An Improved Spatial Stochastic Frontier Model

The research on the source of China's economic growth is a hot topic in the economics field in recent years.Since the reform and opening up,China's economic development has achieved unprecedented achievements,but there is a huge controversy about the driving force of China's economic growth.Factor input is the core of contemporary regional development,while the contribution of total factor productivity to China's development is controversial.Spatial correlation analysis is a powerful tool for regional economic research.According to "The First Law of Geography" : Everything is related to everything else,but near things are more related to each other.In this context,it is crucial to select a right model for the study.This paper firstly expands the application of the Geographic Weighted Model(GWR)based on cross-section data to panel data,uses golden section method to determine the smooth parameters,and adding the composite disturbance item in the stochastic frontier model(SFA)into the improved GWR model.Secondly,inspired by spatial error model(SEA),adding spatial correlation into error vectors.In the likelihood function of the model,the Monte Carlo method is used to calculate the multidimensional normal integrals,and the genetic algorithm is used to maximize the likelihood function.Finally,for accurate calculation,the k-value assignment method of the k-nearest neighbor weight matrix is modified.Provincial panel data for 2000-2014(excluding Tibet,Hong Kong,Macao,and Taiwan)was chosen for study.While obtaining the total factor productivity(TFP)of each province,we can also get their respective frontier production functions with the new approach.After that,the Dagum Gini coefficient was used to study the unbalanced development of China's provincial TFP development level,and the Spatial Markov Chain method was used to explore the Temporal and Spatial Patterns of Provincial TFP Development.At the end,on the basis of analyzing the empirical results,this paper proposed policy recommendations.The main conclusions of this paper are:Firstly,factor input,especially capital input,remains the main source of economic growth in most regions.Secondly,the central and western regions are facing a serious problem of labor loss.Thirdly,China's economic growth is accompanied with high pollution and high energy consumption.Fourthly,poor use of new technologies remains a huge obstacle of China's development at this stage.Fifthly,Total Factor Productivity in the middle and west regions is often underestimated in existing research.Sixthly,China's TFP gap has not changed significantly in study period,but the regional development of TFP tends to convergence while the interregional gap tends to expand.Seventhly,there is ‘club convergence' and spatial polarization phenomenon in provincial TFP development during study phase.High level and low level convergence clubs are most stable,which means a strong technical barrier.Eighthly,the convergence process of changes in regional TFP levels is not independent in space because of impact comes from neighborhoods.Ninthly,areas whose TFP level moved up are concentrated in the central and eastern coastal areas,specially,TFP levels in central and west China had been underestimated in many past papers.Provinces downwardly transferred are almost remote or backward areas.