CHINA S GROWTH AND PRODUCTIVITY PERFORMANCE DEBATE REVISITED * - Accounting for China s Sources of Growth with A New Data Set

Transcription

1 CHINA S GROWTH AND PRODUCTIVITY PERFORMANCE DEBATE REVISITED * - Accouning for China s Sources of Growh wih A New Daa Se Harry X. Wu Insiue of Economic Research, Hiosubashi Universiy, Tokyo harry.wu@ier.hi-u.ac.jp & The Conference Board China Cener, Beijing harry.wu@conference-board.org ABSTRACT This sudy presens a series of careful invesigaions in boh inpu and oupu daa problems in Chinese official saisics, especially hose ofen ignored or mishandled in he exising sudies, including an enormous srucural break in employmen saisics, an implausibly high pos-reform labor produciviy growh in he so-called nonmaerial services, serious inconsisencies in oupu measures beween indusrial saisics and naional accouns and beween volume movemens and changes in real values, as well as concepual problems and flaws in he official measures of fixed asse invesmen and invesmen prices. Alernaive adjusmens for hese problems are he basis for he consrucion of a new se of daa for he aggregae economy and is five major secors in I find ha he impac of exernal shocks is more pronounced using my alernaive GDP esimaes han wih official esimaes. Under he mos reasonable scenarios, my resuls show ha China s annul TFP growh is -0.5 and 1.1 percen per annum for he planning and reform period, respecively, much slower han 0.1 and 3.2 percen if using unadjused official daa. However, China s pos-reform bes TFP growh is found for a 4.1 percen per annum and poores TFP performance is found for a -0.8 percen per annum. On average, and for mos of he sub-periods examined, he resuls are very sensiive o how GDP esimaes are adjused, followed by wha invesmen deflaor is used. Bu i is less sensiive o how employmen daa are adjused and wha depreciaion rae is adoped. Keywords: Gross value added; he Gerschenkron effec, numbers employed and hours worked; physical capial sock; human capial sock; oal facor produciviy; economic developmen and reform in China JEL Classificaion: E10, E24, C82, O47 * Par of his research was conduced when I worked wih he lae Professor Angus Maddison in o updae our earlier daa work for reassessing China s long-run economic performance. Subsanial revision o he earlier daa work began wih my ineracions wih The Conference Board o improve he Chinese daa in he TCB Toal Economy Daabase. I am in deb o helpful commens and suggesions from Angus Maddison, Bar van Ark, Kyoji Fukao, Alan Heson, Shi Li, Carlo Milana, Marcel Timmer, Gaaizen J. de Vries, Ximing Yue, Zhaoyong Zhang, as well as paricipans a he 2010 AEA Meeings and seminars and workshops a RIETI (Japan), GGDC a Groningen Universiy, Universiy of Queensland, Universiy of Wesern Ausralia, Birkbeck College a Universiy of London, CEPII (Paris), and OECD Developmen Cener. I also hank Reiko Ashizawa, Vivian Chen and Xiaoqin Li for imely assisance, and Esher Shea and Nicholas Wu for reading and ediing he earlier versions of he manuscrip. I am solely responsible for remaining errors and omissions. As of Ocober 6, 2013 Page 1

2 1. INTRODUCTION The qualiy of China s pos-reform growh has been he subjec of a heaed debae in he lieraure. I draws a significan aenion whenever he China model of reform and developmen is appraised or quesioned. The cener of he debae is wheher China s growh over he reform period is significanly aribued o produciviy growh or mainly driven by facor accumulaion. Unforunaely, alhough more and more sudies have paricipaed in he debae in he pas wo decades, i has remained unseled and inconclusive. Based on he esimaes of oal facor produciviy (TFP) for he aggregae Chinese economy in he lieraure, we may approximaely caegorize he exising sudies ino wo opposie camps, opimisic versus pessimisic. 1 The opimiss may be represened by he mos recen sudies by Perkins and Rawski (2008) and Bosworh and Collins (2008) boh aribuing over 40 percen of China s pos-reform growh o TFP, ha is, 3.8 percen of annual TFP growh for he period in he former sudy and 3.6 percen for in he laer. 2 The pessimiss may be represened by Young s sudy (2003) which only esimaes China s TFP growh a 1.4 percen per year for he period Since Young only covers he nonagriculural economy, one may argue ha his esimae for he TFP growh should have been even slower if agriculure were included, ha is, a bes TFP conribued no more han 15 percen of he growh in ha period. 3 There are, however, esimaes sanding in beween which includes an esimae of 2.4 percen of annual TFP growh for he period in Wang and Yao (2002) and 2.5 percen for in Cao e al. (2009). 1 For he purpose of his sudy, we mainly cover he esimaes using he aggregae daa unless sudies using regional and secoral daa ha cover he whole economy, e.g. Cao e al. (2009) and Y. Wu (2003), excep Young (2003) who only covers he non-agriculural economy. 2 There are sudies ha obain he esimaes of annual TFP growh rae around 3 percen including he work by Ren and Sun (2005) which esimaed an annual TFP growh a 3.2 percen for , Maddison s revised esimae (2007) of 3 percen for , and an esimae of abou 3 percen by He and Kuijs (2007) (an approximae average of 3.3 for and 2.8 for ), hough he periods covered in hese sudies are less comparable. 3 Kalirajan e al. (1996) found ha TFP growh in Chinese agriculure was negaive in 16 of China s 29 provinces in afer a posiive growh in almos all provinces in Mao and Koo (1997) found ha 17 ou of China s 29 provinces experienced a decline in echnical efficiency in in agriculural producion. As of Ocober 6, 2013 Page 2

