Analyzing Producivy Growh: Evidence from China s Manufacuring Indusries Kui-Wai Li a* and Tung Liu b a Universy of Geneva and Cy Universy of Hong Kong b Ball Sae Universy, USA *Corresponding auhor: Tel.: +852-34428805; fax: 852-34420195 E-mail address: efkwli@cyu.edu.hk (Kui-Wai Li)
Analyzing Producivy Growh: Evidence from China s Manufacuring Indusries Absrac This aricle examines he growh aribues of manufacuring indusries in China for he sample period of 1999-2007. All manufacuring indusries are grouped ino and four main indusry groups and four geographical regions. A revised Solow s growh mehod is used o decompose he growh aribues ino inpu growh, scale effec, echnical progress, and echnical efficiency change. A sochasic fronier model is applied o he ranslog producion funcion. The empirical findings show a srong presence of echnical progress, while labor inpu has rapidly been replaced by human capal. Srucural ransformaion in he indusrial secor is eviden, so as regional imbalances. Keywords: China indusries, producivy, efficiency, echnical progress, sochasic fronier model JEL classificaion: L60, O14, O47, O53
I Inroducion China since he early 1950s had adoped socialis collecivizaion from he Sovie Union and pursued mainly heavy indusries. Economic reform in 1978 came wh a change of ideology under he lae Deng Xiao-ping s moo ha does no maer wheher is a black or whe ca, so long as caches mice. A he ime of economic reform in 1978, here was a coexisence of excess supply in heavy indusrial goods and excess demand in consumer goods (Perkins, 1988, 1994; Wu, 2005). The producion of ligh manufacuring indusries began o grow wh he esablishmen of Special Economic Zones ha provided invesmen advanages and low producion cos along he coas of Souhern China. Alhough indusrial reform began in he mid- 1980s, was no unil he mid-1990s when Shanghai was designaed for developmen in highechnology indusries (Wu, 2005). Wh a devaluaion of 30 percen in he yuan in 1994, China s manufacured expors have since expanded coninuously. According o he Naional Bureau of Saisics of China in 2006, manufacuring is he larges secor occupying 48.3 percen of real GDP, followed by service secor wh 40.2 percen and primary secor has fallen o 11.5 percen. Empirical sudies on China s manufacuring secor have concenraed on a number of discussion areas. One area of discussion relaes o he ransformaion o non-sae-owned enerprises (Perkins, 1988). The 1997 sae-owned enerprises reform has promoed he sraegy of grasping he large, releasing he small (juada fangxiao) and expanded he number of nonsae owned enerprises (Wu, 2005). In dealing wh he daa problem, he repored indusrial oupu daa have been adjused by differen benchmark years and benchmark indusries in Ren and Zheng (2006), Wu (2002) and Maddison and Wu (2008). Szirmai e al. (2005) has poined ou ha he consrucion of China s indusrial daa is needed more a he beginning sage of he reform era when daa were missing, bu daa reliabily has increased in recen years. Anoher area of discussion relaes o he producivy and efficiency performance of indusries, as sudies have been conduced on individual indusries, such as iron and seel, oil and aerospace, elecommunicaion and insurance (Jefferson, 1990; Movshuk, 2004; Ma e al., 2002; Mu and Lee, 2005; Yao e al., 2007). In sudying producivy changes in China s indusries, he Malmquis index ha decomposes producivy ino efficiency and echnological changes has been applied in Ma e al. (2002) and Movshuk (2004), bu Sun e al. (1999) and Yao e al. (2007) have argued ha sochasic fronier and daa envelopmen analyses are more effecive in measuring echnical efficiency in China s indusries. 2
Based on he daa of 161 hree-dig manufacuring indusries in 31 provinces for he recen sample period of 1999-2007, his aricle sudies he growh facors in China s manufacuring indusries, and in paricular, he issues of indusrial producivy, echnological progress and efficiency (Aigner e al., 1977; Kumbharkar and Lovell, 2000). In he empirical analysis, we apply he sochasic fronier model o he ranslog producion funcion. This approach relaxes Solow s (1975) assumpion of consan reurn o scale and opimal producion capacy. The ranslog producion funcion allows a flexible nonlinear funcional form wh nonconsan reurns o scale, while he sochasic fronier model provides he possibily of deviaions beween acual and opimal oupu due o echnical inefficiency. Secion II shows he indusrial daa and he descripive performance of indusries in China. Secion III discusses he mehodology ha decomposes TFP ino hree separae aribues, while Secion IV repors he empirical resuls and he performance of differen indusries across provinces. The las secion concludes he paper. II Indusrial Daa and Performance The online daa Suppor Sysem for China Saisics Applicaion from he All China Markeing Research (ACMR) in Beijing provide a comprehensive indusrial oupu and relaed informaion for a large number of four-dig manufacuring indusries for he 31 provinces in China since 1999. The daa for he 1999-2002 years show a oal of 511 four-dig GB/T-94 coded manufacuring indusries, bu since 2003 he number has been reduced o 473 four-dig GB/T-2002 coded indusries. These wo ses of indusry codes are compared, compiled and adjused o eliminae classificaion and reporing inconsisencies. The sandardized and adjused four-dig manufacuring indusries for he sample period of 1999-2007 are hen aggregaed ino 161 hree-dig and 29 wo-dig indusries, which are furher grouped ino four main indusrial groups according o heir dig caegory and level of echnology (see Appendix Table A1). While he Processing Indusry is based more on agriculure or raw maerials, he Ligh Manufacuring Indusry relaes mainly o labor-inensive indusries. The Meal and Machinery Indusry conains he radional heavy indusries, while he High-echnology Indusry relaes more o he modern echnology-inensive indusries. Wh a oal of 161 indusries in 31 provinces from 1999 o 2007, gives abou 29,812 daa values afer excluding missing observaions (see Appendix Table A2). 3
