quarterly journal of economics - stanford …web.stanford.edu/~klenow/the razor's edge.pdf ·...

THE

QUARTERLY JOURNALOF ECONOMICS

Vol CXV Issue 4November 2000

THE RAZORrsquoS EDGE DISTORTIONS AND INCREMENTALREFORM IN THE PEOPLErsquoS REPUBLIC OF CHINA

lsquolsquoSages say the path [to salvation] is narrow and difficult to treadnarrow as the edge of a razorrsquorsquo

ALWYN YOUNG

In a partially reformed economy distortions beget distortions Segments of theeconomy that are freed from centralized control respond to the rent-seekingopportunities implicit in the remaining distortions of the economy The battle tocapture and then protect these rents leads to the creation of new distortions evenas the reform process tries to move forward In this paper I illustrate this idea witha study of the Peoplersquos Republic of China Under the plan prices were skewed so asto concentrate prots and hence revenue in industry As control over factorallocations was loosened local governments throughout the economy sought tocapture these rents by developing high margin industries Continued reform andgrowing interregional competition between duplicative industries threatened theprotability of these industrial structures leading local governments to impose avariety of interregional barriers to trade Thus the reform process led to thefragmentation of the domestic market and the distortion of regional productionaway from patterns of comparative advantage

This research was supported by the Canadian Institute for AdvancedResearch and a Sloan Foundation Fellowship I am enormously grateful for thedetailed comments and suggestions of Lawrence Katz and Olivier Blanchardwhich took me much further than I would have gone by myself Him Chung MaryAnn Connolly and Jean Hung provided long-suffering research assistance Asalways my greatest debt is to Ho Veng-Si who rst brought this idea to myattention in 1985 and whose comments guided this project from start to nish

From the Katha-Upanishad as translated by Mascaro [1965] A similarmessage appears in the New Testament lsquolsquoEnter ye in at the strait gate for wide isthe gate and broad is the way that leadeth to destruction and many there bewhich go in thereat Because strait is the gate and narrow is the way whichleadeth unto life and few there be that nd itrsquorsquo Matthew 713ndash14

r 2000 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics November 2000

1091

I INTRODUCTION

Analyses of the transition to market in centrally plannedeconomies have identied two potential pitfalls in the process ofincremental economic reform First the removal of government-mandated controls in one part of the economy in the context ofcontinued distortions elsewhere can for standard second-bestreasons lead to a fall in output [Murphy Shleifer and Vishny1992] Second political opposition may develop along the path ofreform stalling the reform process or reversing it altogether[Dewatripont and Roland 1995 Martinelli and Tommasi 1994] Inthis paper I use the reform experience of the Peoplersquos Republic ofChina to highlight a third potential pitfall Incremental reformreleases segments of the economy from centralized control whilemaintaining for a prolonged period many of the distortions of thecentral plan The freed segments of the economy nd it protableto exploit the rent-seeking opportunities implicit in the remainingdistortions of the economy Their attempts to capture and thenprotect these rents leads to the creation of new distortions In thissense the reform process can be derailed while some distortionsare eliminated moving the lsquolsquotrainrsquorsquo of the economy to marketothers are added drawing it off on unexpected tangents

As early as 1986 the government of Zhejiang province beganto restrict the sale of silk cocoons to other parts of China Eachharvest season the government bureaucracy from the provinciallevel all the way down to village units and including police andmilitia forces was mobilized to ensure that Zhejiangrsquos silk farm-ers sold their silk cocoons locally for local processing or directexport Under the plan the prices of raw materials such as silkwere kept articially low while the prices of nal goods such assilk fabric were kept high During the 1980s with the relaxationof central controls on rural industry locally controlled silk-processing industries developed rapidly in Zhejiang a traditionalproducer of silk cocoons to capture the rents implied by thearticial price wedge The trade embargos of the mid-1980sonward appeared as other provinces tried to purchase silk co-coons at above planned procurement prices directly from Zhe-jiang farmers bidding away the rents the Zhejiang governmentcould acquire from local processing or direct export1 In the rsthalf of 1988 silk factories in Shanghai located next door toZhejiang and Jiangsu the principal domestic producers of silk

1 See Chen [1994] preface pp 1ndash2 and text pp 7ndash8

QUARTERLY JOURNAL OF ECONOMICS1092

received only 40 tons of their planned allocation of 2000 tons ofraw silk In a country which all by itself accounted for 60 percentof world production and 90 percent of world exports of raw silkShanghai found itself in the surprising position of using valuableforeign exchange to import silk for its factories2

The lsquolsquosilkworm cocoon warrsquorsquo described above was just one ofthe many interregional trade conicts in both raw materials andnished manufactured goods which appeared in the PeoplersquosRepublic during the 1980s and 1990s While a full description ofthe factors behind this interregional conict will be given furtherbelow it is worth outlining the main features of the argument atthis stage Under the central plan raw material prices were keptlow and nal goods prices high generating substantial surplusesin manufacturing and processing industries which funded thegovernment budget As central controls over factor and materialallocations were relaxed local governments throughout the coun-try moved to develop manufacturing industries and restrict theoutward movement of raw materials in an attempt to capture therents implicit in centrally mandated price wedges Growingovercapacity in a number of industries combined with the elimi-nation of central controls over prices and the growth of interre-gional free marketing threatened industrial prot margins andled to burgeoning trade conicts as local governments tried tocontrol prices and limit lsquolsquoforeignrsquorsquo competition using a variety ofadministrative and physical barriers to interregional trade Hav-ing developed nancial and political interests in a particularindustrial and price structure local governments continued todefend those interests even after the original motivating factors(ie centrally mandated distortions) had largely disappearedThe removal of these new distortions posed an additional chal-lenge to reformers intent upon moving the economy toward freemarket institutions

This paper proceeds as follows Section II lsquolsquoStoriesrsquorsquo developsa fuller description of how incremental reforms in the PeoplersquosRepublic devolved power to local governments and led to thedevelopment of competitive local industrial policies and interre-gional barriers to trade Section III lsquolsquoDatarsquorsquo leavens the argumentwith data on the interregional dispersion of output and prices Ishow that there was widespread convergence in the structure of

2 Far Eastern Economic Review (hereinafter FEER) lsquolsquoIn a Spinrsquorsquo March 231989

THE RAZORrsquoS EDGE 1093

output during the reform period as the different provincesduplicated each otherrsquos industries Regarding prices I observe adivergence in regional prices in the late 1980s followed byuctuating rounds of convergence and divergence (without overalltrend) during the 1990s These data are supportive of the anec-dotal narrative They are however also consistent with a varietyof other explanations One might argue for example that the dataon production are explained by the opening of China to the worldeconomy which given similarities in factor endowments has ledthe different regions to nd their comparative advantage in acommon industrial structure To put this and other alternativeexplanations to rest Section III also presents data on the varianceof regional labor allocations and labor productivities showingthat these diverged substantially during the reform period Thecombination of converging outputs and diverging factor intensi-ties is grossly inconsistent with appeals to free market mecha-nisms and comparative advantage but is again compatible withthe possibility of growing interregional barriers to trade

To establish the argument that the reform process led to thecreation of new distortions one needs ultimately to providedirect evidence of increasingly irrational factor allocations To thisend Section III draws as its nal piece of evidence on data onfactor allocations and yields in agriculture I show that whilelabor intensity (labor per hectare) and yields (grain tons perhectare) were positively correlated under the plan this relationcompletely disappeared during the reform period as the provinceswith the highest agricultural yields shed labor Focusing on amore precise measure of agricultural comparative advantage thequality of the weather I show that during the reform periodprovinces with good weather increasingly applied less laborfertilizer irrigation and power resources to agricultural cultiva-tion These results are consistent with a movement towardregional autarky Section IV concludes the paper while an Appen-dix details the statistical sources used in each part of theanalysis3