3 There are also conradicory findings for more comparable bu shorer periods. For example, for he reform period up o he mid-1990s (1978/ /95), China s annual TFP growh rae is esimaed a 3.8 and 3.9 percen in Borenszein and Osry (1996) and Hu and Khan (1997), and even as high as 4.2 percen in Fan e al. (1999). These high TFP resuls can be compared wih a much lower esimae a 1.1 percen by Woo (1998) and 1.4 percen ( ) by Y. Wu (2003). Maddison s (1998a) earlier esimae of 2.2 percen sands in beween. 4 Oher examples can be found for he nex period beween he mid-1990s o he early and mid 2000s. An opimisic esimae of he annual TFP growh rae for his period can be as high as 3.9 percen ( ) in Bosworh and Collins (2008) and 2.8 percen ( ) by He and Kuijs (2007), compared wih a very pessimisic resul of only 0.6 percen ( ) by Zheng and Hu (2005) or even a negaive value of -0.3 percen ( ) by Cao e al. (2009). Drawn on hese very differen findings, wo conflicing views abou he produciviy performance of he Chinese economy have emerged in he debae. On one side, Bosworh and Collins (2008, p.53) concluded ha heir findings had se China apar from he Eas Asian miracle of he 1970s and 1980s, which was more heavily based on invesmen in physical capial, and ha China sands ou for he sheer magniude of is gains in oal facor produciviy. By conras, Young (2003) concluded ha he produciviy performance of China s nonagriculural secor during he reform period is respecable bu no ousanding a all. This echoes Krugman (1994) s earlier commen ha if aking a longer view and benchmarking he measure on he early 1960s raher han 1978, China appears o be more like he Eas Asian economies, ha is, only modes growh in efficiency, wih mos growh driven by inpus (pp.75-76). I is difficul o sele he debae by any conjecure abou he performance of he Chinese economy. For example, China s miraculous performance should be aribued o is srucural changes ha are in line wih China s comparaive advanage (Lin, Cai and Li, 1996), refleced by is rapid growh of expor-oriened indusries in he pas wo decades. However, empirical sudies have shown ha oher Eas Asian economies 4 In his lieraure review here I concenrae on he resuls using he aggregae daa and adoping he index number approach. Bu here are some sudies ha op for he regression approach, e.g. one by Chow and Li (2002) which gives a an esimae of 3 percen for As of Ocober 6, 2013 Page 3

4 ha pursued he same expor-oriened developmen did no do well in erms of TFP (Young, 1992, 1994a, 1994b and 1995; Kim and Lau, 1994). Some argue ha China s beer performance pos-reform is benefied from a srong auhoriarian regime ha can mainain sabiliy and implemen indusrial policies wih a surdy insiuional suppor or a powerful poliical srucure hanks o Mao s cenral planning legacy (Oi, 1992; Popov, 2007). However, his reasoning is ofen aken as a negaive facor for he inefficiency of he oher Eas Asian counries. Xu (2011), on he oher hands, rejecs such a simple inerpreaion of China s auhoriarianism and aribues China s ousanding produciviy performance (based on he findings of Perkins and Rawski, 2008) o wha he calls a regionally decenralized auhoriarian (RDA) sysem, which is a combinaion of poliical cenralizaion and economic decenralizaion (Qian e al., 2006 and 2007). Under he RDA, Chinese regions compee fiercely agains each oher for beer performance rankings and regional officials careers are linked o heir performance in he ournamens (Li and Zhou, 2005). Besides, financial benefis also incenivize officials deeply involved in local business (Walder, 1995; Oi, 1992). Neverheless, while i may no be a surprise why he RDA can promoe a super fas oupu growh, i is by no means clear ha hrough wha mechanism i can find is way o drive a superb TFP growh. Insead, here is ample evidence o show he RDA induces various governmen subsidies for he sake of growh. Typically, he cenral governmen suppresses he prices of energy and oher primary inpus and he local auhoriies end o exernalize facor coss of he local seleced indusries, including he coss of land, energy, labor, capial and environmen (Huang and Tao, 2010; Geng and N Diaye, 2012). While i may no be difficul o undersand ha a he micro level underpaid coss can exaggerae profi, hence encouraging overinvesmen and inefficien use of capial, i is hard o accep ha such a negaive exernaliy may lead o a beer TFP growh. I is obvious ha hese conjecures can easily lead o differen expecaions o empirical resuls. Bringing in and relying on any of he heoreical argumens can only confuse raher han enlighen our undersanding of he mos likely problem underlying because here are likely several forces working in differen direcions. Wha we are As of Ocober 6, 2013 Page 4

5 facing is an empirical raher han a heoreical problem. 5 Since all he sudies adoped he neoclassical framework and he majoriy follows he growh accouning approach, a raher simple logic can sugges ha he core of he problem lies wih daa and measuremen. Indeed as I have wrien elsewhere he Chinese official saisics suffer from many deficiencies ranging from inconsisencies in definiion, classificaion and coverage, o he legacy of he maerial produc sysem (MPS) in naional accouning adoped in he earlier period of cenral planning from he Sovie Union, and from errors caused by inappropriae saisical approaches o daa fabricaions caused by insiuional deficiencies. 6 Regreably, despie more han wo decades of significan scholarly effors in accouning for China s sources of growh, researchers sill have o ge back o he daa fundamenals and ask wheher deficiencies in official saisics have caused significan biases and how o choose appropriae approaches o ackle hem or how o jusify he reasons for ignoring hem. 2. TAKING DATA PROBLEMS MORE SERIOUSLY While geing back o he daa fundamenals does no sound exciing, i is he only way o sele he debae. Taking daa problems more seriously does no mean ha he daa issue is he mos imporan one in his area of research, bu i is scienifically and logically an essenial issue. Growh accouning is highly daa-driven and is resuls are highly sensiive o wha daa are used and how variables are measured. Researchers in his field should have be reminded ha here were dedicaed economiss from he 1950s o 1980s whose carefully sudied daa and measuremen problems in he accouning for he sources of growh in he US economy seled inensive debaes abou he US produciviy performance (see Jorgenson, 1990, for a comprehensive review of he conribuion of he relaed sudies). 7 I is necessary o observe he imporan principles in dealing wih daa and measuremen problems in such sudies. Firs, a argeed daa problem should be fully discussed no only wih clear evidence bu also wih an undersanding of he behaviors 5 The relevance of he neoclassical orhodoxy in he case of China and many oher developing and ransiion economies is highly quesionable for is srong insiuional and behavioral assumpions (see Pack, 1993; Felipe, 1999), bu i is a differen issue here. Afer all, he debae is no abou wha heory or mehodology is more appropriae in he case of China bu which resul is closer o he realiy given he heory and mehodology. 6 See Wu (2000, 2002a, 2007 and 2011), Maddison and Wu (2008) and Wu and Yue (2012) for more discussions. 7 Also see oher aricles on his opic in he same book edied by Bernd and Triple (1990). As of Ocober 6, 2013 Page 5