The 31 provinces and auonomous areas in China are geographically grouped ino four regions. The Easern region includes he eleven provinces of Beijing, Tianjin, Hebei, Shanxi, Shanghai, Jiangsu, Zhejiang, Anhui, Shandong, Henan, and Hubei. Some provinces in he Easern region are coasal provinces ha experienced high inial economic growh and expor expansion afer he open door policy. The Souhern region conains he eigh provinces and auonomous areas of Fujian, Jiangxi, Hunan, Guangdong, Guangxi, Hainan, Chongqing, and Sichuan. Since economic reform in 1978, he Souhern region adjacen o Hong Kong and Taiwan has been he recipien of foreign direc invesmen ha concenraed in labor-inensive manufacuring. The Wesern region is remoe and no easily accessible, and consiss of a oal of nine inner provinces and auonomous areas of Inner Mongolia, Guizhou, Yunnan, Tibe, Shaanxi, Gansu, Qinghai, Ningxia and Xinjiang. The Norheasern region adjacen o Japan and Souh Korea is he radional heavy indusry area ha composes of he hree provinces of Jilin, Heilongjiang and Liaoning. Table 1 summarizes he daa on he four groups of indusries in four regions in he sample period. The oal numbers of enerprises beween 1999 and 2007 have doubled. The Meal and Machinery Indusry is he larges indusrial secor in erms of indusrial sales oupu value and he number of enerprises and employees. The Processing Indusry is he second larges indusry in erms of he number of enerprises and employees, indicaing ha he Processing Indusry is sill based mainly on raw labor inpus. However, s indusrial sales oupu value is he smalles among he four indusrial groups. The Ligh Manufacuring Indusry has expanded since economic reform in 1978, wh he oal numbers of enerprises and employees close o, bu sill lower han hose in he Processing Indusry. However, due o heir high value-added conen, he Ligh Manufacuring Indusry has overaken he Processing Indusry for he indusrial sales oupu value. The High-echnology Indusry has experienced he highes growh in he number of enerprises (288%) and employees (284%) beween 1999 and 2007 among he four indusrial groups, and s sales oupu value in 2007 has reached he level ha lies beween he Processing Indusry and he Ligh Manufacuring Indusry. Mos imporanly, he High-echnology Indusry has he highes percenage of indusrial sales oupu value designaed for expor, abou hree o four imes higher han he oher hree indusrial groups. The expor share in he Processing Indusry oupu is he second larges, bu s share has dropped considerably beween he wo sample years. Boh he Ligh Manufacuring Indusry and he Meal and Machinery Indusry 4
show he lowes expor poenial wh mild changes only, suggesing ha such oupus are geared mainly for he domesic marke. The regional performance in he sales oupu value and he numbers of enerprises and employees are similar among he four indusrial groups. The Easern region has he larges shares in sales oupu values and numbers of enerprises and employees. Each of hese values in he Easern region is much larger han he sum of he respecive values in he oher hree regions. The Easern region is hus he cener of he manufacuring indusrial producion. Beween he wo sample years, sales oupu value and he numbers of indusrial enerprises and employee have expanded mos in he Easern region, followed by he Souhern region. The oal number of indusrial employees in he weaker Wesern and Norheasern regions has remained que saic. Among he four regions, he Souhern region is sronges in erms of expor delivery, followed by he Easern region. Beween 1999 and 2007, however, is only he High-echnology Indusry ha has significanly expanded in he percenage share of value of expor, noably in he wo richer Easern and Souhern regions. All oher indusries and regions show eher a significan drop or a very mild increase in he performance of expor beween he wo sample years. Prior o he 1997 reform in sae-owned enerprises (SOEs), a large number of convenional SOEs were loss-making, faced wh huge iner-enerprises debs (he riangular debs) and relied heavily on sae subsidy. The percenage of loss-making enerprises has declined by half in many cases beween 1999 and 2007, especially in he wo prosperous Easern and Souhern regions. The Easern region has he lowes percenage of loss-making enerprises, while he weaker Wesern region sill experiences a large (over 20%) percenage of loss-making enerprises. The values of oal indusrial oupu deflaed by he GDP deflaor are used as he dependen variable in he producion funcion. The independen variables include labor, physical capal and human capal. The oal fixed asses deflaed by he invesmen index are used as he proxy variable for physical capal. Since here is a lack of daa on he educaional level of indusrial workers, a reliable proxy alernaive would be he wage paymens and relaed labor coss which are included in he operaing expenses. Deflaed by he GDP deflaor, he operaing expenses is hus used as he proxy for human capal. 1 1 The missing 1999-2002 operaing expenses daa are esimaed by inerpolaion from he 2003-2007 daa. 5
Indusry / Region Table 1 Performance of Indusries in he Four Regions China: 1999 and 2007 Toal No. of Toal No. of Indusrial Sales % of Value of Enerprises Employees Oupu Value Expor Delivery % of Lossmaking Enerprises (10,000) (Million 2,000 yuan) 1999 2007 1999 2007 1999 2007 1999 2007 1999 2007 All Indusries 139,080 303,513 4,241 6,669 5,743 27,131 18.49 21.13 27.30 13.49 Processing 39,919 77,844 1,188 1,804 1,475 4,998 24.27 18.41 29.80 13.41 Easern 23,666 50,747 716 1,075 904 3,141 23.67 18.78 24.16 12.43 Souhern 10,525 19,513 310 565 391 1,260 31.97 21.98 36.56 13.92 Wesern 3,113 2,946 82 76 103 268 5.83 4.10 42.92 22.74 Norheasern 2,615 4,638 80 88 77 329 16.88 12.77 38.01 16.00 Ligh Manufacuring 32,002 64,157 906 1,231 1,315 5,925 11.63 11.14 25.80 14.37 Easern 18,692 38,172 528 677 808 3,657 10.40 11.05 20.00 12.21 Souhern 8,361 18,367 219 379 293 1,288 19.11 15.53 32.78 14.93 Wesern 2,519 2,979 67 82 74 367 4.05 3.00 36.92 33.23 Norheasern 2,430 4,639 92 93 140 613 7.14 7.34 34.94 17.81 Meal & Machinery 53,310 121,607 1,706 2,380 1,933 10,750 12.11 12.64 26.58 12.57 Easern 32,697 76,870 986 1,371 1,246 6,763 10.19 12.05 21.01 10.74 Souhern 12,924 31,123 397 680 432 2,490 20.14 16.47 35.75 13.73 Wesern 3,778 4,332 135 135 97 464 4.12 5.39 38.27 30.84 Norheasern 3,911 9,282 188 194 158 1,033 10.13 10.55 31.58 15.34 High-echnology 13,849 39,905 441 1,254 1,020 5,458 31.08 51.19 26.31 15.03 Easern 8,360 24,237 225 600 559 3,176 22.54 47.29 22.85 12.54 Souhern 4,201 13,462 167 598 400 2,053 43.50 60.30 30.45 18.60 Wesern 503 489 18 21 21 60 9.52 6.67 39.56 30.67 Norheasern 785 1,717 31 35 40 169 37.5 29.59 32.61 17.82 Source: Suppor Sysem for China Saisical Applicaion, All China Markeing Research, Beijing. 6
