory power of the market-based model in Taiwan is higher than that in China, and so Hypothesis 4 is also established. 16

Conclusion The empirical results are different from those for Europe and the U. S. A in that the market-based model for both Taiwan and China fails to outperform the accounting-based model. Each of the credit risk models for Taiwan is, however, superior to those for China, indicating that the performance of the market-based model when compared with that of the accounting-based model should be related to the development of the securities market. 17

Conclusion In developing economies stock markets are more volatile during the rapid economic growth process, therefore a high volatility of stock return may overestimate the default probability. Thus, the market-based credit risk assessment model may be less suitable in such instances. 18

Thank you for your attention 19