3 This is not the rst paper on interregional trade conict in the PeoplersquosRepublic which has received a great deal of attention in the Chinese media andscholarly journals (see the citations in Section II below as well as the translatedarticles in Chinese Economic Studies [Fall 1993]) A World Bank study led byKumar [1994a 1994b] represents the best attempt to move beyond anecdotes andpresent statistical evidence of aggregate effects Kumar compared the degree ofindustrial dispersion in China the United States and the European Communitybreaking each political unit into twelve regions and found that Chinarsquos productionstructure was substantially more diversied It is difficult to evaluate this


II STORIES

Any review of the reform period must perforce take as itsstarting point the legacies of 30 years of central planning Threeaspects of the planning system in particular had a profoundinuence on the development of local policy during the 1980s4

First as shown in Table I revenue collection under the plan was

comparison however given differences in stages of development and the size of theeconomic units involved Kumar also used data provided by the State StatisticalBureau to examine trends in interprovincial retail purchases nding evidence of adecline in interprovincial trade However unbeknownst to the World Bankmission the data provided by the SSB (aside from containing numerous errors andmixed data types) included the value of interprovincial trade transactions withnonstate commercial departments in the value of intraprovincial purchases (ieexcluded them from the measure of interprovincial trade) This makes interpreta-tion of the declining lsquolsquotrade ratiorsquorsquo problematic In an appendix available from theauthor I provide a more complete discussion of the statistical denitions underly-ing these data on commercial transactions

Other statistical studies of this issue have been of lesser quality For examplea team from the Chinese Academy of Social Sciences [1993] presents a variety ofdata on the interprovincial trade of six provinces The tables in the paper aremutually contradictory and the article concludes at different points that for mostof these provinces the relative role of interprovincial trade was both rising andfalling (during the same time period)

4 Public administration in the Peoplersquos Republic is formally divided into velevels (1) central government (2) provinces including provincial-level cities (3)prefectures including prefectural-level cities (4) counties including county-levelcities and (5) townships [Wong Heady and Woo 1995 p 82] The levels ofsubnational government have evolved over time eg the townships were estab-lished in 1983 to replace the old communes while by the mid-1990s villagesalthough having no independent budgetary status had developed their ownsystems of revenue and expenditure [Wong 1997 pp 190ndash199] There is alsosubstantial regional variation in administrative relations eg in Jiangsu theprovincial government bypasses the prefectural level and signs revenue contractsdirectly with the counties [Wong Heady and Woo 1995 p 99] while in otherregions cities have taken over the direct administration of neighboring counties[Unger 1987 p 37] Given these difficulties in the discussion that follows I eschewany attempt to tease out subtle distinctions between levels of government InsteadI follow the standard PRC practice of using the term lsquolsquolocalrsquorsquo to encompass all levelsof government at the provincial level and below and focus on the overall process bywhich power was devolved into the hands of these subnationallevelsof government

TABLE IDISTRIBUTION OF GOVERNMENT BUDGETARY REVENUE BY SECTOR (SHARES)

Industry Agriculture Commerce Transportation Construction Other

1952 036 021 024 005 000 0131965 073 007 007 009 000 0041978 075 003 012 007 000 0031995 044 006 033 004 002 012

Industry includes mining manufacturing and utilities Revenue is classied according to the nature ofthe institution not the type of revenue eg lsquolsquoindustryrsquorsquo includes nonindustrial revenue collected by industrialministries


concentrated in the industrial sector which by the mid-1960saccounted for three quarters of the consolidated budgetary reve-nue of the central and local governments5 With the prices of rawmaterials kept articially low relative to the prices of nalindustrial goods industrial processing enterprises reaped unusu-ally high returns which were then handed over to the govern-ment either in the form of turnover (indirect) taxes or through thedirect remittance of prots Wong [1992] explains that this systemwas patterned after the Soviet model of the 1920s in which lsquolsquopricescissorsrsquorsquo generated large industrial surpluses that could then beused to fund the development of industry In the context ofgovernment control of most economic sectors this motivation ispeculiar Presumably prots could just as easily have beenconcentrated in raw material sectors such as agriculture or nalsales sectors such as commerce and then used to nance thedevelopment of industry6 It could be the case that these rents andprice wedges simply represented the combination of ideologicalbias (eg frivolous consumer goods such as watches and fansshould be expensive while valuable raw materials such as coalshould be available to all at low cost industry is important andhence industry and not agriculture or commerce should generaterevenue) and practical expediency (eg it is easier to collect therevenues of a few silk-processing centers than to tax the incomesof thousands of silk-growing collectives) For whatever reasonindustry was paramount as a source of revenue During thereform period with the rise of raw material prices and theintroduction of contractingincentive prot retention systemsthe role of industry in revenue collection declined Neverthelessthe sector still accounted for almost half of all budgetary revenuein 1995

Second as shown in Figure I historically most of governmentrevenue was nominally lsquolsquolocal revenuersquorsquo collected by local govern-ment authorities and then handed over to the central governmentIn the mid-1970s just prior to the reforms local revenue amountedto almost seven times central revenueAs the gure shows during

5 As the Chinese denition of revenue has evolved over time I adjust theofficial data to construct a consistent series This and issues such as extra-budgetary and off-budget nance which are not included in the table arediscussed in the Appendix

6 For example the large turnover taxes which were included in the factorysales price of manufactures and therefore collected from industry could just aseasily have been levied after goods were shipped to distributors and hence beencollected from the commercial sector


the 1980s the ratio of local to central revenue collection declinedrapidly as the central government increased its share of revenuecollection These gures however conceal the fact that with theexception of some centrally controlled enterprises most of so-called lsquolsquocentralrsquorsquo revenue has actually always been collected bylocal authorities simply because the central government hasnever possessed a nationwide tax bureaucracy7 Thus under theplan and even during the reform period most revenue wascollected by local governments designated as lsquolsquolocalrsquorsquo or lsquolsquocentralrsquorsquoand then transferred from local coffers to central coffers andfrequently back again The obfuscation inherent in the bilateraltransfers of this system should not blind one however to thecrucial fact that it was local governments and not the centralgovernment that had the long-standing scal ties with enter-prises throughout the economy

A third legacy of the pre-reform era was the duplication ofindustries across provinces and the active involvement of provin-cial and local authorities in economic planning Enjoying by themid-1960s hostile relations with its neighbors in almost every

7 Wong [1991a] In 1989 Chinarsquos State Tax Administration had a total staff of450 persons in Beijing and another 150 persons nationwide The remaining half amillion tax collection officials were appointed by local governments In 1994separate central tax bureaus were established with some fanfare in the variousprovinces Early foreign press reports however indicated that these new bureauspossessed lsquolsquotwo name plates but one staffrsquorsquo ie were still part of the localbureaucracy See FEER lsquolsquoFiscal Feudalismrsquorsquo April 6 1989 and lsquolsquoThe Grip SlipsrsquorsquoMay 11 1995

