Research On China’s Macroeconomic Forecast Based On Interaction MIDAS Model Based On State Transition

Gross domestic product(GDP),as a proxy variable of the macro economy,is also an important indicator to measure the overall operation of the macro economy,reflecting the rise and fall of a country’s economy.Therefore,in order to achieve moderate and stable economic development,the proxy variable of GDP should be monitored and warned,so as to grasp the trend of economic changes,so as to stabilize the economy as much as possible and achieve policy goals.On the one hand,there are many indicators that affect macroeconomic changes,such as money supply,industrial added value,etc.,and their sampling frequencies are different.On the other hand,the indicators that affect macroeconomic changes are not independent of each other,but show interwoven correlations and interactions,and this interaction has state transitions.Therefore,from the perspective of economic prosperity monitoring,this paper selects the consistent indicator group and the leading indicator group,and constructs the M-MIDAS-AR prediction model and the M-MS-MIDAS-AR prediction of state transition under the interaction of the consistent and leading indicators.Model,forecasting research on my country’s macro economy.The main conclusions are as follows:1.Use the cross-correlation coefficient to describe the interaction between the consistent index and the leading index,take Beta,Almon and Exp-Almon as the weight function,and the univariate MIDAS-AR model with the maximum lag order of 23 as the leading model,determine the optimal weight function and the maximum lag order,and construct the M-MIDAS-AR regression model and prediction model based on consistent indicators,leading indicators and their interactions.2.Multivariate MIDAS-AR forecasting model based on interaction,interaction plays an important role,successfully correcting the forecast results of current GDP by consistent indicators,leading indicators and lagging GDP indicators,reducing forecast errors.3.Based on the ICSS algorithm,the test results show that the interaction between the consensus index and the leading index has obvious state transition characteristics.4.Based on the interaction state transition characteristics,a two-state and three-state interaction M-MS-MIDAS-AR regression model was constructed.Based on the goodness-of-fit test,it was found that there was a three-state interaction M-MS-MIDAS-AR The regression model successfully captures the transmission mechanism of the interaction state transfer mechanism to the macro economy.5.Based on the interaction M(3)-MS(3)-MIDAS-AR(1)regression model,it is found that GDP has strong inertial characteristics,and the GDP of one lag period contributes as much as 95% to the current GDP change.Consistent indicators,leading indicators,and interactions all have short-term effects on GDP,and the time limit is about 5 months,but the impact of the three on GDP is different.The interaction in state 1 has the greatest impact on GDP,and the range of change is(-1,1),followed by state 3,state 2 has the least impact.Consensus indicators are more powerful than leading indicators.6.Based on the interaction M-MS-MIDAS-AR-h forecast model,forecast the GDP data in the sample from the first quarter of 2019 to the third quarter of 2021.In the forward 1-12 step prediction,the prediction errors of the M(3)-MS(3)-MIDAS-AR-h model are smaller than those of the traditional multivariate mixing regression model,and the overall prediction error is reduced by more than 30%,especially in the In the 9-12 step prediction,the error is reduced by about 60%,indicating that the long-term prediction advantage of M(3)-MS(3)-MIDAS-AR-h is more significant.