Assess the Impact of Asset Price Shocks on the Banking System

Abstract

In order to analyze the impact of asset price shocks on the banking system, this paper develops a macro stress-testing framework to assess liquidity risk, credit risk and market risk. Firstly, using the Monte Carlo method to simulate market risk path generated by the financial asset price shocks; secondly, using Morton model to analyze the linkage between market and default risks of banks, while the linkage between default risk and deposit outflows is estimated econometrically; Contagion risk is also incorporated through banks’ linkage in the interbank and capital markets. Finally, the framework is applied to a group of banks in China, based on publicly available data as at the end of 2009. Its test results show that: the liquidity risk of the bank system is very low, the probability of no bank default is 99.32%, and the entire bank system is stable.

Appendix

1.1 Econometric Estimation of the Relationship Between the Probability of Default and the Monthly Retail Deposit Outflow Rate

To reveal the empirical relationship between PD and the monthly retail deposit outflow rate, the following panel data regression equation is estimated:

$$ {{\hbox{G}}_{i,t}} = {\partial_i} + {\beta_1}\ln ({R_{i,t}}) + {\beta_2}\ln ({R_{ - i,t}}) + {\beta_3}\ln (P{D_{i,t}}) + {\beta_4}{Y_t} + {\varepsilon_{i,t}} $$

(12)

Where G_i,t, is the monthly growth rate of China dollar retail deposits of bank i at time t. R_i,t is the retail deposit rate offered by bank i at t, while R_−i,t is that offered by other banks in the market. The estimated coefficients of R_i,t and R_-i,t, (i.e., β₁ and β₂respectively) are expected to be positive and negative respectively. PD_i,t is the default probability of bank i at t, which is calculated based on the Briys and de Varenne model. The empirical relationship between PD and the monthly retail deposit outflow rate is revealed by the estimated value of β₃, which is expected to be negative. Y_t is the year-on-year growth rate of GDP in China, and the estimated coefficient of Y_t is expected to be positive, as the growth rate of retail deposits should be higher under good economic conditions.

We estimate (12) using its first difference form with the generalized least squares method. β₃ is estimated to be −0.2111, which is statistically significant at the 5% level. This suggests that a bank with high default risk (i.e., PD_i,t closer to 1) would lead to a monthly retail deposit outflow rate of about 21.44%.

The 95% confidence interval of β₃ is approximately between −0.42 and −0.01. In the stress-testing framework, instead of setting the monthly retail deposit outflow rate to be the point estimate (i.e., −0.2111), a more severe rate, which is the lower bound of the confidence interval, is assumed (i.e., monthly retail deposit outflow rate = −0.42 PD_i,t).

Other parameters, β₁, β₂ and β₄ are estimated to be 0.4738, −0.2748, and 1.2387 respectively, with β₁ and β₂ being statistically significant at the 1% level and β₄ being statistically significant at the 10% level. Overall, the estimation result is consistent with the economic intuitions in (12).