FIGURE IRatio of Local to Central Revenue and Expenditure


direction the Peoplersquos Republic began a conscious attempt todevelop duplicate sets of industries in each region and province soas to be better able to sustain economic activity in the event of aforeign invasion This policy extended beyond the duplication ofkey military industries to a general program of promoting localself-sufficiency in most areas of industrial production Pressed bythe complexity of directing the growing number of small localenterprises and probably dislocated by the political events of thetime the central planning apparatus during the Cultural Revolu-tion (1966ndash1976) increasingly relinquished control of detailedplanning to provincial authorities focusing instead on managingthe interprovincial transfer of key materials and products8 Withmaterial supplies only ensured when one actually produced themoneself and with the central regime actively encouraging andfunding the local development of industries each province countycity and locality tried to develop its own duplicate set of indus-tries9 It is important to note however that despite the wide-spread industrial diversication central controls over entry intohigh margin industrial sectors were maintained [Wong 1988] andconsequently industrial rents remained concentrated in a fewselect regions Thus in 1981 Shanghai with its important process-ing industries earned budgetary revenues equal to ten times totalexpenditures and while producing only 7 percent of GDP ac-counted for 45 percent of the total net local to center transfer ofbudgetary funds in the economy10

8 See Naughton [1991] and Wong [1985 1991b] The description abovesounds more coherent than the system actually was For example almost allenterprises were decentralized to local control but when their managementproved too complex for provincial authorities put under the planning of centralministries (while remaining locally lsquolsquocontrolledrsquorsquo enterprises)At the same time thestaff of the central planning agencies (material supply statistics etc) wasliquidated falling to a total of 610 employees in 1970 In the words of no less anexpert than Perkins [1988] lsquolsquoBy some mechanism inputs and outputs and theirallocationbetween enterprises were coordinated in a way that avoided the chaos ofthe Great Leap Forward and this coordination was in no sense achieved through arevived market mechanism It is not clear just who in the bureaucracy didmuch of the planning and control of enterprises but planning and control throughthe bureaucracy did take placersquorsquo

9 Wong [1985] recounts how in Huangshi city (Hubei) there were four ironand steel mills one set up by and beholden to the central government oneestablished by the provincial authorities to meet their needs one set up (near theother two mills) by the municipal government to meet its requirements and oneestablished by the suburban county which needing only 3000 tons of pig iron ayear could not get its requirements satised by allocations from other producers

10 I calculate Shanghairsquos net transfer as the difference between its revenuesand expenditures since the provinces are not supposed to engage in debt nanceand with the exception of the rollover of small surpluses and decits appear tomaintain budget balance on an annual basis Estimated from Chinarsquos PublicFinance Statistics 1950ndash1985 [pp 54 and 92]


With this background in mind I now turn to a review ofdevelopments during the reform period From almost the verybeginning the central government sought to improve the effi-ciency of industrial production and public administration byhardening the budget constraints of both state enterprises andlocal governments Beginning in 1979 the historical system of fullremission of enterprise surpluses (and central coverage of losses)was replaced with contracts specifying the division of protsbetween the government and enterprises with incentives forexceeding historical values Unfortunately each enterprise ineach locality operated under unique historical circumstancesSome enterprises had unusually large or small capital stocksothers produced goods where prots were high or low because ofdistortions in state prices whereas still others were engaged inthe difficult production of goods in unusual locations (as man-dated by the pre-reform policy of industrial diversication)Consequently contracts had to be negotiated on an enterprise-by-enterprise basis with opportunities for renegotiation and renege-ment (by both sides)11 Similarly beginning in 1980 under thepolicy of fenzao chifan (lsquolsquoeating in different kitchensrsquorsquo) the centralgovernment attempted to separate the central and local budgetsby establishing revenue contracts with the local authorities As inthe case of industrial enterprises each provincersquos circumstanceswere unique requiring a different contracting system Thus someprovinces remitted a lump-sum tax others remitted a proportionof total revenue while in still other cases revenue sources weredivided between central local shared and lsquolsquoadjustmentrsquorsquo (subjectto yet another formula for sharing) Budgetary contracts whileoften lsquolsquoestablishedrsquorsquo for ve-year periods were again much likeenterprise contracts subject to frequent renegotiation as circum-stances and policies changed

For all its imperfections the contracting system undoubtedlyhardened local and enterprise budget constraints and improvedscal and industrial efficiency by providing an objective function(albeit moving) with marginal rates of central taxation well below100 percent In the process however it inadvertently devolvedpower to local authorities and strengthened the ties between local

11 The li gai shui (tax-for-prot) system advanced in 1982ndash1985 sought tolimit the need for detailed bargaining by introducing a complex set of interlockingand compensating universal tax rates that varying by product assets etc wouldcompensate for all of the pre- and post-plan distortions However the li gai shuiinitiative met with strong opposition and by 1986 was abandoned in favor of areturn to enterprise-by-enterprise contracting


governments and state enterprises Enterprise contracting shiftedpower from the center to local governments since these with theirhistorical tax and administrative ties with enterprises and de-tailed knowledge of local circumstances were best positioned tonegotiate and monitor contracts for themselves and the centralgovernment The hardening of local budgets increased the inter-est of local authorities in industrial enterprises where the mainrevenue surpluses were to be found Local governments couldcontrol input prices and costs minimizing reported lsquolsquoprotsrsquorsquo(which might have to be shared with the center) while stillmaximizing the surpluses available for local coffers While localgovernments could draw revenue from industry their broaderinterest centered around the nancial well-being of state enter-prises as these traditionally provided housing and a wide varietyof social services (eg health retirement disability burialrecreation etc) to their workers In sum in attempting to hardenbudget constraints in the context of continued price distortionsand no attempt to clarify the residual ownership of enterprises orseparate their social from their industrial functions the centralgovernment merely transferred enterprise control and the respon-sibility for sustaining enterprise well-being down to the locallevel12

A second critical reform initiative focused on the develop-ment of rural industry as a means of absorbing surplus resources(ie labor) in the countryside Beginning in 1977 the decline incentral procurement of farm products and capital equipment andthe introduction of rural labor responsibility systems freed mate-

12 For a history of reforms in enterprise taxation and how these enlargedlocal power see Naughton [1985] and Wong [1987 1992] FEER [lsquolsquoThe Grip SlipsrsquorsquoMay 11 1995] provides a recent review of the degree to which central governmentcontrol over nance has devolved into the hands of local authorities Song [1992]and Wong [1992] review the different types of central-local contracting systems

I should note that there also appears to have been a more direct transfer ofpolitical power down to local levels China like other communist countries uses anomenklatura system in which each level of government has a list of positionsover which it exercises power of appointment (which usually includes personnel inthe next lower level of government) In 1983 as part of a plan to decentralize andimprove the efficiency of personnel management two-thirds of the positionscontrolled by the Central Committee were transferred down to lower levelsProvincial and prefectural party committees also transferred control down to lowerlevels with as an extreme example the provincial government of Heilongjiangretaining only 23 percent of its original positions [Burns 1987] It is difficulthowever to evaluate the practical signicance of these reforms as there wasalready signicant consultation with lower level authorities who had no formalnomenklatura authority (see Manion [1985 p 217]) while the length of theoriginal lists (the Central Committee exercised direct control over 13000 appoint-ments and oversight over tens of thousands more) would have led to substantialrubber stamping of lower level proposals


rial capital and labor resources for use by rural enterprisesWhile rural industry as it developed during the Cultural Revolu-tion had previously been restricted to sectors with small marginssuch as agricultural machinery repair and farm tools theserestrictions were now removed Further the central governmentactively encouraged the development of rural industry by instruct-ing the Agricultural Bank of China to provide low-interest loansrequiring that half of state budgetary allocations to communes beused for rural enterprise development and instructing centraland local authorities to incorporate the supply of rural enterprisesinto their plans13 Most importantly three-year tax holidays weregranted to particular industries (eg cement plants) and to allenterprises that might have some initial difficulty in paying taxeswhich in practice ended up granting exemption on all incomeand turnover taxes to all new enterprises With turnoverindirecttaxes as high as 66 percent in some sectors tax exemptions werecrucial allowing rural enterprises even inefficient rural enter-prises to capture the rents implicit in the margins enjoyed byrms in the urban processing centers Local governments spurredon by central initiatives aimed at hardening their budget con-straints naturally supported this shifting of rents which in theabsence of well-developed private capital markets to supportinvestment was mostly undertaken by collective organizationsWhile tax exemptions denied the central government any share ofthe captured rents informal local levies could determine theirdistribution between the collective organizations and local govern-ment coffers In 1984 contributions for local social expenditureand other levies amounted to 44 percent of the nationwideafter-tax prots of rural enterprises [Wong 1988]

