Var Methods For The Dynamic Impawn Rate In Invetory Financing

Inventory financing, as one of main business models of supply chain finance in China, make inventory as the pledge to strongly mitigate credit risk of loans. In current banking practice, banks still rely on the experience to determine the impawn rate, which would be far from consistent with the risk tolerance level of banks. Therefore, as a core issue, it is important to set impawn rate not only for the risk control of supply chain finance but also for promotion of the development of the business. There are many theoretical modeling and cases based on individual samples while the empirical analysis based on large number of samples are scarce in existing literature.Different from the existing research statically setting impawn rate in the impawn period, this paper, taking account comprehensively of macroeconomic environment, the credit level of counterparty, the liquidity of pledged inventory and the risk preference of banks, first proposes a dynamic model setting dynamic impawn rate by dividing the impawn period into different risk windows to trade off the dilemma between risk horizon and impawn period. Compared with pledging loan of bonds and stocks, this paper proposes that, the essence of setting impawn rate in inventory financing is to resolve the long-term risk forecasting due to the insufficient liquidity of inventory.Based on the dataset of spot steel, usually traded in the OTC (over-the-counter markets), this paper establish the model of AR(1)-GARCH(1,1)-GED, which can better depict the features of the autocorrelation, heteroskedasticity, leptokurtosis and fat-tails of the returns, and forecast VaR of steel during the different risk windows in the impawn period by methods of out-of-sample, ultimately get the impawn rate consistent with the risk tolerance of banks under the consideration of the cost of loans.Furthermore, this paper presents an application of Copula-ARMA-GARCH model in the estimation of a inventory portfolio’s VaR with rolling time window, composed of1#Cu and Rebar(HRB400). This model could better depict the characteristics of the autocorrelation, heteroskedasticity, leptokurtosis and fat-tails and leverage effects of the marginal distribution, but also the conditional dependence of the joint distribution. The empirical results shows that compared with the histroial simulation method, the results of Copula-ARMA-GARCH model with MonteCarlo simulation capture the VaR more successfully in out sample. In addition, the impawn rate of this model is more effective than experience methods in short term risk windows.In summary, both the model of AR(1)-GARCH(1,1)-GED and Copula-ARMA-GARCH could control the risk better while reducing the efficiency loss compared with existing methods. It puts forward a dynamic impawn rate mode and framework for banks.