6 of agens involved in he daa generaing process. Second, any assumpion ha is adoped o solve a daa problem should be compared wih alernaive scenarios and suppored by sensiiviy ess. Third, daa work for any indusry or secor of an economy should be considered for accouning ideniy and inersecoral coherence given by a se of conrol oals in a naional accouns framework. Fourh, any adjusmen ha affecs eiher level or growh rae in one ime poin mus be jusified for he flow or sock effec over a longer period. Las bu no leas, all kinds of daa work mus be made ransparen and uncondiionally available for oher researchers o repea he same exercise. Daa problems have indeed been reaed as a fundamenal issue in some growh accouning sudies on China or sudies ha ry o ackle major measuremen problems encounered in accouning for China s produciviy performance. Insead of aking official daa for graned or simply filling daa gaps by srong assumpions, hose researchers have made significan effors in idenifying daa problems in official saisics, 8 invesigaing heir naure, and proposing alernaive esimaes wih esable reasons and empirical suppors. They have also ried o make heir daa and measuremen work ransparen. Examples of such effors include sudies by Maddison (1998a, 2007), Wu (2002a, 2007, 2008a, 2011), Maddison and Wu (2008) and Keidel (1992 and 2001) on oupu level and growh rae, Woo (1998), Ren (1997) and Young (2003) on deflaors, Rawski and Mead (1998) on farm employmen, Young (2003) on human capial, Wu (2002b) on employmen, Chow (1993), Holz (2006b and 2006c), Wang and Szirmai (2012) and Wu (2008b) on invesmen flows and capial sock. While i will ake some ime for researchers o decide if hese effors o fix he daa problems can be accepable, here are sill unsolved daa problems ha are obsrucing a proper produciviy assessmen of he Chinese economy. Firs, as discussed in Maddison and Wu (2008), here is an enormous break in he official aggregae employmen daa, available wih hree broad secors, showing an implausible 17-percen jump in 1990 over This can be compared o a very plausible 1.5-percen increase in anoher employmen series wih much more deailed 8 In mos cases, official saisics in his sudy, especially he saisics of naional accouns, refer o he resuls of reconsruced and subsequenly revised GDP esimaes since he reform and paricularly since he ransiion in official saisical sysem from MPS (maerial produc sysem) o SNA (sysem of naional accouns) in 1987 marked by he 1987 SNA-MPS hybridized inpu-oupu able (DNEB and ONIOS, 1991); see SSB-Hiosubashi (1997), DNEA (1997, 2004 and 2007) and revisions in China Saisical Yearbooks by NBS (various issues). As of Ocober 6, 2013 Page 6

7 indusry breakdowns, which is repored in he same saisical yearbook. As suggesed by Yue (2005 and 2006), his problem migh be caused by a clash beween populaion census-based esimaes and annual esimaes hrough a long-esablished daa reporing sysem. This requires a careful invesigaion of he cause of he break and a proper adjusmen wih secoral foundaion raher han simply smoohing i. Second, he Chinese official saisics show ha he labor produciviy of he socalled non-maerial services (a MPS concep ha refers o services ha are no considered conribuing o maerial producion, including all non-marke services under SNA) grew a an asonishing annual rae of 5.4 percen for he enire reform period and even a 7.1 percen per annum for Such a performance has never been observed in he human hisory in normal siuaion ha shows labor produciviy in services would grow only by abou one or even less han one percen a year, because of he highly labor-inensive naure of mos services, especially non-maerial services (van Ark, 1996). Based on his observaion, Maddison proposed a zerolabor-produciviy-growh hypohesis o gauge he real oupu of hose services (Maddison, 1998a and 2007), bu i has been debaed (Maddison 2006 and Holz, 2006a). The hird problem is wih he official esimaes of he real indusrial oupu. There was an unnoiced conradicion beween he naional accouns and he indusrial accouns in he official esimaes of value added since he lae 1990s. The former covers he naional economy whereas he laer only covers enerprises a/ above he designaed size (Appendix A). When he size crierion of five million yuan of annual sales was inroduced in 1998, he value added and employmen of he abovedesignaed-size enerprises accouned for abou 60 percen of he indusrial oals. However, by 2006 he value added by he above-designaed-size enerprises was equal o he naional indusrial GDP and by 2008 i exceeded he naional indusrial GDP by 10 percen (see more deails in Appendix A). This has raised serious quesions no only abou he qualiy of he official esimaes of he indusrial value added and employmen daa, bu more imporanly, abou flaws inheried in he official saisical sysem. In fac, before his new problem wih he official indusrial saisics had surfaced, I already found ha official esimaes exaggeraed China s real indusrial growh by As of Ocober 6, 2013 Page 7