The growth of rural industry as the arbitrager of articialprice wedges inevitably led to the development of interregionaltrade barriers Local governments in regions that traditionallyproduced raw materials moved downstream into processingdiverting outward-bound raw material shipments at low plannedprices to their own factories Traditional processors starved ofmaterial inputs responded by sending buyers to raw materialdistricts offering above-plan prices directly to raw materialproducers such as farmers Since their nancial interests were

13 As with all Chinese policies there were zigs and zags In 1981ndash1982 ruralentry into some sectors was restricted and some rural enterprises were forciblyshut down By 1984 however these policies had been reversed in favor of furtherbank lending and the removal of most restrictions on entry


linked to processing and not raw material production govern-ments in raw-material-producing districts moved to prevent thesesales using trade barriers to turn themselves into monopsonybuyers of raw materials at low prices This led to the developmentof lsquolsquowarsrsquorsquo over coal cotton jute silkworm cocoons and tobaccoamong other materials14 In the area of nished goods such ashigh margin light industrial consumer products the duplicationof production throughout the country led reportedly to growingovercapacity As central controls over prices were relaxed andprivate marketing channels developed rural producers foundthemselves in competition with each other and with the tradi-tional more efficient producers of nished products To protecttheir industrial interests provincial county and even city govern-ments found it expedient to erect barriers to trade so as tomaintain high local nal industrial goods prices Aside from tariffbarriers (ie special charges levied at roadblocks) nontariffmethods such as physical barriers outright prohibition low-interest loans and other nancial benets for commercial estab-lishments marketing local goods nes for commercial establish-ments marketing nonlocal goods legal restrictions on pricedifferences between local and nonlocal goods sold in commercialestablishments local purchasing quotas and administrative trivia(eg medical sanitation epidemic prevention product qualitymeasurement and other such licenses and certicates) were usedto hamper trade in products as varied as textiles automobilestrucks perfumes beverages plastics matches household electri-cal appliances electrical machinery bicycles pens alcohol laun-dry detergents and soaps tires tractors engines processed foodsand food avorings15 The legal system was also subverted as

14 In the Chinese press coverage of the battles to control resources thediversion of planned allocations to local industry the interpretation of localmotivations as an attempt to capture the rents implicit in price wedges and therole the hardening of local budget constraints played in spurring official interven-tion can be found in Beijing Domestic Service 1100 GMT May 24 1986 BeijingXinhua Domestic Service 1133 GMT December 6 1994 China Daily (BeijingBusiness Weekly) December 10 1990 p 4 Jingji Cankao (Beijing) June 4 1990 p1 April 27 1990 p 1 June 24 1990 p 1 and July 20 1990 p 1 Jingji Guanli(Beijing) No 12 (December 1989) pp 18ndash21 Jingji Ribao (Beijing) November 111989 p 2 Jingjixue Zhoubao April 30 1989 p 1 Renmin Ribao June 11 1990 p6 Zhongguo Tongxun She 0355 GMT August 11 1989 and Zhongguo Wujia No 2(February 1995) pp 13ndash17 See also FEER lsquolsquoTrouble at Millrsquorsquo March 7 1991 andlsquolsquoProtection Has a Pricersquorsquo August 29 1996 With the exception of the FEER allChinese and Hong Kong news services cited in this section are as reported andtranslated in the Foreign Broadcasting Information Service

15 For coverage of the trade barriers reference to the growing overcapacityof a number of industries brought about by excessive duplication and interpreta-tion of the interventions of local governments as stemming from their nancial


enterprises were encouraged not to pay nonlocal bills the courtsignored nonlocal pleas rulings and nes were issued againstnonlocal producers and judges who ruled in favor of nonlocalrms were punished16

While the central regime railed in a number of circularsagainst interprovincial trade wars sent down task forces tomediate conicts at provincial borders and even went so far as toestablish a separate ministry the Ministry of Internal Tradewhose main purpose was to enliven lsquolsquocirculationrsquorsquo and distributionits other policies inadvertently encouraged interregional con-ict17 Thus for example the allocation of special trading rightson a regional basis in particular to the Special Economic Zonesand Guangdong province led to conict as other regions tried toprevent the loss of export earnings brought about by the diversionof their exports to these gateways to the international market18

Perhaps the most destructive element however was the centralregimersquos policy on price controls During the Cultural Revolution

relations with industrial enterprises see China Daily (Beijing) June 9 1990 p 1Ching Chi Tao Pao (Hong Kong) No 9 (March 5 1990) pp 8ndash9 and No 37ndash38(September 28 1992) pp 12ndash15 Fazhi Ribao April 19 1993 p 1 GuangmingRibao November 3 1990 p 3 Guangxi Ribao February 3 1994 Hangzhou-Zhejiang Provincial Broadcast Service 1000 GMT December 21 1990 JingjiCankao (Beijing) June 24 1990 p 1 and June 4 1990 p 1 Jingji Ribao (Beijing)July 14 1990 p 2 Liaowang (Beijing) No 45 (November 6 1989) pp 18ndash20Qunyan (Beijing) No 2 (February 7 1991) pp 22ndash23 Renmin Ribao July 30 1990p 2 and June 5 1990 p 5 Shanxi Ribao December 1 1995 p 1 Ta Kung Pao(Hong Kong) June 22 1990 p 2 Tangtai (Hong Kong) No 41 (September 8 1990)pp 13ndash14 Wen Wei Po (Hong Kong) January 24 1991 p 6 Xinhua DomesticService 0133 GMT July 18 1984 0815 GMT May 8 1990 1031 GMT April 101990 and 0400 GMT March 29 1992 and Zhongguo Tongxun She (Hong Kong)0909 GMT February 11 1990 See also The Economist lsquolsquoChinarsquos Feuding RegionsrsquorsquoApril 20 1996 and FEER lsquolsquoOpening Doorsrsquorsquo (April 2 1992)

16 On the use of the legal system see Jingji Cankao (Beijing)April 1 1990 p4 Renmin Ribao December 10 1991 p 5 and Tangtai (Hong Kong) No 41September 8 1990 pp 13ndash14

17 Chen [1994 pp 2 33ndash36] indicates that central task forces were sentdown on an annual basis to mediate disputes at Zhejiangrsquos borders He also notesthat circulars and directives against interregional trade barriers were issued bythe State Council Party Congress Peoplersquos Congress etc in 1980 1981 19821984 1985 1987 1988 1989 and 1990 The establishment of the Ministry ofInternal Trade and its mission are highlighted in Beijing Central TelevisionProgram One Network 1100 GMT April 10 1993 China Daily July 29 1993 p 1and Renmin Ribao August 2 1993 p 2

