The Warning Models Of Debt Risk For Chinese Local Government

Local government debt is a major threat to the stability and sustainability of the local government finance. Becasuse the National Budget Act gives no permission to local government financial deficit and issuance of bonds, the debt is carried out in the dark. It is urgent but difficult, therefore, to build a set of warning system for the local government debt under current condition. On the basis of investigation, this article establishes two local government debt risk warning models, which have different adaptabilities and complementarities from a static perspective. Based on this model, this article further establishs an individual local government and multiple local governments debt risk warning model of dynamic linkage from a dynamic perspective.This paper consists of seven chapters. The first chapter analyses the topic basis, research significance, research method and, technique and the major contents of this research. The second chapter summarizes the risk warning researches on government debt or finance at home and abroad. The third chapter introduces the definition of "local government debt risk" analyses its producing mechanism and establishes a set of "risk warning indicators system" The forth chapater builds the risk warning model using Fisher discriminant analytical method. Under the large sample size and inconsistency of data circumstances, the fifth chapter establishes the debt risk warning model using multi-layer perceptron neural networks. Based on multi-agent modeling method, the sixth chapter builds the individual local government debt risk warning model and multiple local governments debt risk warning model of dynamic linkage using simphony Repast. The seventh chapter is a conclusion and prospect.The main content of the dissertation is as follows:(1) Establishing the debt risk warning indicator system for local government. Based on current high-frequency indicator of typical literature and other texts, combining with rational analysis, this study develops a systematic, comprehensive and observable local government debt risk warning indicator system. On the basis of "balance" concept, debt "dichotomy" and financial risk matrix, according to systematic and observability principle, this study establishes the debt risk warning indicator system by means of the qualitative analysis method, which including deleting, modifing, reserving and increasing some indicators. The essential characteristics of Chinese local government debt risk source are reflected by active liability indicator, such as the financing platform of fiscal revenue ratio. Through direct debt and contingent debt classification using the financial risk matrix theory, all kinds of local government contingent debts were reflected by the warning indicator system.(2) Building the risk warning model for small sample size based on the static perspective. Through the analysis of the distribution of debt risk indicator data and the characteristics of discriminant analysis method, this paper develops the local government debt risk warning model by means of Fisher discriminant analytical method. In order to guarantee the availability and applicability of this method, this paper conducts factor identification, reduces the number of parameters using factor analytical method, then the problem of small sample sizes is solved. Meanwhile, this study gives the discriminant function through case studing, on the basis of seven cities in Shandong Province for several years (6 years or 8 years) data. This discriminant function can classify those samples accurately for 91.3%.(3) Establishing the debt risk warning model under the large sample size and inconsistency of data circumstances, based on the static perspective. Through the analysis of the characteristics of different artificial neural network method and debt risk warning itself, this paper develops the local government debt risk warning model by means of multi-layer perceptron neural networks. In order to guarantee the availability and applicability of this method, this paper conducts factor identification, reduces the number of parameters using the fault tolerance well of neural network, then the problem of large sample sizes and inconsistency of data circumstances is solved. Meanwhile, this study gives multi-layer perceptron neural network through case studing, on the basis of seven cities in Shandong Province for several years data. This neural network can accurately simulate the expert judgments because its target error is 10-5.(4) Buliding the individual local government debt risk warning model and multiple local governments debt risk warning model of dynamic linkage based on the dynamic perspective. Through the analysis of the characteristics of complex adaptive, this paper develops the local government debt risk warning evolving model by means of multi-agent modeling theory method, based on the dynamic perspective. This results ensure that the multi-agent modeling theory can be applied in local government debt risk warning problem. According to the number of governments, this paper develops the individual local government debt risk warning model and multiple local governments debt risk warning model of dynamic linkage respectively. And then the study simulates the model through mainstream multi-oubject Repast Simphony tool, which ensures the reliability of simulating. Moreover, this model can make up some deficiencies of existing research based on the static perspective.