The Indusrial Oupu Value a curren prices in housand yuan is he oal value of indusrial producs, wheher sold or ready for sales, in a cerain period of ime. I includes he value of all final producs which are warehoused afer inspecion and packaging, or do no need any furher processing. I also includes processing of foreign producs, home-made semiproducs, and changes in invenories. The Indusrial Oupu Value is calculaed wh he facory mehod, in ha each individual indusrial enerprise is regarded as a separae eny, and he final oupu of all he producion acivies is considered. This calculaion mehod avoids double couning whin he same enerprise. In he consrucion of he physical capal in he sudy of indusry producivy in China, invesmen figures have reliably been used as a proxy (Jefferson, 1989; Jefferson e al., 1992, 1996; Chen e al., 1988a, 1988b; Chow and Li, 2002; Li, 2003). The oal fixed asse invesmen expressed in housand yuan shows he enerprise s value of ne fixed asses ha includes liquidaion of fixed asses, consrucion in progress, and loss of funds. The operaing expenses expressed in housand yuan refers o firms expenses ha include saff wages, ravel, uilies, lease coss (excluding finance lease charges), repairs, saff welfare, saff educaion funds, union funds, labor proecion, labor insurance, he board of direcors fees, and managemen fees paid. 2 III Sochasic Fronier Model and Decomposion of Oupu Growh The growh aribues of indusrial oupu is divided ino inpu growh and TFP growh. In urn, he TFP growh is decomposed ino adjused scale effec, echnical progress, and echnical efficiency change (Kumbhakar and Lovell, 2000). The adjused scale effec shows he reurns o scale effec from combined inpus. Technical progress shows he rae of echnological change and is indicaed by an ouward shif in he indusry s producion possibily fronier. Technical efficiency change refers o a movemen from a posion whin o a posion on he producion fronier. 2 Oher ems in he operaions expenses are depreciaion of fixed asses, business promoion expenses, business enerainmen, elecronic equipmen running coss, secury coss, propery insurance companies coss, posal coss, foreign fees, prining coss, claims invesigaion coss, real esae ax, ravel ax, land ax, samp duy, meeing fees, legal fees, noary fees, consuling fees, amorizaion of inangible asses, amorizaion of long-erm prepaid expenses, heaing coss, aud fees, echnology ransfer fees, research and developmen fees, green fees and adverising. 7
The economic heory on efficien producion argues ha producers always produce a he maximized oupu level wh given inpus. The empirical esimaion on oupu, cos, and prof funcions could produce variaion in producion efficiency (Farrell, 1957). The sochasic fronier producion funcion whou random shock can be saed as: y i f x )exp( u ), (1) ( i i where y i is he observed scalar oupu and x i is a vecor of inpus for i h firm. The posive value of u i is supposed o measure echnical inefficiency. Technical efficiency can be wren as: yi TEi exp( ui), (2) fi which is he raio of observed oupu o maximum feasible oupu. I shows he oupu of he i h firm relaive o he sochasic fronier oupu ha could be produced by a fully-efficien firm uilizing he same vecor of inpus. Such an oupu-oriened measure of echnical efficiency akes a value beween zero and one. If TEi 1, hen he firm is echnically efficien. By incorporaing echnical progress ino he echnical inefficiency specified in Equaion (1), we represen he producion funcion a ime, whou he subscrip i for firm, as: y f ) u ( x 1, x2,, xk, e, (3) where y is oupu and x j is he j inpu, j 1,2,, k, a ime. Taking logarhm-differeniaion of Equaion (3) wh respec o ime, gives: f x j f 1 u y x j, (4) x f f j j where y 1 y is he growh of oupu and y x 1 j x j is he growh of inpu j x j x. The echnical progress is f A 1 y u. Technical efficiency is TE e, while he growh of he f f echnical efficiency is TE u. Denoe e j f x j as he oupu elasicy wh respec o x f j inpu x j. The oupu growh can be represened as: 8
y e x A TE. (5) j j j The firs erm in he above decomposion can furher be decomposed ino wo differen erms using he cos minimizaion condion. Consider he cos minimizaion problem of he objecive funcion: min x j C, where C j w j x j, subjec o he consrain in Equaion (3). In he Lagrangian form, he objecive funcion and he consrain are wren as: L ( x j j u, ) w j x j ( y fe ), (6) where is he Lagrange muliplier. The firs-order condion for minimizaion is: w j f x u e. (7) j Muliplying boh sides by x j, w j x j f x j x j e u f x j x f j fe u e j y. (8) Taking he sum for all inpu, he oal cos w j x j j j j C is: C e y e y, (9) where e is he sum of oupu elasicies o inpu. I can be shown ha e is a measure of e j j reurns o scale. Suppose changes in all inpus have he same scale, x j ax j. Consider he changes in oupu f by aking he oal derivaive of f x, x,, x, ) and subsuing x ino f, we have: j ax j ( 1 2 n f f f f ax j x j f fa fa e j fa afe A j x j j x j f f j. (10) 9
Whou considering echnical progress, he producion funcion shows increasing (consan, decreasing) reurns o scale when e 1 (= 1, < 1). Dividing Equaion (8) by Equaion (9), he cos share for inpu j is: s j w j j j. (11) C x e e This shows ha he cos share is always equal o he relaive oupu elasicy in he case of cos minimizaion. For he consan reurns o scale, e 1, he cos share is equal o oupu elasicy. Insering e 1 ino Equaion (5) and rearrange erms, we can rewre he oupu growh as: e y e e j j x j e x j A ( 1) TE. (12) j e j e Using he cos share Equaion (11), y s j x j e 1) j ( s x A TE. (13) j j j In he above, oupu growh is decomposed ino four sources: inpu growh adjused scale effec ( e 1), echnical progress s x, A, and he growh of echnical efficiency TE. The firs erm represens he conribuion of inpu growh o he oupu growh. The growh of aggregae inpu ( ) is he weighed sum of all inpu growh. The weigh is he cos share of he inpu, which is also he raio of oupu elasicy of an inpu. The second erm is he adjused scale effec. The conribuion of increasing reurns o scale o oupu growh is posive ( e 1) 0, and he scale effec of ( e 1) is adjused by he growh of aggregae inpu ( ). For consan reurns o scale ( e 1) or zero inpu growh ( 0 ), he adjused scale effec is zero. The hird erm A is a measure of echnical progress and he las erm TE refers o he change in echnical efficiency. This decomposion is differen from Solow s growh decomposion (Solow, 1957) in wo ways. Firs, his decomposion allows non-consan reurns o scale. Second, considers he change in echnical efficiency. j j j 10
