Bingxin ZHANG

Guangdong Agricultural，Industrial and Business Polytechnic College，Guangzhou 510530，China

1 Introduction

Since the reform and opening up，annual growth rate of GDP in China has achieved 9.7%.The urbanization rate，which is the proportion of urban population in the total population，has increased from 19.39%in 1980 to49.95%in 2010 and area of urban land has climbed from 6 720 km2to39 758.4 km2with the annual growth rate reaching 6.1%.Construction land，an important form of land use，is the carrier of production and life in the city，whose expansion provides resources and capital guarantee for the quick industrialization as well as urbanization and further boosts economic growth.However，economic growth will also result in the aggregation of resources and labor force，promoting the further expansion of urban land，which causes a series of problems，such as damaging the security of national grain and decreasing biodiversity.Therefore，it is quite significant to study the relationship between economic development and urban construction land.

The relationship between urban construction land and economic development is mainly studied from two theoretical perspectives.The first perspective is contribution of land to economic growth，which can be traced back to the view of classical economics that land is the mother of wealth，with Adam Smith and David Ricardo as the representatives.Some scholars in China has have tested by using the extended Solow Model.The empirical study on the regional［1－3］and national［4－6］level has showed that expansion of construction land can promote economic growth in China and land is an essential production factor.The second perspective is that economic development is the driving force for expansion of construction land，which has been widely accepted theoretically and proved from the urban，regional as well as national dimensions.Si Chenglan's［7］research on Nanjing city，Ge Chunye's［8］research on Chongqing city and Wu Dafang's［9］research on Zhuhai have supported the conclusion from the city level.As for regional and national dimension，Huang Jikun's research on regional disparity of land expansion has showed that the plane expansion of construction land in central and western areas is largely due to economic growth［10］.Chen Chun believes that GDP has the largest driving force for the change of construction land from national，eastern，central and western dimensions［11］.Zhao Ke's investigation on the driving force for land expansion by using panel data analysis has discovered that economic development，population increase and adjustment of industrial structure are the main driving force［12］.With spatial econometric model，Ye Hao has discovered that social economic development boosted land expansion not only within the area but also in the surrounding areas［13］.

The above mentioned researches have achieved significant progress，but only in one aspect of the relationship between economic development and urban land，which may cause deviation in evaluating the empirical model due to ignoring the endogeneity of variables and affect robustness of conclusion to some extent.This article investigates the mutual effect between economic development and urban land by using data of31 provinces，cities and autonomous regions（Hong Kong，Macao and Taiwan being not included）from 1997 to 2011.

2 Model setup，data processing and research method

2.1 Model setupThe following empirical model has been specified

In formula（1），LANDitstands for urban land scale（unit：km2）in i province，city or autonomous in year t，i＝1，2，…，31；t＝1997，1998，…，2011.AGDPitmeans real GDP per capita in province i in year t（unit：yuan）.Consumer price index based on 1996 was used to remove the effect of price on gross domestic product.βimeans cointegration coefficient in each panel unit.αirefers to different fixed effect in cointegration among each panel unit.δitstands for error term.In order to eliminate heteroscedasticity as much as possible，logarithmic treatment on LANDitand AGDPithas been conducted.

2.2 Data sources and processingData about the area of urban construction land，CPI and regional gross product per capita in the research comes from China Statistical Yearbook.If land area in one year is lost，average area of the previous and following year would be adopted.However，data of land area in Shanghai is seriously lost，this research takes urban land area as explained variable and time as explanatory variable and conducts regression treatment as well as autocorrelation processing on the sequence.The fitted values of those lost data are calculated based on the regression result.In the statistical yearbook，urban construction land area in Tibet in 2011 is nearly 50%less than that in 2010，which is obviously impractical，therefore this research adoptsbuilt-up area.

2.3 Research methodThis research mainly focuses on the long-term equilibrium and short-term dynamic relation between urban construction land and economic development.Test on their causality has mainly adopted the following analytical methods.Firstly，stationarity properties of economic development（ln AGDP）and urban land area（ln LAND）in 31 provinces between 1997 and 2011 are tested by using panel unit root method.Secondly，if ln AGDP and ln LAND are integrated in the same order，the cointegration relation is to be tested with Pedroni test，Kao test and Johansen－Fisher test and cointegration equation is to be evaluated.Lastly，if there is cointegration relation between ln AGDP and ln LAND，the equation of panel error correction of economic development and urban land scale will be established in order to judge whether there is causality based on the significance of regression coefficients.

