The Replication Of PMI Index:Variables,Path Analysis,and Prediction Of Index

Although the global economy gradually picks up after the international financial crisis,the internal effect of crisis still exists.With the influence of non-economic factors,such as politics,the development of our country is facing with many uncertainties and challenges.In order to scientifically?timely monitor the development of economic,China set up the Purchasing Managers' Index——PMI index system.Using the un-quantified survey data,China announces last month's PMI index in the early next month.It can synthetically reflect the development of macro-economy.Under this background,this paper select 36 objective quantified data as possible affecting variables based on previous research,such as PMI(-1)(lagged PMI index)?the ratio of excessive reserve?the required reserve ratio?the imports amount?the exports amount?the average exchange rate?M0 that circulates outside the banking system.At the same time,based on non-parametric variables elimination method of mixed data proposing by Li and,Racine(2004),the paper copied variables that there is correlation between PMI index and variables.What is more,this paper structure a semi-parametric?time-varying complete model and path model to analyse the combined effect and the individual effect of selected variables.Finally,the model is extended to a prediction model about PMI index.So it pushes the descriptive statistic of PMI index to inferential statistic.It fills in the blanks of existing literature in this area.The specific work and conclusions are as follows:First of all,based on non-parametric variables elimination method of mixed data,the paper selects variables that there is correlation and linear relationship.The paper copies variables that there is correlation between PMI index and variables.The result shows the linear variables are:PMI(-1)?the exports amount(CE)?industrial added value(GZ);the non-linear variables are:the turnover of stock(GE)?the public revenue(GR)?the public balance of payments(GGCE)?tax revenue(SR)?the average exchange rate(PL)?the interest rate of demand deposit(HL).Secondly,through the empirical analysis of semi-parametric?time-varying complete model and path model,we find combined effects can better explain the PMI index.At the same time,we find different non-linear variables have different effect on PMI index.The variables of positive effect are:GE?GR?GGCE.There is no negative variable.The SR,PL and HL have little effect on PMI index.Finally,through the comparison of coefficient of variation,we find the rank of impact on the PMI index is:GGCE,GE,HL,PL,GR,SR(after excluding).Finally,through the empirical analysis of semi-parametric?time-varying prediction model,we find selected variables can not only explain PMI index in the future,but also provides a more precise prediction model.It can be a strong evidence for enterprises,financial institutions and the government to judge the economic situation and make the plan of development.