8 using a commodiy index approach (Wu, 1997 and 2002a), 9 which was adoped in Maddison (1998a and 2007) and Maddison and Wu (2008). In a more recen work (Wu, 2011), I show ha he impac of exernal shocks on he indusrial GDP growh is more pronounced using my revised commodiy index han official esimaes. I is now clear ha here is a need o revisi he earlier work no only in he wake of he new problem bu also o furher address wo problems in he commodiy index approach, ha is, he Gerschenkron effec, also known as subsiuion bias, caused by he single 1987 price weighs and he bias of a fixed 1987 value added raio (Wu and Yue, 2000; Maddison and Wu, 2008; Wu, 2002a and 2011). I is also necessary o invesigae wheher my commodiy-index approach indeed ends o underesimae qualiy improvemen (Holz, 2006a; Rawski, 2008). This sudy aemps o improve Chinese daa for he sandard growh accouning from wo dimensions. On he oupu side (he numeraor), firs, I es he sensiiviy of Maddison s value added esimaes for non-maerial services under his zero-laborproduciviy hypohesis using alernaive assumpions and incorporaing he annual movemens in he official esimaes and my new esimaes of he miliary personnel (par of he non-maerial service employmen). Second, I subsanially revise my earlier value added esimaes for he indusrial secor based on a commodiy index approach by inroducing more price weighs and ime-varian value added raios, and examine he resuls for poenial downward bias due o underesimaion of qualiy change. On he inpu side (he denominaor), based on a careful examinaion using earlier census daa, I firs work on he measure of labor inpu. This includes an adjusmen o he break in he official employmen series a broad-secor level wih alernaive scenarios, a conversion of he naural employmen numbers ino o full-ime equivalen (FTE) numbers employed based on new evidence on hours worked, and finally augmening he numbers by an educaion-based human capial index. Second, I consruc a ne capial sock series afer a careful examinaion of he available naional asse surveys and problems wih he iniial sock and invesmen deflaors and a choice of measures on capial consumpion. 9 Oher sudies ha adoped alernaive price indices (Jefferson e al., 1996; Ren, 1997; Woo, 1998; Young, 2003) and energy consumpion approximaion (Adams and Chen, 1996; Rawski, 2001) also suppor he upward bias hypohesis. As of Ocober 6, 2013 Page 8

9 The res of he paper is o be organized as follows. Secion 3 explains why official esimaes for he real oupu may be biased upwards. Secion 4 firs adjuss he serious break in he official employmen saisics, hen, makes a new esimaion for he employmen in non-maerial service, and finally convers employmen in naural numbers ino full-ime equivalen numbers based on hours worked. Secion 5 provides new esimaes for he value added of China s non-maerial services. Secion 6 consrucs an educaion-based measure for China s human capial sock. Secion 7 subsanially revises my commodiy-index based esimaes for he value added of Chinese indusry, focusing on a es of he Gerschenkron effec and he impac of exernal shocks using muliple price weighs. Secion 8 consrucs alernaive series of ne capial socks for he Chinese economy using differen deflaors and depreciaion raes. Secion 9 discusses he growh accouning resuls using his new daa se. Secion 10 summarizes he new daa effors and concludes he paper wih implicaions of he new findings and he direcion of fuure research. 3. THE UPWARD-BIAS HYPOTHESIS I has long been believed ha because of heoreical, mehodological and insiuional problems he official esimaes of China s GDP growh have been biased upwards. In his secion, I firs show ha why he Marxis heory-based MPS ends o exaggerae growh compared wih SNA. I hen, using he indusrial secor as an example, show why he comparable price sysem adoped ogeher wih MPS in growh indexing inroduces more upward bias because of he improper linking segmened indices ino one in he official pracice. Finally, I discuss why insiuional deficiencies also play a role in generaing he upward-biased growh esimaes. Why does MPS end o exaggerae he real growh? Since China s saisical pracice is sill influenced by many cenral planning legacies (Xu, 2002a) and here are sill gaps beween he adoped SNA sandards and Chinese pracices (Xu, 2009), i is necessary o undersand he key differences beween he MPS and he SNA and heir implicaions in measuring he real oupu level and growh rae in a more rigorous way. Before proceeding furher, i should be noed ha our approach is a value-added one, which consrucs oupu from he producion-side of he naional accouns. Also, for he sake of simpliciy, our discussions As of Ocober 6, 2013 Page 9

10 and mahemaical expressions below are in he real erms, leaving he price problem of official saisics in measuring real value added o Secion 7. By he MPS sandard of indusrial classificaion, here are five maerial secors in he Chinese saisics, i.e. agriculure, indusry, consrucion, ransporaion and elecommunicaion, and commerce, of which ransporaion and elecommunicaion, and commerce are he so-call maerial services. Such classificaion was common in he pracice of all he former cenrally planned economies. However, i should be noed ha he maerial service secors only include services ha are used in he maerial producion and he maerial services. Consumer services, e.g. passenger ransporaion and residenial elephone, are excluded because hey are considered unproducive in he Marxian orhodoxy. Perhaps conrary o he common percepion, he MPS does no compleely ignore he conribuion by he non-maerial services. In he calculaion of he NMP (ne maerial produc), he non-maerial services ha are used (and hence paid) by he maerial secors are kep ogeher wih he newly added value by he maerial producion, such as banking or financial services, scienific research, and legal and oher business services. The res of he non-maerial services including residenial governmen services are ignored under he MPS. 10 As shown by he formula below, he gross value of oupu of non-maerial services ns C consiss of wo componens: he gross value of he non-maerial services used (paid) by he maerial secors, ns1 C, and he gross value of he res of non-maerial services used by consumers ha are excluded from he MPS, ns2 C : (3.1) C ns ns1 ns2 C C. Now, le he value of all maerial inpus be m C, he value of depreciaion of fixed capial be m D, and ne value added (excluding depreciaion) from he maerial 10 Taking he naional accouns saisics for 1991 in nominal erms as an example (he earlies daa available wih deails of 2-digi services), if assuming 100% of he value added by scienific research services, 70% of he value added by financial services, and 20% of he value added by all oher non-maerial services are used for producers, here would be abou 60% of non-maerial services for consumers ha were ignored under he MPS (NBS, 2001, Table 3-5). As of Ocober 6, 2013 Page 10

