On The Definition, Testing And Mechanism Of The Club Convergence Of Regional Economic Growth

As an important type of convergence club convergence has extended the researches on convergence from the single steady state to many steady states so to expand the development fields of convergence theory. In reality club convergence embodied the characteristic of"within group convergence and between groups divergence"has better explain capability for the general phenomenon about co-existence of rich regions and poverty regions. Therefore since the middle 1990s club convergence has gradually been the hot spot of internal and external researches on convergence. But totally the researches on club convergence still are a field which is in the beginning period and urgently needs to be reinforced. Almost all related literature limited to test existence of club convergence but involved little about what club convergence was, what theoretical foundations for testing were, how to choose suited methods of regional classification and convergence testing, and how club convergence formed so on. The significances of this study lie in trying to solve the above mentioned problems using Chinese regions which foreign literature pay little attentions as research objectivities. In order to do this on the basis of the convergence theories in Neoclassical growth, the Neoclassical stochastic growth theories, the factor spillovers theories in New growth and the geography spillovers theories in New economic geography and with the technical supports including Mathematical Analysis, Classification and Regression Tree Analysis, Exploratory Spatial Data Analysis, Cross-section approach, Panel approach, Spatial autoregressive model, Spatial error model, Spatial cross-regressive model, Distribution model and GIS, our study systemically investigates the definition of club convergence, the theoretical foundations for distinguishing club convergence, the methods system for distinguishing club convergence, distinguishing Chinese club convergence and probing the forming mechanisms by using the methods from putting forwards hypothesizes to examine them.The thesis has gotten the following main conclusions.First, it needs to consider the differences between time dimension, space dimension and divergences among test methods when we definite club convergence. The thesis firstly puts forwards the Time and Space Coupling Club Convergence which combines time dimension and space dimension against their separation lying in current literature so to make up lacks created by their separation. In addition considering the divergences resulted from different test methods the thesis further divides club convergence into Cross-section Time Club Convergence, Stochastic Time Club Convergence, Space Club Convergence, Cross-section Time and Space Coupling Club Convergence and Stochastic Time and Space Coupling Club Convergence.Second, the important step of testing club convergence hypothesis, that is regional classification to examine convergence, also needs to seek suitable methods from the aspects of time, space and time and space coupling. In the discussing of regional classification methods through comparing the advantages and disadvantages of related methods this thesis deems that Classification and Regression Tree Analysis can entirely consider similarities of various kinds of influence factors about economic growth and extract their important degree so it suits to divide time regional groups. And Exploratory Spatial Data Analysis has the advantage of including spatial factors so to be the suitable method to divide space regional groups. Then above two methods'combination suits to divide time and space coupling regional groups. In the discussing of examining convergence, this thesis firstly according to the unique global steady state in Neoclassical convergence model, many global steady states in Neoclassical convergence model, many local steady states in Neoclassical convergence model, the Neoclassical stochastic growth model and spatial expanding of classical club convergence model, respectively establishes the theoretical foundations for examining Cross-section Time Club Convergence, Stochastic Time Club Convergence and Space Club Convergence. Then under the above theoretical directions the thesis respectively determines according test methods. Finally the thesis deems that with regard to Time Club Convergence its distinguishing should abide by firstly using Classification and Regression Tree Analysis to divide time regional groups and then employing cross-section approach or panel approach to test Cross-section Time Club Convergence in the same group and testing Stochastic Time Club Convergence with time series analysis methods and panel unit root test. With regard to Space Club Convergence the thesis deems that its distinguishing should abide by firstly using Exploratory Spatial Data Analysis to divide space regional groups and then choosing suitable spatial autoregressive model or spatial error model and spatial cross-regressive model to examine Space Club Convergence in the same group. In addition the distinguishing of Time and Space Coupling Club Convergence is to simultaneously test the club convergence of time dimension and space dimension.Third, there exit three time convergence clubs, two space convergence clubs and two time and space coupling convergence clubs among Chinese regions between the periods from 1990 to 2006. And the sequence of convergence rate from high to low is time and space coupling convergence clubs to time convergence