Wh he decomposion of oupu growh as shown by Equaion (13), we can easily derive he decomposion of he growh of TFP. Define he TFP for a producion funcion wh muliple inpus a ime as: TFP y, (14) where is he aggregae inpu. Assuming j j j j s x j j, he growh of TFP is: TFP ( e 1) s x TE. (15) The decomposion of oupu and producivy growh shown in Equaions (13) and (15) will empirically be applied o he daa. The sochasic fronier model assumes deviaions from he efficien fronier wh random shock (Aigner e al., 1977). The specificaion of echnical inefficiency in Equaion (1) migh also capure oher random shocks ha are eher beyond he conrol of he firm or no direcly aribuable o he underlying echnology. The random shocks can be included in he ranslog producion fronier funcion by adding a wo-sided error erm (Greene, 1980). 3 The sochasic fronier model wh panel daa hen becomes: lny K L H 2 2 2 0 K ln L ln H ln KK (ln ) LL (ln ) HH (ln ) KL ln K ln L ln K ln H ln L ln H D v u, (16) KH where i =1, N indusries and =1, T; lny is he log of real indusrial oupu for i h indusry a ime. ln K is he log of oal fixed asse invesmen used as a proxy for physical capal, ln L is he log of oal number of employed workers, ln H is he log of operaing expenses used as a proxy for he human capal variable. LH K D is he ime dummy variable ha capures echnical progress and he parameer can be used o measure echnical progress over ime. The random 2 error v is symmeric and normally disribued wh v ~ N (0, ). The echnical inefficiency erm u can eher be ime invarian or ime varian (Kumbhakar and Lovell, 2000). In he case v L H of ime invarian echnical inefficiency, 2 u ui ~ N (, u ), where μ is he mode of he 3 We follow he classical growh model wh exogenous inpus. If he endogeney of inpus occur, he esimaed coefficiens will be biased and he conclusion from his paper may be conservaive. Liu and Li (2006) conrols endogeney of human capal by applying he wo lags of human capal as insrumens. However, heir esimaion of he producion funcion does no include echnical inefficiency. 11
runcaed half-normal disribuion. In he case of ime varian echnical inefficiency, u can be expressed as a monoonic decay funcion as u u, where exp( ( T)), and is an i unknown scalar parameer for echnical inefficiency. u can eher be increasing (if 0 ), decreasing (if 0) or remained consan (if 0 ) (Baese and Coelli, 1992). The minimummean-square-error predicor of he echnical efficiency of he i h indusry a ime is shown as (Baese and Coelli, 1988, 1992, 1995; Baese and Corra, 1977; Coelli, 1996; Kumbhakar and Lovell, 2000): TE ( u where v u. E exp( ) ), (17) Ignoring he random shock erm, he decomposion of oupu growh, Y, and oal facor producivy growh, Y TFP TFP, can be derived from Equaions (13) and (15) as: Scale TE, (18) Scale TE, (19) ek el e H K L H, (20) e e e Scale ek el e H ( e 1)( K L H ), (21) e e e where K, L, and H are growh of inpus; e is reurns o scale. e K, e L, and e H are oupu elasicies for physical capal, labor, and human capal, respecively: ek 2 ln K ln L ln H, (22) K KK KL KH el eh 2 ln L ln K ln H, (23) L H LL HH KL KH LH LH 2 ln H ln K ln L. (24) To derive he esimaes for he componens in he growh and producivy decomposion, we firs use he maximum likelihood mehod o esimae he parameers in Equaion (16). Subsuing he esimaed coefficiens of ' s in Equaion (16) ino he above hree equaions gives ê K, L ê, and ê H. The esimaes of oupu elasicies and individual inpu growh can be used o esimae he firs wo componens of oupu growh: growh of aggregae inpu and 12
adjused scale effecs. Given he esimaed coefficien of in Equaion (16), he esimae of echnical progress is ˆ 1 ( ˆ ˆ 1). In he case of he ime-varying decay echnical T 1 efficiency u exp( ( T )) u, he change in echnical efficiency can be esimaed by: i TE ˆ 1 ˆ exp( ˆ( T )) uˆ ii NT, (25) i, where û is he esimae of echnical efficiency and ˆ is he esimae of ime-varying decay parameer. For he ime invarian echnical efficiency, we have 0 and T 0 for all. E IV Empirical Resuls The sochasic fronier producion shown in Equaion (16) conains a oal of eigh ime dummy variables wh 1999 as he benchmark year ha serve o capure echnical progress over ime. When he regression includes he indusrial daa from all provinces, we include hree regional dummy variables of Souhern, Wesern, and Norheasern regions, while he Easern region is regarded as he benchmark (Szirmai e al., 2005). In esimaing echnical efficiency, we consider wo ypes of sochasic fronier models wh ime-invarian echnical inefficiency, inefficiency, i u u for all, and ime-varying decay echnical i u exp( ( T )) u. For he ime-invarian echnical inefficiency model, he esimae of he change of echnical efficiency is zero, namely TE 0. For he ime-varying decay echnical inefficiency model, a posive esimae of he ime-varying decay parameer gives posive TE and implies an improvemen of echnical inefficiency. When we apply he ime-varying decay echnical inefficiency model o differen daa ses, he resul shows ha here are several cases wh negaive growh in echnical efficiency ( TE 0) for a given daa se. We find ha a negaive growh in echnical inefficiency is always accompanied by an unusual large echnical progress. However, he sum of esimaes of he echnical progress and he growh in echnical efficiency ( TE ) is close o he esimae of echnical progress for he imeinvarian echnical inefficiency model. We consider his is an idenificaion problem in he esimaion and use he resuls from he ime-invarian echnical inefficiency model. 13