11 producion be economy as: m V. We can define he gross maerial produc or GMP for he oal (3.2) GMP C m V m D m C ns1 Then, he sandard measure of NMP can be derived by subracing Eq. 3.2, which equals he sum of he ne value added ( V non-maerial services ( C ns1 ), ha is: m m C and m D from ) and he paymens o he (3.3) NMP GMP ( C m D m ) C ns1 V m Apparenly, neiher he GMP nor he NMP is compaible wih he SNA concep of gross value added or GVA (GDP), which includes ne value added and depreciaion of all economic aciviies, maerial or non-maerial (and marke or non-marke), as shown in Eq. 3.4 below: * m * m ns1 ns1 ns2 ns2 (3.4) GVA ( V D ) ( V D ) ( V D ). The hree componens on he righ hand side of Eq. 3.4 are: 1) he gross valued added by he maerial secors under he MPS plus he missing maerial services for m consumers, i.e. ( V * + m D * ), 2) he gross value added by he non-maerial services ns1 paid by he maerial secors under he MPS, i.e. ( V + D ns1 ), and 3) he gross value added by he res of non-maerial services ha is gone missing under he MPS, i.e. ns2 ns2 ( V + D ). Clearly, he GMP has a serious double couning problem because i includes he inermediae inpus of all he maerial secors ( C ). However, boh he GMP and he NMP ignore he conribuion by a major par of he non-maerial services ( V + D ) as well as he maerial services for consumers (= ( V D ) ( V D ) ). ns2 i m ns2 i * m * m m m i i i i Besides, he NMP is no free of he double couning problem because i includes he gross value of oupu raher han he gross value added of he non-maerial services consumed by he maerial secors (noe ha C /( V D ) 1). Finally, he NMP ignores he value of capial consumpion. ns1 ns1 ns 1 As of Ocober 6, 2013 Page 11

12 The differences beween he MPS and he SNA imply ha, firsly, he GMP (as well as he NMP bu o a less exen) ends o exaggerae he real oupu growh if he growh of inermediae inpus is faser han ha of he value added. In oher words, using our noaions, if m C grows faser han m V, holding he growh of m D consan, he GDP/GMP raio will decline over ime and, consequenly, he GMP will have a higher growh rae han he GDP. Scholarly work has shown ha his is indeed he case for a ypical cenrally planned economy (e.g. see he Sovie case by Maddison, 1998b). Wu and Yue (2000) and Wu (2011) have also shown ha he Chinese economy has experienced a declining value added raio over ime (Table 6). Secondly, if he excluded non-maerial services ends o grow a a much slower pace compared wih he res of he economy, especially manufacuring, which is a widely observed phenomenon in general (van Ark, 1996) and in he cenrally planned economy in paricular because is indusrial policy ends o sacrifice services, he real growh of oupu is ineviably o be exaggeraed (Maddison, 1998b). Prior o 1992, China s saisical auhoriies followed he Sovie MPS which involved double couning and excluded a large par of service aciviies, herefore as above discussed i underesimaed he level of he real oupu while oversaing he growh rae of he real oupu (Keidel, 1992; Rawski, 1993; World Bank, 1994; Woo, 1998; Maddison, 1998a; Wu, 1997, 2000 and 2002a). Why does he comparable price sysem exaggerae he real growh? The official growh indexing relies on a comparable price sysem, adoped ogeher wih he MPS in he early 1950s. As Maddison (2007) noed, i used segmened price weighs wih overlong inervals beween adjacen benchmark years, hence ineviably underesimaing price changes and exaggeraing he real growh. The problem can be well explained by he Gerschenkron effec (Gerschenkron, 1951), i.e., a comparison of wo siuaions in erms of oupu growh, weighed a he base-year prices, can be expeced o be biased upwards because he price movemens are inversely relaed o he quaniy movemens when he normal demand relaionship is held, which is usually he case. By he same oken, he growh esimaes prior o he base year (in an index ha is based on a year beween he beginning and he ending poins) can be expeced o be downward biased. This effec is also known as he subsiuion bias, and he longer he ime span, he sronger he bias. As of Ocober 6, 2013 Page 12

13 There have been five ses of consan prices, based on 1952, 1957, 1970, 1980 and 1990, respecively. The number of commodiies in he sample se is presumably adjused when consrucing each se of he consan prices (Xu and Gu, 1997). The 1990 consan prices was used for he period I is one of he wo ses of he consan prices in he sysem ha was used for more han a decade. The oher is he 1957 consan prices. I should also be noed ha hese consan prices are adminisraive prices excep for he 1990 consan prices ha also include marke or semi-marke prices as China jus began a dual-rack price reform. In he growh indexing, hese consan prices are linked ino one index, called comparable price index (hereafer CPPI for simpliciy and disinguished from CPI). In Wu (2011), I prove ha linking he quaniy index esimaed by segmened consan prices weighs as in he Chinese naional accouns pracice also inroduces an upward bias and can be explained by he Gerschenkron effec (Appendix B). Since he CPPI had been used up o 2002 and i sill serves as he basis in measuring price changes aferwards i is wrong or groundless o argue ha he applicaion of he comparable price sysem has lile impac on he esimaion of he real GDP growh in he pos-reform period (Holz, 2006a) FIGURE 1 COMPARISON OF CPPI, PPI AND GDP-IPI FOR CHINESE INDUSTRY (Indices 1980=100; annual growh rae in percen) Las year = = CPPI PPI GDP(I)-PI CPPI PPI GDP(I)-PI Sources: The daa for calculaing he implici indusrial GDP deflaor (GDP(I)-PI) are from China Saisical Yearbook (NBS, Tables 2-1 and 2-4, and earlier issues). The PPI daa are direcly from China Saisical Yearbook (NBS, 2012, Tables 9-11 and 9-12, and earlier issues). Noes: CPPI is consruced using GVO daa a curren prices and a differen consan prices available from China Indusrial Economy Saisical Yearbook (DITS, various issues, up o 2003). As of Ocober 6, 2013 Page 13