clubs to space convergence clubs. During dividing time regional groups the thesis thinks that choosing prediction variables should firstly follow regional group limits of club convergence definition and then consider important effect factors of economic growth especially those factors which have Chinese characteristics. So this thesis altogether chooses 34 indexes as prediction variables in CART which estimate the following 12 aspects such as inertia effect, investment level, population structure, industrial structure, marketization level, FDI level, infrastructure level, location condition, resource possession, the ratio of capital to labor, school degree and government policy. And this thesis designs 10 plans including straightforward administration cities or not and arranging force root or not. Through comparing the results of each plan the thesis finally gains 3 time regional groups and those groups are in more accordance with club convergence's characteristic of"within group convergence and between groups divergence"than traditional four regions. By using Exploratory Spatial Data Analysis this thesis gains 4 space regional groups which respectively stand for different spatial dependence patterns. And then through choosing simultaneous parts of time regional groups and space regional groups this thesis gains 10 time and space coupling regional groups. The results of cross-section regression, panel data analysis and unit root testing in the same time regional group entirely show that there exist Cross-section and Stochastic Time Club Convergence in the three time regional groups and this conclusion is very steady. The regression results of spatial autoregressive model, spatial error model and spatial cross-regressive model in the same space regional group entirely show that there exist two space convergence clubs namely HH group and LL group. In addition following the procedure of examining space convergence club to time convergence club in the same time and space coupling regional group this thesis finds that there exist two cross-section time and space coupling convergence clubs and two stochastic time and space coupling convergence clubs in 2HH group and 1LL group.The size relation of convergence rates among time convergence clubs, space convergence clubs and time and space coupling convergence clubs full shows that comparing to time convergence clubs space convergence clubs have the advantage of involving spatial factors but the similarities in the same space regional group limit in the aspect of spatial dependence mode of sole index and whether existing synthesized similarities of many indexes which affect initial economic growth and structure similarity like time regional groups certified by CART or not is indefinite. So the convergence rates of space convergence clubs generally are beneath the convergence rates of time convergence clubs. And time and space coupling convergence clubs have another similarity namely spatial dependence mode similarity comparing to time convergence clubs so their convergence rates are higher.Fourth, according to regard club convergence as one type of convergence and simultaneously consider its individualism and Chinese characteristics the thesis gains forming mechanisms of Chinese convergence clubs by using the methods from putting forwards hypothesizes to examine them.The forming mechanism of Chinese time convergence clubs can be stated that history factor (initial income), structure factor (employment structure) and geography factor (resource possession) are three factors which affect the conformations of time convergence clubs. And marketization force (nationalization level), government force (policy level) and interaction force are three forces affecting the conformations of time convergence clubs. The alone functions of three factors and collaborative functions of three factors under marketization force and government force cause growth aggregations of similar regions. Because of local growth spillovers these aggregations emerge convergence and then form time convergence clubs.The forming mechanism of Chinese space convergence clubs can be stated that under the effects of spatial heterogeneous economic behaviors and local geographic spillovers uneven spatial distributions emerge. That is two different spatial dependence mode namely HH group and LL group. The interaction strength among regions in HH group (LL group) is higher than the interaction strength between HH group and LL group (unified force is higher than separation force). So there emerge growth aggregations in HH group (LL group). With sustained growth the degree of aggregations increases so that economic growth disparities between two groups are produced and then both HH group and LL group form space convergence club.The forming mechanism of Chinese time and space coupling convergence clubs is integration of time dimension convergence clubs and space dimension convergence clubs.On the basis of the above discussions the author thinks that proceeding to impel marketization reform, striving to improve the levels of schooling and infrastructure and optimizing regional industrial structure can encourage local economic growth under the assistances of marketization force and government force so to narrow disparities. Furthermore local governments should break administrative boundaries, actively seek opportunities of economic cooperation, reinforce interregional economic relation so to spread the circumscription of growth spillovers and then to achieve coordinative development of interregional economy.