Esimaes for he Aggregae and he Four Indusrial Groups Table 2 presens he esimaes of Equaion (16) for he aggregaes of all indusries and he four groups of manufacuring indusries. Columns (1) and (2) in Table 2 show, respecively, he esimaes for he ime-decay varian and ime-invarian echnical inefficiency models for he aggregaes of all indusries. We remove he Souhern region in column (1) since is no significan in he esimaion. The esimaes of µ show ha he mode for he runcaed half-normal disribuion of he echnical inefficiency erm u is significanly greaer han zero for column (1), bu no significan in column (2). The echnical inefficiency erm in Column (2) appears o have half-normal disribuion, insead of runcaed half-normal disribuion. In spe of he difference in he assumpion of ime-varian and ime-invarian echnical inefficiency, mos esimaed coefficiens from hese wo models are very close. Since he esimae for he ime-varying decay parameer η shown in column (1) is no significan, he ime-invarian echnical inefficiency model shown in column (2) is more appropriae han he ime-varying decay echnical inefficiency model shown in column (1). For he four indusrial groups, we only find ha he High-echnology Indusry has he ime-varying decay echnical inefficiency propery, as shown in column (6). The esimaed coefficien of he ime-varying decay parameer (η) for he High-echnology Indusry is 0.034, which indicaes an improvemen in echnical efficiency. The res of he hree columns for he oher hree indusrial groups conain he esimaes from he ime-invarian of echnical inefficiency model. Table 2 shows ha almos all esimaed coefficiens of he inpus (physical capal, labor, 2 and human capal) are significan a 5 percen level. We apply es for he significance of he 2 six nonlinear erms. The saisics (nonlinear) are all significan and show ha he ranslog funcion is more appropriae han he Cobb-Douglas funcion for he aggregaes and all four indusrial groups. Because of he nonlinear relaionship in he ranslog funcion, he relaionship beween inpus and oupu should bes be described by he oupu elasicies (shown in Table 3). 14
Table 2 Regression Esimaes of he Indusry Aggregaes and Four Indusrial Groups All All Processing Ligh Meal and Indusries Indusries Indusry Manufacuring Machinery (1) (2) (3) (4) (5) Highechnology (6) ln K -0.143-0.138-0.588-0.222 0.032 0.148 (0.026) (0.025) (0.073) (0.037) (0.045) (0.083) ln L 0.994 1.017 1.411 1.112 0.933 0.534 (0.030) (0.029) (0.073) (0.047) (0.054) (0.090) ln H 0.631 0.707 0.696 0.427 0.729 0.999 (0.036) (0.035) (0.077) (0.064) (0.055) (0.116) ln K ln K 0.014 0.014 0.039 0.017 0.011-0.029 (0.003) (0.003) (0.007) (0.004) (0.005) (0.010) ln L ln L -0.022-0.021-0.026-0.014-0.011-0.059 (0.004) (0.004) (0.009) (0.007) (0.007) (0.013) ln H ln H -0.052-0.055-0.092-0.085-0.053-0.043 (0.006) (0.006) (0.013) (0.011) (0.009) (0.016) ln K ln L 0.006 0.004-0.017-0.007-0.004 0.101 (0.006) (0.006) (0.015) (0.010) (0.011) (0.022) ln K ln H 0.037 0.033 0.057 0.080 0.031 0.019 (0.007) (0.007) (0.015) (0.012) (0.012) (0.020) ln L ln H -0.043-0.043-0.070-0.074-0.040-0.056 (0.008) (0.008) (0.016) (0.014) (0.013) (0.022) Souh -0.155-0.129-0.109-0.201-0.171 (0.021) (0.051) (0.036) (0.031) (0.057) Norheas -0.169-0.236-0.165-0.192-0.292-0.203 (0.028) (0.029) (0.072) (0.051) (0.043) (0.080) Wes -0.432-0.544-0.514-0.570-0.573-0.474 (0.022) (0.023) (0.055) (0.041) (0.033) (0.069) Consan 6.473 7.171 6.711 7.724 5.107 5.942 (0.451) (44.028) (0.354) (38.965) (1.179) (0.392) µ 2.443 3.334 1.698 3.249 2.049 1.842 (0.443) (44.027) (0.204) (38.964) (1.166) (0.177) η 0.002 0.034 (0.002) (0.005) 2 ( nonlinear ) 369.20 382.26 203.19 263.01 70.69 59.59 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 2 ( regions ) 382.72 576.50 88.94 198.80 298.26 48.00 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 2 ( ime ) 336.28 (0.000) 5612.82 (0.000) 948.67 (0.000) 1400.91 (0.000) 3088.17 (0.000) 15.18 (0.056) Nobs 28,915 28,915 6,233 8,028 11,692 3,859 Noe: The figures in he parenhesis under he esimaes are sandard errors. The figures in he parenhesis under he chi-squares saisics are p-values. Nobs = number of observaions. The daa from Guizhou and Tibe are removed from he esimaion for columns (1) and (2). 15
Table 2 shows ha he esimaed coefficiens of regional dummies are all negaive and significan. The 2 ess for he join hypohesis es of all regional dummy variables, 2 (regions), are all significan as well. This implies ha oupu growh in he Souhern, Norheasern, and Wesern regions is smaller han oupu growh in he Easern region. The coefficien for he Wesern region is he smalles, suggesing ha he Wesern region has he slowes growh rae, as s growh is abou 0.5 percen lower han he Easern region. The resuls of hese dummy variables provide an evidence of imbalanced growh among he four regions. The esimaed coefficiens for he ime dummy variables are no shown in Table 2, bu 2 he join hypohesis ess for he coefficiens of ime dummy variables (ime) are displayed. All he es saisics are significan a 5 percen level for he aggregaes and indusrial groups, excep for he High-echnology Indusry. The p-value of he 2 (ime) es for he Highechnology Indusry is 5.6 percen, suggesing ha he echnical progress esimaes may no be significan for he High-echnology Indusry. Equaions (18) and (19) show he decomposion of oupu growh (Y ) and oal facor producivy growh (TFP ). To esimae he componens of oupu and producivy growh, we use Equaions (21) (25) o derive he esimaes for he oupu elasicy wh respec o he hree inpus of capal, labor and human capal ( e K, el and e H, respecively), reurns o scale ( e ), inpu growh ( ), adjused scale effec ( Scale ), rae of echnical progress ( ), and growh of echnical inefficiency (TE ). Table 3 shows hese esimaes for he indusry aggregaes and he four indusrial groups. Noe ha he esimaes for he indusry aggregaes and he hree indusrial groups are based on he model wh ime invarian echnical efficiency and TE 0 ; only he esimaes for he High-echnology Indusry are from he model wh ime-varying decay echnical efficiency. Table 3 produces a number of observaions. Labor has he larges oupu elasicy among he hree inpus. For he aggregaes of all indusries, he oupu elasicies for labor, human capal, and physical capal are 0.617, 0.501 and 0.335, respecively. The esimaes of hese hree elasicies in four indusrial groups are similar o hose for he aggregaes; he oupu elasicies for labor (beween 0.601 and 0.658) are he larges, followed by he oupu elasicies for human capal (beween 0.427 and 0.558) and he oupu elasicies for physical capal (beween 0.311 and 0.372). 16