3 Empirical analysis

3.1 Panel unit root testFor regression with time sequence，the stationarity should be first tested.Otherwise，the estimation would be wrong if the two non-stationary series are not integrated of the same order.But the estimation would be correct if there were cointegration relationship between the non-stationary variables of the same order.Panel data possesses the features of both cross-section and time sequence data，therefore there also exists nonstationary problem.Testing the cointegration between economic development and urban construction land requires the panel unit root test on their panel data.Supposing ln AGDP and ln LAND are in accordance with the following AR（1）：

In formula（2），yitrefers to panel data to be tested，namely panel ln AGDP or panel ln LAND.Parameterρiis autoregressive coefficient.xitmeans exogenous variable in the model，including fixed effect and time trend of cross section.i＝1，2，…，N，N stands for the number of cross section.t＝1，2，…，T，T is the observation period of i cross section.μitmeans stochastic error，supposing it is independent and homoscedastic.As for formula（2），if＜1，sequence is stationary（weak）；if｜ρi｜＝1，sequence yitis nonstationary.

Unit root test of panel datamainly involves LLC test，Breitung test，Hadri test，IPS test，F(xiàn)isher－ADF test and Fisher－PP test.LLC test，Breitung test and Hadri test are for the same root while Fisher－ADF test and Fisher－PP test for different roots.Null hypothesis of Hadri test is stationary panel sequence while that of other tests is nonstationary panel sequence.This research has adopted LLC test，Breitung test，IPS test and Fisher－ADF test to conductunit root test on lnAGD and ln LAND（shown in Table 1）to further judge the stationarity of panel data.

Table 1 Results of unit root test of ln AGDP and ln LAND

Based on the testing results，it can be concluded that original sequence of ln AGDP panel data was non stationary under LLC test，Breitung test，IPS test and Fisher－ADF test while its first order difference sequence was stationary，hence ln AGDP－I（1）.There was differentiation in stationarity test on original sequence of panel ln LAND.Under the 5%significance level，LLC test could reject null hypothesis and accept alternative hypothesis and each section sequence was stationary.However，Breitung test，IPS test and Fisher－ADF test could not reject the hypothesis of nonstationary ln LAND.Since both IPS test and Fisher－ADF test adopted the hypothesis that there were different unit roots in each cross section sequence in panel data，there was obvious difference in economic development as well as construction land scale among31 provinces in China.Therefore，the research believes that original sequence of ln LAND is nonstationary and its first order difference sequence is stationary under the 1%significance level.Panel data of ln-LAND is also first order cointegration sequence，namely ln LANDI（1）.

3.2 Panel cointegration testCointegration theory believes that though individual variable sequence may be nonstationary，linear combination of nonstationary sequence may be stationary.So，if there is cointegration relation between ln AGDP and ln-LAND，there would be no spurious regression in their regression analysis.The testing results show that both ln AGDP and ln LAND are first order integrated panel sequence and are accessible to cointegration test.

Pedroni test（1999）and Kao test（1999），as the methods for panel data cointegration test，are based on Engle and Granger two-step method as well as Johansen－Fisher，which is based on the theory of Johansen cointegration test.This research adopts Pedroni test to check whether there is cointegration relation between ln AGDP and ln LAND and the testing results are shown in Table2.

Table 2 Results of Panel Pedroni Cointegration Test

There are altogether seven statistical values in Pedroni test，namely Panel V，Panel Rho，Panel PP and Panel ADF，Group Rho，Group PP and Group ADF.Panel V，Panel Rho，Panel PP and Panel ADF are all described with dimensions within groups while Group Rho，Group PP and Group ADF are described with dimensions between groups.According to Pedroni research conclusion，under the condition of small samples，statistical values of Panel ADF and Group ADF possess higher quality than those of others while Panel V and Group Rho have worst effect，which should be followed when testing results are different.This research judges whether there is cointegration relation between ln AGDP and ln LAND based on the statistical values of Panel ADF and Group ADF.According to Table 2，all statistical values except Panel V and Group Rho can reject the hypothesis that there isno cointegration relation between ln AGDP and ln LAND under the significance leve of 1%，therefore it can be concluded from Pedroni test that there is cointegration relation between ln AGDP and ln LAND.