18 While most regions were only allowed to retain about a quarter of theforeign exchange generated by their exports Shenzhen SEZ (in Guangdong) wasallowed to retain 100 percent In 1988 Hunan and Guangxi were reported to bepatrolling their border with Guangdong province restricting the movement ofgoods to their neighbor Senior officials in both provinces complained that goodsshipped to Guangdong for subsequentexport as lsquolsquoGuangdongrsquorsquo goods robbed them ofvaluable foreign exchange FEER lsquolsquoBeggar thy Neighbourrsquorsquo January 12 1989 Seealso China Daily April 4 1989 p 4 and Zhongguo Xinwen She 1344 GMT April 21992


in recognition of the fact that the policy of forced industrialdiversication led to inefficient production the central regimeallowed localities to set their own higher nal goods prices [Wong1987] This policy was maintained and apparently expandedduring the reform period With each round of price liberalizationwhile the central government reduced the number of centrallymandated price controls it simultaneously acquiesced to explic-itly allowed or in some cases even vigorously mandated the localmaintenance of price controls19 In a planned economy wherestate organs controlled the interregional movement of goods pricedifferentials could easily be maintained In a market economyhowever with atomistic private arbitragers hard at work largeinterregional differentials in the prices of traded goods could onlybe maintained with barriers to trade Thus in the area of pricesand the interregional movement of goods the central regimersquoslsquolsquoreformsrsquorsquo ultimately degenerated as they did in so many otherareas into a general devolution of power into the hands ofprovincial and subprovincial governmental authorities

By 1991 according to some estimates somewhat less than

19 The nationwide decontrol of grain prices in 1992ndash1993 provides awonderful example of policies working at cross purposes While the price of grainwas supposed to move to market levels government pronouncements at both thenational and local level stressed the importance of provinces prefectures and evenlocalities keeping adequate grain reserves so as to manage the price of grain intheir jurisdiction Similarly localities were urged to improve lsquolsquoprice inspectionrsquorsquo (forwhat purpose) and to switch from mandatory grain procurement lsquolsquoquotasrsquorsquo tovoluntary purchase lsquolsquocontractsrsquorsquo which would ensure that prices did not rise toohigh or fall too low One wonders how any of this could be accomplished withoutbarriers to trade See Jiangsu Peoplersquos Radio Network 1015 GMT March 23 1993Liaowang Overseas Edition (Chinese Beijing) No 51 (December 21 1992) p 2Xinhua Domestic Service 0734 GMT April 2 1993 and 2102 GMT April 22 1993and Zhongguo Xinwen She (in English) 0844 GMT October 3 1992

Explicit reference to the rights of local authorities (even below the provinciallevel) to set their own prices is made in Jingji Ribao (Beijing) June 8 1990 p 3Xinhua (Domestic Service) 0830 GMT February 24 1993 and 0531 GMT March 131994 Xinhua (English) 1435 GMT September 2 1989 Zhongguo Wujia No 8(August 1995) pp 16ndash18 and Zhongguo Xinwen She 0915 GMT September 41993 and 1228 GMT October 19 1994 Examples of general reports on provincialprice controls and subsidiesare given by Changsha Hunan Provincial Service 2300GMT January 24 1990 China Daily (Business Weekly) January 10 1993 p 3Harbin HeilongjiangPeoplersquos Radio Network 1000 GMT April 8 1994 HefeiAnhuiProvincial Service 1100 GMT December 30 1988 Nanfang Ribao (Guangzhou)March 19 1991 p 1 Nanjing Jiangsu Peoplersquos Radio Network 2300 GMT July 231993 Renmin Ribao August 6 1990 p 2 and October 6 1990 pp 1 and 4 ShanghaiCity Service 0900 GMT September 13 1989 Wuhan Hubei Peoplersquos RadioNetwork 1000 GMT May 12 1994 Xinhua (English) 1103 GMT December 301988 1507 GMT January 20 1990 and 0946 GMT November 30 1992 andZhejiang Ribao July 27 1994 p 1 The FEER (lsquolsquoThe Power of the Pursersquorsquo June 181987) reports that central attempts to free (and rationalize) the prices faced byindustrial enterprises were often undermined by localities and governmentdepartments which simply imposed new controls


one-third of enterprise transactions took place at centrally xedprices [Gelb Jefferson and Singh 1993] Since centrally man-dated price distortions are gradually disappearing what forceone might wonder serves to prevent local governments fromreverting to free market principles The answer to this question isas old as the history of protectionism itself local governments nownd their nancial and political interests embedded in a particu-lar industrial structure These interests can be defended using avariety of mechanisms In raw-material-producing areas exportbarriers convert processing factories into monopsony buyers andthereby maintain the price distortions of the pre-reform eraElsewhere import barriers while impoverishing the local economyas a whole can induce articially high returns in particularindustrial sectors (just as in the import substitution industries ofso many other countries) Finally when trade barriers cannot befruitfully enforced or are viewed as being too costly to other localinterests there is always the banking system which can berepeatedly pressured into extending credit to enterprises provid-ing local revenue and employment at the expense of nationalination20 Thus while it is probably the case that efficiencyconsiderations the demands of other social and political groupsand interregional competition for factors of production all placeconstraints on the distortionary activities of local governments itis also not unreasonable to argue that there is considerablehysteresis in public nance and political relations As the nextsection shows many of the forceful trends of the 1980s eg theduplication of industrial structure and the growing dispersion ofprices had by the mid-1990s come to a stop They show howeverlittle tendency to reverse themselves

III DATA

While colorful anecdotes about interregional trade wars andcomplex descriptions of the precise pattern of arbitrage opportuni-ties created by central policies make for interesting reading in thecontext of Chinarsquos economy where policy switches back and forthwhere contradictory policies coexist side-by-side and where toconfound everything almost every policy is lsquolsquoadjustedrsquorsquo to suit local

20 On the ability of regional officials to force lending from branches of thestate banking system and their successful resistance to attempts to centralizecontrol of the banking system see FEER lsquolsquoAt your Servicersquorsquo July 7 1994 andlsquolsquoVital Signsrsquorsquo March 30 1995


circumstances they are hardly compelling For every anecdote oninterregional conict one can bring out ve on growing interre-gional cooperation and of course vice versa Similarly accounts ofthe distorted objectives and motivations of local governmentseasily become when viewed through different lenses ne ex-amples of benecial local intervention helping entrepreneurialcollective enterprises use their innate skills and resources todevelop efficient and competitive industrial enterprises21 In thissection I move beyond anecdotes and stories to the analysis ofaggregate data22

A Outputs

Figure II presents the sum of the absolute and squareddeviations of the sectoral output shares of Chinarsquos differentprovinces from the group average23 The longer time series basedupon the socialist measure of lsquolsquonational incomersquorsquo which excludesnonmaterial sectors such as nance and government administra-tion divides output into ve sectors agriculture industry con-struction transport and commerce This series indicates a fairlystable degree of dispersion during the 1950s and 1960s with a

21 As strange as it may seem to readers it is not uncommon for differentobservers to hold completely opposite views on the meaning of almost any Chinesepolicy initiative and even the general thrust of government policy For exampleJefferson and Rawski [1994 p 62] describe the 1992 State Council Regulations onTransforming the Management Mechanisms of State Owned Industrial Enter-prises as part of lsquolsquoan unprecedented and virtually unrestrained push toward themarketrsquorsquo during the 1990s In contrast Li [1996] views the 1990ndash1993 period as onein which government control was reasserted seeing the Regulations in particularas part of a government attempt to lsquolsquoconsolidate their strategic control rights overenterprisesrsquorsquo