The larges cos share is Labor. For he aggregaes of all indusries, labor s cos share is 42.5 percen; human and physical capal s shares are 34.5 percen and 23.1 percen, respecively. For he four indusrial groups, he cos shares for he hree inpus are similar o hose in he aggregaes. The Ligh Manufacuring Indusry has he larges labor cos share (44.6%). The High-echnology Indusry has he larges cos share in human capal (36.5%). The Processing Indusry has he larges cos share in physical capal (26.5%). These cos shares are in line wh expecaions given he labor-inensive and capal-inensive naure of differen indusrial groups. Table 3 Growh Decomposion for he Aggregaes and Four Indusrial Groups Oupu Elasicy Cos Share e K e L e e H s K s L Aggregaes 0.335 0.617 0.501 1.453 0.231 0.425 0.345 Processing 0.372 0.605 0.427 1.404 0.265 0.431 0.304 Ligh Manufacuring 0.322 0.618 0.446 1.386 0.232 0.446 0.322 Meal and Machinery 0.337 0.601 0.538 1.477 0.228 0.407 0.364 High-echnology 0.311 0.658 0.558 1.527 0.204 0.431 0.366 Inpu Growh Effec (%) Scale Effec (%) s K K s L L s H H e 1 (e 1) Aggregaes 2.00 1.50 3.39 6.89 0.45 3.08 Processing 1.34 0.85 2.98 5.17 0.39 2.04 Ligh Manufacuring 2.35 1.71 3.01 7.07 0.38 2.71 Meal and Machinery 2.07 1.38 3.67 7.12 0.47 3.35 High-echnology 2.08 2.56 3.73 8.37 0.52 4.35 Esimaed Y Scale TE TF P Y (1) (2) (3) (4) (5) (3)+(4)+(5) (7) (7)-(1) Aggregaes 18.06 6.89 3.08 8.10 0 11.18 18.19 0.13 Processing 14.63 5.17 2.04 7.43 0 9.46 14.51-0.13 Ligh Manufacuring 16.95 7.07 2.71 7.17 0 9.88 17.54 0.60 Meal and Machinery 19.83 7.12 3.35 9.36 0 12.71 19.92 0.10 High-echnology 20.56 8.37 4.35 0.92 6.92 12.19 20.26-0.30 Noe: Esimaed Y or esimaed growh of oupu is he sum of inpu growh ( ) and TFP. TFP is he sum of adjused scale effec (Scale), echnical progress ( ) and change in echnical efficiency (TE ). Y is he acual oupu growh and he las column shows he esimaion errors. s H 17
Human capal grows faser han he oher wo inpus. The inpu growh for he aggregaes of all indusries amouned o 6.89 percen, bu human capal has he fases growh wh 3.39 percen, while he growh for physical capal and labor are 2.0 percen and 1.5 percen, respecively. All four indusrial groups have similar paerns of inpu growh. There are signs of srucure change in he manufacuring indusries. The inpu growh has shifed from he Processing Indusry o he oher indusrial groups. The High-echnology Indusry has he highes inpu growh, followed by he Meal and Machinery indusry, he Ligh Manufacuring Indusry, and he Processing Indusry, wh 8.37 percen, 7.2 percen, 7.07 percen, 5.17 percen, respecively. The high inpu growh in boh High-echnology Indusry and Meal and Machinery Indusry is due mainly o he growh in human capal. The low inpu growh for Processing Indusry is because of low labor growh. Indusrial producion exhibs increasing reurns o scale. The sum of hree oupu elasicies for he aggregaes of all indusries and each of four indusrial groups is greaer han one. The reurns o scale for he aggregaes are 1.453; he reurns of scale for four indusrial groups have a range from he Ligh Manufacuring Indusry (1.386) o he High-echnology Indusry (1.527). Indusrial producion shows posive adjused scale effecs, due probably o he posive inpu growh and increasing reurns o scale. The adjused scale effec is 3.08 percen for he aggregaes of all indusries. Among he four groups of indusries, he High-echnology Indusry has he highes adjused scale effecs (4.35%). The inpu growh for Ligh Manufacuring Indusry and Meal and Machinery Indusry are similar, he adjused scale effec for he Meal and Machinery Indusry is higher since has a higher reurns o scale. The Processing Indusry has he lowes adjused scale effec, caused mainly by he low inpu growh. Only he High-echnology Indusry shows a posive improvemen of echnical efficiency. Based on he es saisics of he ime-varying decay parameer (no shown in he able), he aggregaes of all indusries and he res hree indusrial groups have no changes in echnical efficiency, namely TE 0. For he High-echnology Indusry, he improvemen of echnical efficiency is 6.92 percen and he echnical progress is only 0.92 percen. Technical progress is more imporan han inpu growh and adjused scale effecs for he aggregaes of all indusries and hree indusrial groups (Processing, Ligh Manufacuring and Meal and Machinery). For he aggregaes of all indusries, he conribuion of echnical progress, inpu growh, and adjused scale effecs o he oupu growh are 45 percen, 38 percen, and 17 18
percen (8.1%, 6.89%, and 3.08% ou of 18.06%), respecively. The TFP, which is he sum of he adjused scale effecs and echnical progress, conribues abou 62 percen o oal oupu growh. Anoher sign of srucure change in he manufacuring indusries is he high TFP growh in he Meal and Machinery Indusry and he High-echnology Indusry. The TFP growhs for hese wo indusrial groups are higher han hose for he Processing and Ligh Manufacuring indusrial groups (12.71% and 12.19% vs. 9.46% and 9.88%, respecively). The high TFP growh, ogeher wh high inpu growh, leads o higher oupu growhs for he Meal and Machinery Indusry and he High-echnology Indusry han he oher wo indusrial groups (19.83% and 20.56% vs. 14.63% and 16.95%, respecively). The Meal and Machinery Indusry has he highes rae of echnological progress (9.36%) and he highes growh rae of TFP (12.71%), reflecing he coninued imporance of he convenional heavy indusry. The high oupu growh for he High-echnology Indusry can be explained by high inpu growh (8.37%) and adjused scale effecs (4.35%), which are he highes in four indusrial groups. The fas growh of he emerging High-echnology Indusry is he key srucure ransformaion in China s indusrial producion. Lasly, he esimaed decomposion generaed relaively small saisical error. The absolue values of he errors beween he esimaed oupu growh and he acual oupu growh are all less han 1 percen. Esimaes for he 29 Two-dig Indusries The 161 hree-dig indusries are aggregaed ino 29 wo-dig indusries (Appendix Table A1). For example, he Processing Indusry conains 53 hree-dig indusries beween he codes 131 and 195. Among hese indusries, he indusries beween 131 and 139 are aggregaed ino one indusry wh he code of 13 ; he indusries beween 140 and 149 are aggregaed ino he indusry wh he code of 14, and so on. The wo-dig indusries wh a code from 13 o 19, 20 o 29, 30 o 37 and 40 o 43 belong o Processing Indusry, Ligh Manufacuring Indusry, Meal and Machinery Indusry and High-echnology Indusry, respecively. The esimaes of growh decomposion for each of hese 29 wo-dig indusries shown in Table 4 provide a deailed view of he manufacuring indusries. 19