3.3 Evaluation of panel cointegration equationResult of cointegration test shows that there is cointegration relation between panel sequence ln AGDP and ln LAND.Construction land is an essential production factor for urban economic development and the carrier for all the production and living activities in the city while economic development also boosts land expansion.Therefore，urban land use and economic development have mutual effect on each other.Direct OLS regression on formula（1）may cause deviation and low efficiency in evaluating results because of endogenous issues and correlation between explanatory variable and error term.With first－order lag of ln AGDP as the instrumental variable，this research conducts regression with two stage least square method and reaches the following results.

In formula（3），1.963 7 is average section effect.αirefers to individual section effect deviating from average section effect.It is not important and thus not listed.Based on each statistical value，model fitting is relatively ideal.Cointegration equation shows that urban land elastic index of economic development achieves 0.494 9.Per capita GDP increasing by 10%will drive4.94%increase in urban land expansion because economic development promotes accumulation of factors.Besides，industrial development and growing population require more space in the city.Therefore，economic development is an essential driving force for urban land expansion.In order to further test the cointegration relation between ln AGDP and ln LAND，this research has conducted panel unit root test on the residuals in formula（3）and the results are shown in Table 3.

Table 3 Results of Panel Residual Unit Root Test

According to Table 3，conclusion of LLC test，F(xiàn)isher－ADF test and Fisher－PP test was that panel residual was stationary，which proved that there was cointegration relation between ln AGDP and ln LAND and formula（3）was cointegration equation.

3.4 Causality testThere is cointegration relation between ln AGDP and ln LAND，which means that there is at least one－direction causality in the long term，but the direction of causality should be established through panel error correction model，which is shown as follows.

In formula（4）and formula（5），the meanings of i and t are the same as those in formula（1）.ECM refers to error correction term.Δln LAND and ln AGDP stands for the first order differential sequence of ln LAND，ln AGDP.k means lag order number.K refers to the maximum lag order number，which is 2 in this research.λi1andλi2are adjustment coefficients and are in accordance with reversed correction mechanism when their values are within（－1，0），which means that the short-term dynamic relation between variables deviating from long-term equilibrium will be adjusted to equilibrium at a certain speed.αi0，αi1，αi2，βi0，βi1andβi2are regression coefficients.δitandφitare error terms and are independent.Table 4 shows the results of error correction model estimation（constant terms and section effect terms being omitted）.

Table 4 Results of Panel Error Correction Model Estimation

Table 4 shows the results of panelerror correction model estimation.According to the coefficients of error correction terms，error correction coefficients in the equation of economic development are negative with a small value and are insignificant，which means that in the long term，equilibrium between economic development and urban construction land does not adjust the imbalance of economic development.However，in error correction equation of urban construction land，coefficient of error correction term is－0.3383 and is highly significant.Therefore，equilibrium can inhibit the short-term imbalance of urban construction land.The economic meaning of the conclusion is that imbalance of urban construction land cannot be completely adjusted by systematic power and requires external strength like political measures.

Besides，causality between variables can be judged according to coefficients of error correction terms and the significance.In the long term，economic development is the Granger cause of urban land expansion while the reverse relationship cannot be proved by empirical study.The conclusion has economic meanings.Firstly，further improvement of economic development and urbanization rate will cause steady urban land expansion.Therefore，governments at all levels should strictly follow the guideline of＂highly cherishing each inch of land＂and occupy cultivated land as less as possible.Secondly，change in urban construction land is not the Granger cause for economic growth in the long term，which has proved the equation of economic growth in new classical economics.And the reasons may be concluded as follows.Expansion of urban construction land is periodical and regular.Quick land expansion happens in the initial and middle period of urbanization while land expansion will be slow and constant during the mature phase of urbanization.Moreover，production factors like capital，labor force and technology replacing land will result in decoupling between land expansion and economic growth.Therefore，governments at local levels should gradually abandon the city development mode of relying on land management，actively chang the mode of economic growth，increase the utilization of the current land and transform the economic development from being driven by factors to efficiency.