22 A complete description of the sources used for each part of the analysis isgiven in the Appendix while the data themselves are available at my websitegsbuchicagoedufacalwynyoung

23 Specically the absolute and squared deviations equal

oio

jSij 2 Sj o

io

j(Sij 2 Sj)2

respectively where Sij denotes the share of sector j in province irsquos output and Sjequals the average across i of Sij Since this cumulative measure is inuenced bythe number of sectors and provinces used in its construction one should not maketoo much of the levels of the different series The emphasis here is on trends notlevels

As explained in the Appendix Chinarsquos provincial boundaries changed duringthe reform period In general I work with the 30 provincial level administrative(and statistical) divisions in place in 1996 adjusting the data as necessary tomaintain consistent coverage In regard to the national income measure in thegure data on Tibet and Hainan do not extend back before 1979 Hence thatmeasure is based on only 28 provinces Tibet and Hainan accounted for only 08percent of provincial national income in 1992 [China Statistical Yearbook 1993Table 2ndash17]


FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)


secular downward trend appearing around 1969 ie during theCultural Revolution when the central government promotedindustrial diversication Between 1969 and 1978 the sum ofabsolute deviations panel (a) fell 18 percent Surprisingly duringthe reform period convergence continued apace with the sum ofabsolute deviations falling an additional 16 percent from 1978 to1992 The more conventional GDP series which divides outputinto primary secondary and tertiary sectors shows yet moreconvergence with the sum of absolute deviations falling 25percent by 1992 After 1992 the series attens attaining aminimum in 1993 but thereafter returning to the levels of 1992As panel (b) shows measures based upon the squared deviationsfrom the average structure indicate even greater convergencewith the sum of squared deviations falling 39 percent between1969 and 1978 (national income measure) and then a further 48percent by 1997 (GDP measure)

To establish the robustness of the preceding results Table IImodies and decomposes Figure IIrsquos analysis of the convergence

TABLE IICONVERGENCE IN THE STRUCTURE OF OUTPUT

Deviations from provincial average structure

Absolute Squared

Unweighted Weighted Unweighted Weighted

T C I T C I T C I T C I

1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038

Deviations from national structure

Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039

T 5 total C 5 coastal I 5 interior


in the structure of production I compute separately the sum ofdeviations from the provincial average structure and from thenational structure (in effect the output-weighted provincial aver-age structure) As in Figure II I compute the unweighted sum ofabsolute and squared deviations but also consider a weighting ofeach deviation by the provincersquos share of total GDP24 Finally Idecompose the sum of deviations into those of the coastal prov-inces and those of the interior25 As the reader can see theweighted measures show somewhat more convergence indicatingthat the larger provinces moved more rapidly to the averagestructure Beyond that the use of either the provincial average ornational structure as the benchmark has little effect on theresults while a breakdown of the deviations into the contributionsof the coastal and interior provinces indicates that convergencewas widespread ie the sum of deviations fell substantially forboth groups with the relative strength of the effect dependingupon the measure examined

Figure III graphs the shares of primary secondary andtertiary industry in the GDP of each province in 1978 against thecorresponding share in 1997 providing a more immediate visual

24 Thus the sum of absolute weighted deviations from the national struc-ture is given by

oio

j30 p wi Sij 2 Sj

where wi denotes the provincersquos share of total provincial GDP the 30 adjusts for thefact that the average weight is 130 and Sj 5 S iwiSij

25 The coastal provinces are Tianjin Hebei Liaoning Shanghai JiangsuZhejiang Fujian Shandong Guangdong Guangxi and Hainan Together theseprovinces accounted for 49 percent of provincial GDP in 1978 and 56 percent in1997 Jilin touches the coast on a small strip of land just north of Korea but I donot count it as a lsquolsquocoastalrsquorsquo province

FIGURE IIIShares of GDP (30 provinces)


illustration of the convergence in the composition of output duringthe reform period Panel (b) shows the enormous compression inthe variation in the share of secondary industry (mining manufac-turing construction and utilities) between 1978 and 1997 Ashighlighted by the 45 degree line provinces with high initialsecondary shares witnessed a dramatic decline in the share ofthat sector while provinces with low initial shares generallyexperienced a rise in its contribution to GDP Excluding the twooutliers in the lower left-hand corner of the gure (Hainan andTibet) the range of variation in the share of secondary industrywent from between 34 percent and 77 percent in 1978 to between37 percent and 54 percent in 1997 As regards primary industry(agriculture forestry and shing) panel (a) most provinces saw adecline in the share of that sector but the overall distribution wascompressed as by far the largest reductions occurred in thoseprovinces with high initial shares Finally as shown in panel (c)all provinces increased the share of services in total productionand due to the appearance of outliers such as Beijing andShanghai the degree of dispersion rose The disproportionategrowth of service output in the urban centers most likely reectsthe fact that both the development and measurement26 of thisnatural urban activity were neglected under socialist planning

It is important to distinguish between convergence in thecomposition of output which is predicted by a rise in interregionaltrade barriers and convergence in absolute levels of outputwhich is not Figure IV graphs the standard deviation of theprovincial GDPs per capita of primary secondary and tertiaryindustry27 As the gure shows while the variation of output percapita of secondary industry has fallen steadily this is not true forprimary and tertiary output In particular despite the conver-gence in the shares of primary industry (Figure III above) the

26 Since most services involved lsquolsquononmaterial productionrsquorsquo they were notmeasured in the socialist concepts of National Income or Total Product of SocietyThe 1991ndash1992 Census of Services led to an upward revision of about 30 percentin the estimated volume of service activity In the national and provincialstatistical series it was assumed that virtually all of this (previously unmeasured)activity had developed since 1978 (compare for example China StatisticalYearbook [1994 and 1995 Table 2ndash10] It seems likely however that some of thenewly discovered output existed prior to the reform era

27 The outputs in Figure IV as well as those used in Figures II and III and inTable II are computed in nominal prices Although I have collected data oninterregional prices in China (see the next section) I lack the heroism necessary toconvert these into regional PPP deators The variation of the ln outputs percapita or the composition of output computed in constant prices (of either 1978 or1997) exhibits trends similar to those depicted in the gures above


standard deviation of output per capita of primary productsactually increased 37 percent between 1978 and 1997 Tradebarriers tend to lower the internal prices and real outputs ofproducts in which a region has a comparative advantage Thisleads to a convergence in the nominal output shares of differentsectors The impact of this convergence in structure on thevariation of output per capita however is ambiguous and dependsupon the correlation between patterns of comparative and abso-lute advantage For example consider a poor region with anabsolute disadvantage in both agriculture and industry and acomparative advantage in the former With the imposition ofbarriers to trade the regionrsquos output of industrial products willrise while its output per capita of agricultural products alreadybelow the national average will fall This will decrease theinterregional variation of industrial output per capita whileincreasing the variation of agricultural output per capita Further-more putting aside trade barriers the evolution of patterns ofabsolute advantage during the reform period will have inuencedthe interregional dispersion of absolute output levels The imposi-tion of trade barriers has clear implications for the interregionalvariation in output shares it has no prediction regarding thevariation in the absolute output levels

FIGURE IVStandard Deviation of ln(GDPCapita) (30 provinces)