14 In Figure 1, I presen hree official price indices for he indusrial secor as a whole (including all manufacuring, mining and uiliy indusries) for he period They are CPPI, producer price index or PPI and he implici indusrial GDP deflaor or GDP(I)-PI. CPPI is derived from he GOV daa for he period up o 2002 and updaed o 2012 using secor-weighed PPI assuming ha he pos-2002 deflaion follows he movemen of PPI. I should also be noed ha boh CPPI and PPI refer o gross value of oupu, whereas he GDP(I)-PI refers o gross value added. From 1980 o 2012, PPI rose by 4.7 percen per year, GDP(I)-PI by 3.7 percen per year while CPPI by 2.5 percen per year. The annual flucuaions follow a similar paern bu o differen degrees. CPPI appears o be he leas volaile index while PPI is mos volaile. This follows ha given he nominal oupu, CPPI implies he mos rapid real oupu growh, whereas PPI suggess he slowes real oupu growh. Following wha has been discussed, PPI should be more reliable han CPPI and used in deflaing gross value of oupu. However, since PPI can be used as an inpu price deflaor for producer goods or a leas as a close proxy, a faser increasing PPI, compared wih GDP(I)-PI, may imply ha he rise of inpu prices mus be even higher han PPI. If his is he case, here mus be a rising value added raio (value added o gross oupu) due o efficiency improvemen or subsidies. In fac, he value added raio in Chinese indusry has been declining raher han rising (see discussion below and Table 6). I is never clear which price index and wha deflaion procedures have been used in he official esimaion of China s indusrial GDP. This cass a big puzzle and moivaes he effor of using volume movemens o bypass official price measures. Insiuional Deficiencies We now consider anoher problem ha is also relaed o China s pracice of he comparable price sysem. In order o measure price changes and real growh, enerprises are given price manuals ha provide produc-specific consan prices for he base year of he curren period. Enerprises are required o regularly repor iemized oupu a boh he consan and curren prices. However, hese manuals canno cover all iems produced and or sufficienly specify deails. This is paricularly problemaic when here are new producs appearing afer he base year. Enerprises have no guidance on how o properly price hem. Since i is very complicaed o urn new producs ino somehing equivalen lised in he manual for he base year, enerprises end o repor new producs a he curren prices raher han convering As of Ocober 6, 2013 Page 14

15 hem ino he consan prices ha do no exis. This creaes leeway for boh enerprises (sae firms in paricular) and local governmens o exaggerae he number of new producs and o overprice hem in consan erms. A differen bu also similar problem ha violaes he rule of using he assigned consan prices involves smallsized, non-sae enerprises esablished afer he base year. These enerprises end o repor he same figures a boh consan prices and curren prices for convenience or jus ou of ignorance (Rawski, 1993; Woo, 1998). Local governmens also end o close heir eyes o such pracice because of heir poliical incenives o show faser growh (Li and Zhou, 2005; Ma, 1997). Mahemaically, we can assume ha he oupu (GVO) in any given price manual period consiss of wo pars, one using he assigned consan prices for he price manual-lised producs and anoher using he curren prices or some prices ha are differen from and higher han he consan prices for he new producs. So he repored growh rae of gross value of oupu and ) can be defined as I GVO ( R) GVO GVO I GVO II GVO (3.5) g I II I II GVO GVO GVO GVO I where GVO p q and GVO priced by he consan prices c 0 I 0 II 0 p q c p c, 0 0 GVO(R) g beween wo periods (0 I 0, i.e. he firs par (I) of he oupu is II p c p ; GVO ~ p q and GVO ha is, he second par (II) of he oupu is (inappropriaely) priced by he curren prices p or somehing ha is close o he curren prices II 0 II 0 ~ p q p~ as he consan prices 0, inerpreed by enerprises for he reasons discussed; and respecive weighs for he wo pars of he oupu. I II and are he If p p 0 p and c II grows, he repored growh rae will be significanly biased upwards compared wih he acual growh rae ( g GVO ), ha is, I II I GVO ( R) GVO GVO GVO pcq GVO (3.6) g g I II I GVO GVO p q GVO 0 0. c 0 0 Beside his leeway a local and enerprise levels, here has been evidence showing ha pressures upon saisical offices o provide righ numbers ha can show As of Ocober 6, 2013 Page 15

16 expeced performance may also come from powerful auhoriies a he cenral level raher han local governmens (Wu, 2007). 11 Empirical evidence This upward-bias hypohesis has been invesigaed and esed for he oal economy by various empirical sudies using differen approaches ranging from physical oupu or commodiy indicaor (Maddison, 1998a and 2007; Maddison and Wu, 2008), 12 energy consumpion (Adams and Chen, 1996; Rawski, 2001), food consumpion (Garnau and Ma, 1993), o foreign price approximaion (Ren, 1997). Despie differen resuls, all alernaive measures appear o be srongly supporive o he hypohesis. For example, for he period , compared wih he official GDP growh rae of 9.8 percen per annum, i is esimaed as 4.8 percen by he energy approach (Adams and Chen, 1996), 6.8 o 8.5 percen by alernaive price indices (Ren, 1997; Wu, 2000), and 7.5 percen by volume movemen (Maddison, 1998). For he period , he annual growh rae is esimaed by Maddison and Wu (2008) a 7.9 percen per annum compared wih he official esimaed rae of 9.6 percen per annum RECONSTRUCTION OF CHINESE EMPLOYMENT SERIES China s official daa on employmen no only have concepual problems (see Wu, 2002b) bu also suffer from srucural breaks. In paricular, he official oal number of employmen jumped from million in 1989 o million in 1990, suggesing an asonishing 17 percen or 94.2 million increase in one year (Table F-3)! This new oal is available wih hree-secor breakdowns (primary, secondary and eriary) linking o he same breakdowns prior o 1990, bu no wih esimaes a indusry level. However, he exising indusry level esimaes, which follow he pre-1990 radiion, 11 Wu (2007) shows ha he pos-2004 census revision is made direcly o he real oupu, which is equivalen o an implici adjusmen o underlying prices (census by naure does no observe price changes). Afer replicaing he adjusmen procedures using he sandard inerpolaion approach, he shows ha he repored NBS esimaes are arbirarily modified and deliberaely lef 1998 unadjused he year when China was heavily hi by he Asian financial crisis, as clearly shown by my commodiy index (see below), bu he governmen believed ha he Chinese economy sood firmly or remained inac due o he righ leadership and policy. 12 Differen from oher sudies, hese sudies work on secor or indusry level. The commodiyindex approach is used for indusry and agriculure, whereas volume movemen approach is used for non-maerial services, leaving he official esimaes of consrucion and maerial services unadjused. 13 Based on Table 1 of Maddison and Wu (2008), all figures are convered o 1990 Geary-Khamis PPP dollars. As of Ocober 6, 2013 Page 16