Lastly，in the short term，economic development is affected by the inertia property and urban land expansion，but effect of land expansion is insignificant and with time delay.However，urban land expansion is affected by not only internal changes but also economic development.And impact of economic development on urban land is much more significant.From the analysis of causality，economic development and urban construction land are the Granger reasons of each other in short-term.

4 Conclusion and discussion

This research has tested the relationship between economic development（ln AGDP）and urban construction land（ln LAND）by using the data of 31 provinces，cities and autonomous regions in China and reached the following conclusions.

Firstly，panel data of economic development（ln AGDP）and urban construction land（ln LAND）is first-order integrated，which means ln AGDP－I（1）and ln LAND－I（1）.

Secondly，there is cointegration relation between economic development（ln AGDP）and urban construction land（ln LAND），namely long-term equilibrium.Cointegration equation has showed that10%of increase in GDP per capita would drive4.94%of expansion in city construction land.

Thirdly，results of panel error correction model regression have showed that economic development is the Granger cause of urban land expansion in the long term while the reverse relationship has not been proved.In short-term，economic development and urban land expansion are Granger causes of each other.The long-term equilibrium can only regulate the short-term imbalance of urban land，arriving at33.83%.

This research studies the mutual influence of ln AGDP and ln-LAND，which are endogenous variables in nature.In fact，there are other factors like capital，labor force，technology，human resources and systems besides land can affect economic development.Apart from economic development，other variables，such as population，urbanization rate，policy，natural geographical environment，industrial structure adjustment and transportation，can also affect land scale.Based on this research，further researches can include other variables as exogenous variable or control variables to study the relation between ln AGDP and ln LAND by using spatial econometrics analysis.

［1］LIMY，HU ZZ.Empirical study on contribution rate of land to economic growth in Shanghai［J］.Soft Science，2005，19（6）：21－23.（in Chinese）.

［2］WUCF，YANG ZR.Comparatively atudy on land conversion driving factor：Theory and empirical analysis［J］.Journal of Zhejiang University（Humanities and Social Sciences），2008，38（2）：29－37.（in Chinese）.

［3］YU Y，LU XH.Contribution rate of construction land to the secondary and service industry growth：A case study of Wuhan City［J］.Areal Research and Development，2010，29（3）：124－128.（in Chinese）.

［4］FENG L，WEIL，JIANG Y.Study on the contribution of land element to economic growth in China［J］.China Land Science，2008，22（12）：4－10.（in Chinese）.

［5］TAN SK，RAOYX，ZHU XB.Study on the influence of land investment on regional economic growth［J］.China Polulation.Resources and Environment，2012，22（9）：61－67.（in Chinese）.

［6］LIX，ZHANGRP，OUMH，et al.Study on construction land contribution to secondary＆service industry and its spatial correlation［J］.China Polulation.Resources and Environment，2011，21（9）：64－68.（in Chinese）.

［7］SICL，ZHOU YK.Analysis of the changes and the driving forces in land for construction for them in Nanjing［J］.Social Sciences in Nanjing，2008（11）：139－145.（in Chinese）.

［8］GECY，TU JJ，LIK，et al.Analysis of the changes and the driving forces in land for construction for them in Chongqing［J］.Journal of Southwest Agricultural University，2008，30（10）：131－135.（in Chinese）.

［9］WU DF，LIU YY，DONG YX，et al.Analysis on the spatio-temporal change characteristics of construction land and its driving forces in Zhuhai［J］.Economic Geography，2010，30（2）：226－232.（in Chinese）.

［10］HUANG JK，ZHU LF，DENGXZ.The regional difference of China's construction land expansion and its influencing factors［J］.Science in China（Series D），2007，37（9）：1235－1241.（in Chinese）.

［11］CHEN C，F(xiàn)ENG CC.Driving forces for construction land expansion in China［J］.China Polulation.Resources and Environment，2010，20（10）：72－78.（in Chinese）.

［12］ZHAOK，ZHANGAL，LIP.Driving forces of urban construction land expansion：An empirical analysis based on panel data of provinces［J］.Journal of Natural Resources，2011，26（8）：1323－1332.（in Chinese）.

［13］YEH，ZHANGP，PU LJ.Spatial econometrics study of relationship between regional socio-economic development and construction Land in China［J］.Scientia Geographica Sinica，2012，32（2）：149－155.（in Chinese）.