B Prices

Ceteris paribus trade barriers which segment markets willincrease the variation of prices across those markets Conse-quently in this section I examine time trends in the standarddeviation of the ln of regional goods prices

While price indices for the different regions of the Chineseeconomy are widely published and easily available data on theactual prices of goods are sparser and more problematic Usingpublic sources I have been able to collect four different data setson regional prices specically (1) annual data on the retail pricesof 305 consumer goods in 30 cities for the period 1986ndash1993 (2)annual data on the prices of 130 agricultural goods purchased bycommercial establishments in the 30 provinces during the period1986ndash1993 (3) monthly data on the market prices of 49 industrialmaterials in 36 cities for the period 390 to 599 and (4) monthlydata on the market prices of 33 agricultural products in 36 citiesfor the period 693 to 599 While the gures listed above outlinethe full range of products and locations covered by the data it isimportant to note that there are numerous gaps Products enterand leave the published series while numerous citiesprovincesreport no data and the locales omitted in this fashion vary fromproduct to product and period to period28 The annual data formthe basis of the decades old retail price and farm productspurchasing price indices While the reported locales are fairlystable the number of products present varies considerably Over-whelmed by the number of new products appearing in theeconomy the State Statistical Bureau completely abandoned thecollection of these data in 199429 The monthly price data repre-sent a new effort begun at the turn of the decade by the ChinaPrice Information Centre a division of the Planning CommitteeThere is little change over time in the reported products but thereis great (although seemingly random) month-to-month variationin the reporting cities Table III summarizes the characteristics ofthe different series

I take the simplest approach to the analysis of these unbal-anced panels regressing the ln of the standard deviation of the ln

28 All of which provides some insight into what lies behind the ubiquitousprice indices

29 However the various price indices continued to be publishedAccording toofficials at the SSB (personal communication) each locale was instructed tocompile the index in the manner it saw t and report to the SSB which would thencombine the local price indices to form the published series


prices of each product in each time period on a complete set ofproduct and time dummies using the repeated observations of theproduct standard deviations to infer trends in the overall disper-sion of prices30 In Figure V I report the value of the timedummies relative to a base period of zero Panels (a) and (b)examine the dispersion of the annual data using 1986 as the baseperiod As the gure shows the dispersion of both retail andagricultural prices rose rapidly between 1986 and 1989 afterwhich it fell and then rose again Overall there was a highlysignicant positive time trend in retail price dispersion of about24 percent per annum (t 5 80) and a somewhat weaker albeitstill signicant (t 5 21) positive trend of 14 percent per annumin the dispersion of agricultural prices Panels (c) and (d) extendthe analysis to the monthly data taking as the base period 693the initial period of the agricultural market price data andapproximately the endpoint of the annual estimates in panels (a)

30 In essence I ignore the variation in the locales used to compute eachstandard deviation under the assumption that they vary randomly I haveexperimented with other approaches For example I construct a chained balancedpanel by taking paired observations on locales in adjacent periods Taking eachgroup of pairs (representing observations on the prices of a product in commonlocales in two adjacent periods) as a lsquolsquoproductrsquorsquo I then regress the ln standarddeviations on time dummies and a full set of lsquolsquoproductrsquorsquo dummies The movementsof the time dummies are similar although given the chain the standard errorsexplode as one moves away from the base period In another approach I regressthe ln of the squared residuals from a regression of ln prices on time 3 productdummies (ie deviations from time 3 product means) on sets of product time andlocale dummies to search for time-related patterns of heteroskedasticity (adjustedfor locale-specic heteroskedasticity) This approach is not without its problems(see Greene [1993 p 400]) as the time 3 product means depend upon the localespresent in each period In any case the results are once again similar

TABLE IIISUMMARY OF PRICE DATA

Retailprices

Agriculturalpurchase

prices

Industrialmaterials

prices

Agriculturalmarketprices

Geographic unit city province city cityFrequency annual annual monthly monthly of products 305 130 49 33 of time periods 8 8 110 71 prod 3 time combinations 1774 690 3853 1971 of price reports 48717 10361 79824 46135

Product 3 time period combinations and number of cityprovincial price observations used in theanalysis (products with less than ve price observations in a time period are eliminated from the sample)


and (b) As the reader can see the monthly price data have gonethrough bouts of falling and rising dispersion These uctuationsmirror the patterns in the overlapping period of the annual datawith a reduction in dispersion in 1990ndash1991 followed by risingdispersion in 1991ndash1993 but the movements are considerablygrosser with the mean monthly industrial materials price disper-sion rising 23 percent from 1991 to 1993 These bouts of fallingand rising dispersion are consistent with trade wars that are

FIGURE VTrends in the Standard Deviation of ln Product Prices

(time dummies plus or minus two standard errors)


periodically interrupted and suppressed by the central govern-ment only to resume once again when central attention focuseson other matters31 In any case regardless of the interpretationthere is no indication of any trend toward reduced price dispersionin the 1990s32

To summarize the quality of Chinese price data is extremelypoor The data such as they are indicate a rise in dispersionduring the late 1980s During the 1990s price dispersion haswaxed and waned but shows no clear tendency to reverse theincreases of the previous decade

C Alternatives and a Variance Decomposition

While evidence of a convergence in the structure of produc-tion and some divergence of prices is consistent with an increasein interregional trade barriers it is also consistent with a varietyof other hypotheses One might argue that the factor endowmentsand relative productivities of Chinarsquos diverse provinces appearquite dissimilar when viewed in isolation but surprisingly homo-geneous when contrasted with the rest of the world If so thenwhen China was a closed economy its provinces would havespecialized in different sectors but once China opened to the

31 These uctuations in price dispersion do not appear to be driven byvariation in the locales present in the data As noted in the footnote above when Iconstruct a chained balanced sample or estimate time series heteroskedasticitywith controls for locale-specic heteroskedasticity I get similar results If any-thing the magnitude of the uctuations is substantially greater For example inthe chained balanced sample the mean monthly standard deviation of industrialmaterials prices rises 72 percent () between 1991 and 1993

Alternatively one might worry that the uctuations are driven by changes inthe overall rate of price ination To this end I have used balanced sample localepairs to construct period-by-period estimates of the rate of ination in eachproduct and found that product ination during the period has either aninsignicant or a signicantly negative effect on price dispersion In any case theestimated pattern of uctuations is once again largely unaffected

Finally I should also note that it is unlikely that measurement error explainsthese uctuations Measurementerror is included in the estimates of the standarddeviation of ln prices For it to generate the pattern in the gure it would benecessary that the quantity of measurementerror cycled up and down in a seriallycorrelated fashion The only information I have on this is that according to officialsin the Price Information Centre with the introduction of computer coding in thepast two years they have made attempts to weed out errors eg asking forconrmation of numbers that deviate excessively from gures reported in earliermonthsYet during 1998 and 1999 the estimated standard deviations rose rapidlyas part of a new upswing

32 Actually the industrial materials price standard deviations have asignicant positive trend (09 percent per month) while the agricultural marketprice standard deviations have a signicant negative trend ( 2 2 percent permonth) but anyone glancing at the sinusoidal pattern of the monthly dummies inFigure V can see that these coefficients might easily be eliminated with theaddition of a few more months of data


international market each province would naturally nd itscomparative advantage in a common industrial structure Alterna-tively it is possible that interregional patterns of comparativeadvantage converged during the reform period and that thiswithout any appeal to internal trade barriers or the opening to theinternational market explains the convergence in the composi-tion of output Regarding the divergence of prices during the late1980s if the reform period has seen a growth of product varietywith different provinces specializing in different varieties thenabsent any change in interregional price arbitrage this wouldnaturally lead to growing price dispersion within product catego-ries Further with increased regional specialization within prod-uct categories temporary economic shocks might have uneveneffects across localities generating the bouts of converging anddiverging prices observed during the 1990s