17 fall shor of he new esimae of oal employmen in 1990 by 80.1 million. The pos daa series is hen buil on his new level of oal employmen, hence susaining a coninuous gap wih he underlying rend based on he pre-1990 daa series. When his indusry-level esimaion was disconinued in 2002, he gap rose o 99.6 million (NBS, 2009, Table 4-5). Two decades have passed since he gap firs emerged, ye here has been neiher explanaion nor adjusmen by he saisical auhoriy. In his secion I firs adjus he 1990 break in he employmen series by invesigaing he naure of he break and he fundamenal forces ha migh affec he demand and supply of labor a he ime of he break. I provide new esimaes wih hree scenarios. I hen make a new effor o re-esimae he missing miliary personnel and o add i back o China s non-maerial service employmen for he period prior o 1990 a facor ha played an imporan role in Maddison s value added esimaes for non-maerial service (Maddison, 1998a and 2007). Finally, I conver he esimaed naural numbers employed o full-ime-equivalen numbers employed based on he new esimaes of hours worked from wo recen sudies (Wu and Yue, 2012; Wu and Zhang, 2013). An adjusmen o he 1990 break A quick look a he 1990 srucural break agains he background of labor supply and macroeconomic siuaion gives an impression ha he break is raher arificial. On he one hand, he change of working-age populaion around ha ime was sable, i.e. wihou any significan deviaion from he rend. On he oher hand, i was impossible for he demand for labor o have a faser-han-normal increase in he middle of a serious growh slowdown by he official saisics (Table F-7) he growh of GDP dropped sharply from 10.5 percen in 1988 o 3.3 percen in 1989 and sayed a around a similar rae (3.2) in 1990, which was he slowes growh since he reform. As discussed in Yue (2005), he gap is caused by inappropriaely linking he resuls of he 1990 Populaion Census resuls o he annual esimaes ha are based on a regular employmen regisraion and reporing sysem esablished in he early planning ime. The populaion census discovered a large number employed who had been missed by he regular reporing sysem, ye he NBS did no inegrae he resuls wih he annual esimaes a he indusry level. Noneheless, wihou any good reason o ignore he census resuls, beween 1990 and 2002 he NBS coninued is census- As of Ocober 6, 2013 Page 17

18 based esimaion for oal employmen suppored by annual populaion sample surveys and published he resuls parallel o annual indusry-level esimaes in a way ha disguised he huge underlying inconsisency beween he repored oals and he (implici) sum of indusries. If his 80.1 million of addiional workforce recorded in he 1990 Census did no appear suddenly in 1990, which is a reasonable assumpion, a logical inquiry should ask wheher he gap had always exised in he economy bu never covered by he labor saisical sysem or i began from a paricular ime when changes in employmen policy allowed informal employmen o emerge bu no necessarily picked up by he regisraion sysem. A proper invesigaion should be conduced on wo grounds: checking earlier or pre-1990 populaion censuses or sample surveys o see if a similar discrepancy exised in earlier periods and examining changes in employmen policy ha creaed ouside sysem employmen. China only conduced hree populaion censuses before he 1990 Populaion Census in 1953, 1964 and Unforunaely, he available daa from he 1953 and 1964 censuses do no conain any useful employmen informaion. However, he 1982 Populaion Census repors China s oal number of employmen as million, or 68.6 millions more han he annual esimae of million for ha year. Addiional informaion from he 1987 one-percen populaion sample survey gives an esimae of million or 56.7 million more han he annual esimae of million (see Tables F3 for official saisics). I is clear now ha he srucural break sared a leas in 1982 raher han in My nex quesion is when his addiional employmen began o emerge. There has been ample sudies suggesing ha he governmen began o relax is employmen regulaion in he early 1970s o make room for he developmen of rural enerprises (hen named as commune and brigade facories ) and o allow ouside of plan hiring in ciies (Wu, 1994). However, new jobs were creaed in an informal way and many of he new workers were emporal and seasonal in naure and could be engaged in muliple jobs wihin a year, hence hey were insufficienly covered by he labor planning and reporing sysem. I hus feel jusifiable o assume ha he discrepancy began in he early 1970s, which was he ime when China emporally seled down from he chaoic siuaion a he early sage of he Culural Revoluion and reemerged inernaionally (marked by normalizing relaionship wih he US and Japan). As of Ocober 6, 2013 Page 18