Processing Ligh Manufacuring Meal and Machinery High-ech Indusry (code) Table 4 Growh Decomposion for Two-dig Indusries (%) Esimaed Y Scale TE TF P Y (1) (2) (3) (4) (5) (3)+(4)+(5) (7) (7)-(1) 13 15.87 5.97 2.30 7.61 0 9.90 15.83-0.04 14 18.44 6.83 3.33 8.27 0 11.61 17.55-0.89 15 16.52 7.12 1.87 1.88 5.66 9.40 16.57 0.05 16* 3.83 3.72 3.76-3.65 0 0.11 5.48 1.66 17 11.60 1.65 0.68 9.28 0 9.96 11.09-0.51 18 11.27 5.29 2.22 3.76 0 5.98 11.53 0.26 19 15.06 4.74 0.58 9.74 0 10.32 15.01-0.05 20 19.58 8.33 2.94 8.31 0 11.25 19.97 0.39 21 18.59 10.51 5.06 3.02 0 8.08 19.03 0.43 22 12.15 3.64 1.82 6.69 0 8.51 12.66 0.51 23 8.95 4.06 2.24 2.65 0 4.89 10.58 1.63 24 20.50 11.94 4.78 3.78 0 8.56 21.30 0.80 25* 27.69 9.42 2.13 16.13 0 18.26 28.13 0.44 26 18.23 6.73 1.43 10.06 0 11.50 19.01 0.78 27 15.62 7.25 3.47 4.90 0 8.37 15.75 0.13 28 14.83 5.04 2.44 7.34 0 9.79 13.57-1.25 29 14.33 4.89 1.66 7.77 0 9.43 14.88 0.55 30 16.30 6.68 2.02 7.60 0 9.62 16.99 0.68 31 16.58 4.93 2.29 9.36 0 11.65 16.67 0.09 32 22.31 5.87 2.21 14.23 0 16.44 22.18-0.13 33* 30.95 11.22 4.39 15.34 0 19.73 31.57 0.63 34 19.41 8.25 2.62 6.68 1.86 11.16 18.91-0.51 35 20.03 5.95 3.95 10.14 0 14.08 20.32 0.29 36 19.62 6.10 4.60 8.92 0 13.52 19.48-0.14 37 22.47 9.58 4.88 8.01 0 12.89 22.26-0.21 40 21.57 8.07 4.13 5.27 4.10 13.50 21.34-0.23 41 18.11 8.34 4.70 0.31 4.76 9.77 16.35-1.76 42 20.76 8.37 4.52 2.57 5.30 12.39 22.02 1.25 43 24.12 9.79 2.13 12.19 0 14.32 24.56 0.44 Average* 17.42 6.77 2.88 7.77 10.65 17.52 0.10 Sd 3.78 2.29 1.33 2.71 2.60 3.76 0.72 Minimum 8.95 1.65 0.58 2.65 4.89 10.58-1.76 Maximum 24.12 11.94 5.06 14.23 16.44 24.56 1.63 Noe: * Indusrial codes 16, 25, and 33 are removed from he descripive saisics calculaion shown in he las four rows. 20
Due probably o deliberae policy on healh hazards, he obacco processing indusry (code #16) has an unusual low growh of 3.83 percen. The wo unusual high oupu growh indusries are he oil indusry (code #25) and he meal smeling and alloys indusry (code #33), wh growh raes of 27.69 percen and 30.95 percen, respecively. The high oupu growh for hese wo indusries is due o echnical progress and he high demand for energy and meal producs. If we remove hese hree indusries as ouliers, he growh from he remaining 26 wodig indusries sill has a large variaion, ranging from 8.95 percen o 24.12 percen. The high variaion of oupu growh for he wo-dig indusries is accompanied by high variaion in inpu growh, scale effec, and he combined effec of he echnical progress and he change in echnical efficiency. The inpu growh for wo-dig indusrials has an average of 6.77 percen and ranges from 1.65 percen o 11.94 percen (excluding indusry codes #16, #25, and #33). Excep he four wo-dig indusries in High-echnology Indusry, he wo-dig indusries in he res hree main indusrial groups have large variaion of inpu growh. The adjused scale effec shows an average of 2.88 percen and a range from 0.58 percen o 5.06 percen. This effec has a higher variaion for he wo-dig indusries in he Processing Indusry and he Ligh Manufacuring Indusry han he Meal and Machinery Indusry and he High-echnology Indusry, all of which have an adjused scale effec of higher han 2 percen. The combined effec of he echnological progress and he change in he echnical efficiency shows an average of 7.77 percen, higher han he inpu growh, bu he range from 2.65 percen o 14.23 percen is larger han he range for inpu growh. For he performance of TFP growh, he average is 10.65 percen wh a range beween 4.89 percen and 14.23 percen. These large percenage ranges reflec he diverse performance of he individual indusries in he sample period. All bu hree wo-dig indusries enjoyed a growh rae of TFP in excess of 8 percen, wh wo close o 20 percen. The srucure change in he manufacuring indusrials can also be found in he wo-dig indusries esimaes. The inpu growh for all wo-dig indusries in he Processing Indusry are lower han 8 percen while he inpu growh for all four wo-dig indusries in he Highechnology Indusry are more han 8 percen. Mos wo-dig indusries in he Meal and Machinery Indusry and he High-echnology Indusry have he growh of TFP more han 10 percen and mos indusries in he res of he wo main indusries have he growh rae of TFP less han 10 percen. 21
Wh he excepion of five indusries (codes # 15, 34, 40, 41 and 42), here is an absence of echnical efficiency change in he majory of indusries. Three of he five indusries wh echnical efficiency change are in he High-echnology Indusry. The esimaion of growh decomposion in wo-dig indusries also has a small saisical error. There are five indusries ha have an absolue esimaion error greaer han 1 percen and all oher 24 indusries have an error less han 1 percen beween he acual oupu growh and esimaed oupu growh. Our empirical resuls for he manufacuring indusries in Tables 3 and 4 shows ha he indusrial oupu growh can be explained by inpu growh, adjused scale effec, echnical progress, and he change in echnical efficiency. For he aggregaes of all indusries, inpu growh explains 38 percen, adjused scale effec explains 17 percen, and echnical progress explains 45 percen of oupu growh, bu here is no evidence of echnical efficiency change. Conrary o he findings in Young (2000, 2003) and Li (2009) ha China s pos-reform GDP growh has depended largely on capal inpus, our resuls using more recen indusrial daa show ha echnical progress is more imporan han inpu growh for manufacuring indusries. Furhermore, increasing reurns o scale plays an imporan role for he indusrial oupu growh. Tradionally, he conribuion from he scale effec has been ignored. Wh he inclusion of he scale effec in he analysis of growh in TFP, we found ha he scale effec explains abou 27 percen of he growh of TFP (17% ou of 62%). Among he hree inpus in he producion funcion, our resuls show ha human capal conribues abou one-half of oal inpu growh and he conribuion from physical capal has surpassed he conribuion from labor. Despe he large labor force, recen sudies (Fleisher e al., 2010; Li e al., 2009) showed ha human capal has been lacking, especially in he middle-managemen range. Our resuls indicae ha manufacuring indusries do arac he formaion and growh of human capal. The empirical resuls also show srucure change from ligh indusries o heavy and echnology-inensive indusries. In general, he indusries in he Meal and Machinery Indusry and he High-echnology Indusry have a higher oupu growh han he oher wo indusry groups due o high inpu growh and high TFP growh. The high inpu growh for he Meal and Machinery Indusry / High-echnology Indusry is relaed o he growh of human capal / labor and human capal. This suggess ha he labor movemen from ligh indusries o heavy and echnology-inensive indusries is also accompanied by growh in human capal, especially in he High-echnology Indusry. Cerain wo-dig heavy indusries in he Meal and Machinery 22
