Early Warning Study On Sovereign Debt Default Risk

The causes and policy recommendations of Sovereign Default have been extensively studied academically,however,less researches have been performed on the mechanism of Sovereign Default from microcosmic and macroscopic perspective.Policy makers and scholars have mainly been assessing debt risk,sustainability of fiscal policies and steadiness of macro-economy based on their experiences,meanwhile,few studies were leveraged on mathematical analysis and empirical tests.Therefore,under current research mathematics,there were great chance of neglecting the major factors causing Sovereign Default,as a result,ineffective monitoring frequently occurred in current risk alert system,in another word,private sectors,rating agencies,scholars and policy makers had tended to under estimated Sovereign Default risk.It had always been unexpected when Sovereign Default happened.During European debt crisis,Stand&Poor’s,Fitch,Moody’s came under great criticism as they failed to make any effective early warning.As of December 31,2016,China’s foreign currency reserve totaled to 3 trillion US dollar.The significance of the appreciation of such foreign currency reserve had been widely recognized.In many countries,foreign currency reserve was usually invested in its Sovereign wealth fund.As such,building up a Sovereign Default alert system could contribute to assess risk of Sovereign Default on other countries,reduce investment losses and enhance foreign capital injection and self-development of China.Making reference to current researches and literature,we hereby developed a forecast model on Sovereign Default.Such forecast is based on three core factors,namely,sufficient sample database,robust early warning model and efficient early warning method.The third,fourth and fifth chapters of this paper construct such three core elements separately.The calculation of default risk of sovereign debt in specific countries is the application of the above three core elements,which is the main content of Chapter 6 of this paper.In this article,we built up a large,up-to-day database for Sovereign Default events.Utilizing R programming software to analyze the samples selected(all grogram codes were attached for reference),we identified factors causing debt default,compared the effectiveness of various default forecast approaches based data mining and made a default forecast on China,US and Japan.Literature review and introduction of default forecast approaches had been included in Section Two of this article.In this section,we analyzed current status of Sovereign Debt over the world,reviewed and compared the historical Sovereign Default events,reviewed current Sovereign Default theories,introduced those four Sovereign Default forecast approaches that had been applied in current researches and compared the advantages and disadvantages of each approach.The main purpose of Chapter Three is to establish the default warning system and default sample database.The theoretical basis of the establishment of debt default database is the theoretical analysis of the causes of sovereign debt default.This paper analyzes the theoretical causes of sovereign debt default from the perspective of Marxist economic theory and western economic theory.According to the classification principle of debt default factors,the text divides the debt default factors into five categories,namely,domestic economic factors,debt factors,international environmental factors,population factors and reputation factors.We identify the main early warning variables from the five factors,and analyze the influence mechanism of each variable on sovereign debt theoretically.And finally identified the sovereign debt default warning index system of 25 variables.In the case of default events,this article draws on the approach of Canadian bank rating group.On this basis,we have established a 25-factor sample database of defaults,which covers data from 77 countries between 1981 and 2015.Chapter four establishes the debt default warning model from the perspective of empirical analysis and determines the main factors of sovereign debt default based on logit regression method.Chapter Four constructed four main sets of economic variables i.e.,domestic economic variables set,macroeconomic policy variables set,international conditions variables set,and fiscal economic variables set,aggregately with 23 individual explanatory variables to explain the occurrence of sovereign debt default alert.Out of the 23 individual explanatory variables,logit regression of 77 countries from 1981 to 2015 shows that 11 are significantly meaningful,including 5 individual explanatory variables from domestic economic variables set(real GDP growth,unemployment rate,current account balance as of GDP,and export growth and external debt as of GNI);4 individual explanatory variables from macroeconomic policy variables set(inflation rate,M2 as of FX reserve,real exchange rate fluctuation and central government debt as of GDP);1 individual explanatory variable from demographical set,i.e.,dependency rate.In addition,goodwill variable also reflects the influence of concurrent default on future debt default risk.Meanwhile,this research points out that domestic variables are the main drivers for debt default,with minimal influence of international variables,which are far less resilient and significant.After the establishment of the early warning model,this paper uses the logit regression analysis method to identify the main factors of sovereign debt default based on the sample data from 1981 to 2015 in 77 countries.Based on the regression analysis,samples are categorized into subsamples to exam influence under different scenarios according to three dimensions: 1)horizontally into developed countries and developing countries according to different economic development stages;2)horizontally into European countries,Asian countries,Latin American countries,and African countries according to samples’ geographic distributions;3)vertically into pre-1999 period and post-1999 period according to the location of observation in the spectrum of time serials.The result of the categorization divides samples into 20 subsamples,to which individual regression analysis was performed upon.The individual regression analysis toward each subsample collectively enable the verification of explanatory variables in different time frame,different economic development stages,and different geographic locations.This chapter also compares the early warning ability of the early warning model established in this paper with the existing research.Compared with the existing research,the early warning model established in this paper has higher probability of prediction and lower probability probability of occurrence of type 1 errors.Chapter Five exams the most effective method to predict future sovereign debt default alert.The analysis introduces 11 alert methods within 6 categories: BP neural network analysis,decision tree analysis,and collective method within decision tree analysis(stochastic forest,Bagging,and Boosting methods),vector machine method,K approximation method,discriminant analysis(liner discriminant analysis and quadratic discriminant analysis method),and limited variables regression(logit regression and probit regression).After introduction of different methods,this chapter sets pivotal variables in the alert analysis,which was performed and compared across different models and datasets.Subsequently,this chapter evaluates the foresight capability of abovementioned models with matured artificial intelligence methods,e.g.10-knots-cross analysis and expected error cost method.The evaluation than scaled different models using 10-knots-cross analysis and expected error cost method.We believe that the spectrum of foresight capability in terms of predicting sovereign debt default ranges as: vector machine method,stochastic forest and Logit method.Serving as both starting point and destination of this thesis,Chapter Six discusses the application of the theoretical sovereign debt default alert methods.This Chapter verifies that using the big data from previous articles,the most effective method in this thesis can predict the occurrence of sovereign debt default of China,the US,and Japan.Based on the abovementioned research,this thesis can predict the occurrence of sovereign debt default of any given country in future one year with significant confident interval(viz.more than 90%).This thesis was honoured to take the liberty to contribute to future research in sovereign debt default alert studies.Chapter 7 is the conclusion and policy recommendations of this paper.In the seventh chapter,the main conclusions of this paper are summarized,and the policy suggestions are put forward.