19 In he following alernaive adjusmen scenarios, he wo effecs are separaely or joinly considered. Before proceeding furher, he official employmen esimaes have o be revised by aking ino accoun he resuls from he 1982 Populaion Census and he 1987 Populaion Sample Survey (one percen of he populaion). I use he oal numbers of employmen for 1982 and 1987 (he census concep-based one-percen sample survey resuls are muliplied by 100) as he conrol oals for he wo years and use he annual movemens beween he 1982, 1987 and 1990 benchmarks consruc a series of conrol oals. Consequenly, bu no surprisingly, he break is pushed back o 1982 and resuls in 19.3 percen jump in I hen propose hree scenarios for a proper adjusmen o he 1982 employmen daa break, especially for considering he secoral impac of he adjusmen, which are explained below wih he resuls, respecively. Scenario 1: This scenario assumes ha he employmen growh in 1982 follows a linear rend beween 1981 and 1983, or 2.9 percen (i.e. an average of 1981 and 1983 growh raes of 3.2 and 2.7 percen, respecively) insead of he official esimae of 19.3 percen (Table 1). This raises he level of employmen a way back from 1981 o 1949, mainaining he original official growh raes for all oher years. As a resul, he oal employmen is raised by 69.3 million o million in 1981 and by 28.7 million o million in The addiional employmen is hen allocaed ino he exising secors based on he secoral weighs. Bu, his scenario does no consider any policy change effec and assumes ha all he employmen daa prior o 1982 are underesimaed o he same exen as suggesed by he 1982 Census. Scenario 2: This scenario assumes ha he gap idenified by he 1982 census only began in he early 1970s when he governmen began relaxing planning conrols over employmen. In he adjusmen, as Scenario 1, he level of employmen in 1981 is firs raised o iron ou he break. Then, incorporaing he new rend beween 1970 and 1982 and he annual deviaions from he original rend over he same period, a new 14 The adjusmen is made for four secors, namely, agriculure, indusry, consrucion and services. However, he number of agriculural employmen in he 1982 Census (384.2 million) looks oo high, almos he same as ha of he 1990 Census (389.1). Is share in he census oal employmen is 74 percen, which is higher han ha from he regular saisical repor sysem, 68 percen. This is unreasonable in ha he census is expeced o pick up more non-agriculural employmen ha is no covered by he reporing sysem. Taking 1990 and changes beween 1980 and 1990 as references, I reduce he number of agriculural employmen by 10 percen and reallocae he difference o oher secors by heir weighs. The resuls look more plausible wih agriculural employmen accouning for 66.3 percen, indusry 18 percen, consrucion 2.2 percen, and services 13.5 percen. As of Ocober 6, 2013 Page 19

20 series of employmen is esimaed for adding more numbers employed o each year of his period, which is, for example, 69.3 million for 1981 and 4.8 million for The addiional numbers are allocaed based on he exising secoral srucure also like Scenario 1. Scenario 3: This scenario differs from Scenario 2 in he way of allocaing he addiional employmen. I assumes ha he addiional employmen is only engaged in labor-inensive, non-farming aciviies. In services, since hese addiional laborers are likely leas-educaed and largely par-ime and muli-job workers, hey are assumed o only work in maerial services. The non-maerial services are excluded in his allocaion because hey are unlikely o engage in financial, governmenal, and healhcare and educaion services. Resuls based on his scenario are repored in Table D-1. TABLE 1 CHINESE EMPLOYMENT DATA ADJUSTED FOR THE STRUCTURAL BREAK: ALTERNATIVE ESTIMATES COMPARED WITH OFFICIAL STATISTICS (Annual growh rae in percen) Working-age Populaion NBS Original NBS Revised Scenario 1 Resuls Scenario 2/3 Resuls Sources: Daa for working-age populaion and official employmen are from NBS (2010 and oher issues). See he ex for he disincion beween he wo NBS series and explanaion for he hree scenarios. Official saisics refer o end-year numbers whereas he adjused daa are mid-year esimaes. For a proper comparison wih he NBS esimaes, he adjused daa do no include miliary personnel. The compleely adjused, including miliary personnel, and official employmen esimaes are repored in Tables D-1 and D-2, respecively. Table 1 shows my alernaive adjusmens o he official employmen saisics only for he benchmarks and he adjacen years. All he esimaes refer o he aggregae economy. The annual growh rae of he working-age populaion for hese ime poins is also included o show he poenial labor supply. The NBS revised esimaes show he effec of my 1982 Census-based adjusmen ha shifs he break As of Ocober 6, 2013 Page 20

21 backwards from o As discussed, he effecs of he hree scenarios can be shown from backwards. Noe ha Scenarios 2 and 3 are he same a he aggregae level. Their annual growh raes have been significanly raised in he period due o he adjusmen. The revised esimaes in aggregaion and by broad secor based on Scenario 2 and 3 (he same a his level of secoral deails) are repored in Table F-1, which can be compared wih he original official esimaes in Table F-3. Miliary personnel as par of he non-maerial service employmen Maddison (1998a) followed a sandard pracice aemping o add miliary personnel back o China s non-maerial service employmen in consrucing he naional labor accouns. More imporanly, as Maddison argued, he exclusion of miliary personnel would significanly lower he service oupu esimaion especially for he earlier pos-war period when he miliary employmen was sizeable and engaged in many economic aciviies. 15 However, due o very limied informaion he could gaher Maddison simply assumed he size of he miliary personnel was a consan 3 million for he period (Maddison, 1998a, pp ). In his laer work, based on new informaion he assumed ha he official employmen saisics had included miliary personnel in services from 1993 onwards (Maddison 2007, p. 170). The new effor in his sudy is based on more deailed informaion hrough more careful research and daa gahering (documened in deails in Appendix Table C). This new work begins wih reconciling wo employmen series wih one caegorizing oal employmen in maerial and non-maerial and he oher under primary, secondary and eriary. The maerial and non-maerial caegorizaion for employmen is in line wih he official oupu saisics under MPS. Alhough he pracice sopped afer 1993, he available daa are enough for our invesigaion because our focus is on he pre-1990 period. This reconciliaion ensures he compaibiliy of he wo employmen series, and hence mainaining consisency in he non-maerial service employmen where miliary personnel belong o. 15 Apar from defense service, miliary personnel also engaged in consrucion, ransporaion, farming and governmen services in he early period of he People s Republic. Assuming ha hey only engaged in non-maerial (and non-marke) services may exaggerae he inpu and oupu of hese services, bu i will no affec he aggregae analysis. As of Ocober 6, 2013 Page 21