Indusry have low labor growh because hese indusries require more physical capal and human capal han labor. Mos wo-dig indusries in he High-echnology Indusry conain newly formed business and indusries, which induces he high demand for he labor force and high labor growh. For all four main indusrial groups and 29 wo-dig indusries, we only find he improvemen of echnical efficiency in he High-echnical Indusry and five wo-dig indusries, hree of hem are in he High-echnical Indusry. Table 5 Growh Decomposion for Four Indusrial Groups in Four Regions Esimaed Y Scale TE TF P Y (1) (2) (3) (4) (5) (3)+(4)+(5) (7) (7)-(1) Easern Region All 16.74 6.52 2.82 7.41 0 10.23 16.81 0.07 Processing 11.88 4.45 1.70 5.72 0 7.42 12.08 0.21 Ligh Manuf. 14.99 5.95 2.36 6.69 0 9.04 15.50 0.51 Meal & Machi. 18.85 7.17 2.97 8.71 0 11.68 18.87 0.03 High-ech 21.22 8.57 4.77 1.76 6.12 12.65 20.31-0.91 Souhern Region All 20.58 8.82 4.41 3.20 4.15 11.76 21.13 0.54 Processing 17.11 7.32 4.00 5.80 0 9.80 17.37 0.26 Ligh Manuf. 19.75 9.41 4.07 2.54 3.74 10.35 20.57 0.82 Meal & Machi. 22.90 9.15 4.18 9.57 0 13.75 23.48 0.58 High-ech 20.98 9.45 5.62 5.93 0 11.55 21.35 0.37 Wesern Region All 16.53 4.88 2.43 9.22 0 11.65 16.60 0.08 Processing 15.92 4.04 1.73 10.15 0 11.88 14.65-1.26 Ligh Manuf. 15.33 4.85 2.24 8.24 0 10.48 16.10 0.77 Meal & Machi. 16.66 4.51 2.30 9.85 0 12.15 16.82 0.16 High-ech 21.40 7.52 6.11 1.69 6.09 13.88 20.80-0.60 Norheasern Region All 19.28 6.38 2.78 10.11 0 12.89 19.28 0.01 Processing 16.54 4.73 1.00 10.82 0 11.81 16.70 0.16 Ligh Manuf. 19.87 7.92 2.86 9.09 0 11.95 20.84 0.97 Meal & Machi. 21.30 6.52 3.70 8.31 2.77 14.78 20.42-0.88 High-ech 16.61 5.03 2.52 9.06 0 11.58 16.49-0.12 Noe: The daa of Guizhou and Tibe are removed from he esimaion for he Wesern region. 23
Regional Analysis The growh of indusrial oupu may vary in differen regions in China because of heir difference in hisorical background and economic developmen. Using growh decomposion, we can examine he difference in he sources of oupu growh due o regional differences. We firs apply he sochasic fronier model o he daa se from each of he four main indusrial groups in each region. And we ry boh ime-varying decay and ime-invarian echnical inefficien models o each daa se, and pick he resul wh a beer f from he wo models. Table 5 repors he esimaes of he chosen model for he four indusrial groups in he four regions. For hose daa se wh ime-invarian echnical inefficiency, he esimae of change in he echnical efficiency is zero wh TE 0. Among he four regions, he Souhern and Norheas regions have higher oupu growh (20.58% and 19.28%, respecively) han he Easern and Wesern regions (16.74% and 16.53%, respecively). The high oupu growh in he Souhern region is mainly due o high inpu growh and scale effec; he high oupu growh in he Norheasern region is due o high TFP growh. The growh of he Processing Indusry is relaively lower han he growh of he oher hree indusrial groups in he Souhern region. The inpu growhs in he oher hree indusrial groups in he Souhern region are all more han 9 percen and hey experienced he highes inpu growhs among all indusrial groups and regions. The Norheasern region is where radional heavy indusries locae. Wh he improvemen in echnical inefficiency, he Meal and Machinery Indusry in his region has he highes TFP growh among all differen indusrial groups and regions. The low oupu growh in he Wesern region is mainly caused by he low inpu growh. Bu, s high echnical growh (9.22%) is only nex o ha of he Norheasern region (10.11%). The relaively low oupu growh in he Easern region is mainly due o low growh in he Processing Indusry and Ligh Manufacuring Indusry. Boh inpu growh and TFP growh are relaively low for hese wo indusrial groups. Anoher reason of he low indusrial oupu growh in he Easern region is mainly caused by he low oupu growh in several provinces, such as Beijing, Tianjin, Shanghai, Shanxi and Hubei. Beijing, being he capal, has a lowes esimaed oupu indusrial growh. Tianjin and Shanghai are por cy-provinces heavily relied on commerce and expors. Boh Shanxi and Hubei are inland provinces and heir indusrial oupu growhs are no as fas as oher Easern provinces. There is, however, a clear sign of srucural change in he Easern region. The oupu growh of he Meal and Machine Indusry in he 24
Easern region is sill relaively low compared o he Souhern and Norheasern regions because of eher low inpu growh or low TFP growh. Alhough he Easern region does no have comparaive advanages in he Meal and Machinery Indusry when comparing wh he Souhern and Norheasern regions, s growh of he High-echnology Indusry remains high a 21.22 percen. The srong growh provinces in he Easern region are he new indusrial areas of Zhejiang, Anhui, and Shandong, which are suppored by large inpu growh. The improvemen of echnical efficiency has performed differenly among he four regions. The Ligh Manufacuring Indusry in he Souhern region has shown a posive increase. In he radional heavy indusry Norheasern region, obviously he Meal and Machinery Indusry has shown an improvemen in echnical efficiency. Boh he Easern and Wesern regions have shown a highes improvemen in he High-echnology Indusry. The difference performance in he improvemen of echnical efficiency does reflec he comparaive indusrial advanages among he four regions. V Conclusion Armed wh over hree decades of economic reform since 1978, China s indusrial expors by he urn of he 21 s cenury have capured world aenion. This aricle invesigaes ino he facors ha conribue o indusrial oupu growh in differen indusrial groups and geographical regions in China. The empirical findings do provide an X-ray on indusrial performance in China by idenifying s srenghs ha show he various poenials and weaknesses ha require addional improvemens. Having achieved a high level of cheap laborinensive manufacuring expor, China should look ino her nex sage of indusrial developmen if expors, especially in he high-end producs, were o coninue o provide overall economy vigor o economic developmen in China. By using he more recen manufacuring indusrial daa and he appropriae proxy variables on physical capal and human capal, his aricle conducs a comprehensive sudy on he growh and producivy aribues for four main indusrial groups, 29 wo-dig indusries, and four regions. Our growh decomposion mehod and regression resuls esimae he conribuions from inpus growh, scale effec, echnical progress, and echnical efficiency changes o oupu growh. We found ha labor has he larges cos share in he producion, bu s 25