A simple variance decomposition answers some albeit not allof the counterarguments listed above In Table IV I decompose theinterregional variance of the relative output of primary to second-

TABLE IVVARIANCE DECOMPOSITIONmdashPRIMARY VERSUS SECONDARY INDUSTRY

Variance of Covariance of

lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS

China (28 provinces)

1978 071 012 056 0011997 053 015 073 2 017

United States (50 states amp DC)

1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015

All but Tianjin and ZhejiangInner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangxi Henan Hunan Guangdong Sichuan

Guizhou Yunnan Qinghai and Ningxia


ary industry two predominantly tradable sectors33 into thevariance of labor productivities the variance of labor allocationsand the covariance between the two As the reader can see duringthe reform period the interregional variance across 28 of Chinarsquosprovinces of the ln relative output of primary to secondaryindustry went down Surprisingly during the same period thevariance of provincial relative labor productivities and laborallocations both rose In other words while Chinarsquos provincesbecame more similar in terms of the composition of nal outputthey became increasingly dissimilar in terms of sectoral produc-tivities and sectoral labor allocations with the decline in thevariance of outputs only appearing as a consequence of thedevelopment of a strong negative covariance between productivi-ties and factor allocations If the convergence in the composition ofoutput across Chinarsquos provinces is driven by a convergence inpatterns of comparative advantage then one should observe agrowing similarity of labor allocations and productivities (ie theeconomies should be increasingly similar on all dimensions)Similarly common specialization driven by the opening to theinternational market should be accompanied by a growing similar-ity of labor allocations None of this is present in the data

For contrast Table IV also presents a variance decompositionof the regional output of the United States and a longer timeseries for a subset of Chinese provinces Between 1977 and 1994the variance of the relative output of primary to secondaryindustry across U S states fell rapidly Unlike China however inthe United States this convergence was associated with a largedecline in the variance of labor allocations Examining a longertime series for fourteen provinces of the Peoplersquos Republic34

between 1952 and 1965 we observe rapid convergence in thestructure of production driven by a substantial decline in thevariance of labor productivities and labor allocations Duringthe period 1965ndash1978 when central policy and the chaos of theCultural Revolution encouraged regional industrial duplicationthe convergenceof outputs continued but the mechanism changed

33 In Chinese data primary industry includes agriculture forestry andshing while secondary industry includes mining manufacturing constructionand utilities In producing the calculations for the United States (using the BEAdata described in the Appendix) I construct sectoral aggregates using the Chinesedenitions

34 Together these provinces accounted for 52 percent of provincial GDP in1978


as the decline in the variance of relative outputs came from adeterioration of the covariance between productivities and alloca-tions This transformation of the mechanism behind productiveconvergence reached fruition during the reform period (1978ndash1997) when in this sample as in the aggregate economy thevariance of productivities and allocations grew while their covari-ance continued to deteriorate

In sum the convergence in the structure of productionobserved in U S states in recent decades and in Chinese prov-inces in the early postwar era is quite consistent with explana-tions that emphasize either increasing integration into the inter-national market or a convergence in patterns of comparativeadvantage However the continued convergence of the Chineseeconomy during rst the Cultural Revolution and then thereform period is not consistent with such explanations The laterChinese data however are compatible with a rise in tradebarriers A growing diversity of nominal labor productivities couldeasily arise from growing interregional price disparities whichwould increase both the variation of relative prices and due to adecline in factor price equalization the variation of relative reallabor productivities A rise in the variance of labor allocationscould come from increasingly perverse labor allocations as prov-inces poured resources into areas of comparative disadvantage aview that would also not be incompatible with the growingnegative correlationbetween productivities and factor allocations35

35 As rst noted by Ford [1967] and then Falvey [1981] trade barriers tendto raise the relative productivity of labor in industries in which an economy has acomparative advantage This result holds whether the source of comparativeadvantage is aggregate relative factor supplies (capital and labor) the supply ofsector-specic capital (with labor mobile between sectors) or sector-specic totalfactor productivity Further while the impositionof trade barriers should startingfrom free trade lead to a diminishing variation of factor (labor) allocations oncefactor allocations become perverse (ie if regions concentrate factors withinsectors of comparative disadvantage) further impediments to trade will lead to agrowing interregional variation of factor allocations Consequently one couldinterpret the rising variances and deteriorating covariances in Table IV asevidence of increasing trade barriers

The preceding argument however is not quite correct First one can showthat in a two-by-two Heckscher-Ohlin framework in which say all economies havethe same capital-labor ratios but have different relative sectoral total factorproductivities if the elasticity of substitution between capital and labor is greaterthan one then under free trade labor productivity will actually be lower in theindustry in which the economy has a comparative advantage Since trade barrierstend to raise relative labor productivity in the industry in which the economy has acomparative advantage (the Ford-Falvey result) it follows that starting from apoint of free trade a rise in trade barriers will at least initially lead to a decreasein the interregional variation of relative labor productivities

Second and perhaps more importantly the Ford-Falvey analysis is basedupon real productivities As highlighted by the price terms in the column headings


At this point a critic might observe that the precedingdiscussion including the counterarguments that motivated thissection is all based upon the analysis of a free market economymoving from one state to another as a consequence of a change inan underlying lsquolsquoexogenousrsquorsquo variable ie patterns of comparativeadvantage openness to the world or interprovincial trade barri-ers However the Chinese economy in 1978 was a distortedcentrally planned construct As such there is no sensible way touse the customary comparative statics of trade theory to under-stand its evolution since then whether that evolution has beentoward a free market (the conventional view) or distortionarylocal planning (the view of this paper) In other words anadamant skeptic might simply argue that all of the preceding dataon outputs prices labor productivities and factor allocations isfor an appropriately complex set of initial centrally planneddistortions perfectly compatible with a movement to the freemarket

In this paper I have shown that during the reform periodinterregional outputs converged and interregional prices diverged(or at least uctuated wildly) These data are compatible with arise in interregional trade barriers They are also somewhat atodds with a movement to market given the conventional viewthat historically Chinese central planning led to excessive inter-regional duplication of industries To establish the argumenthowever I need to present evidence of a movement of factorallocations away from patterns of comparative advantage ie ofan increasingly distorted interregional allocation of factors Ide-ally such evidence should be uncontaminated by potentially

of Table IV my data concern nominal productivities ie real productivitiesmultiplied by prices Since prices tend to be lower in products in which a region hasa comparative advantage in nominal terms the Ford-Falvey analysis can bereversed ie in the presence of barriers to trade a region might actually have acomparative advantage in products with lower relative nominal productivitiesConsequently all one can say is that the data in the table are compatible with arise in trade barriers ie they are one of many possible congurations (they arehowever regardless of elasticitiesof substitution or prices quite incompatiblewitha convergence in patterns of comparative advantage or the movement of factors inresponse to an opening to the world market)

For what it is worth the analysis in constant prices of either 1978 or 1997looks much the same However this is not a perfect solution as there is still theproblem of the base year prices and their correlation with patterns of comparativeadvantage In any case as I discuss in the text it is somewhat fanciful to use thetraditional constructs of trade theory to understand the transition from distortedcentral planning Consequently rather than pursue the Ford-Falvey emphasis onrelative labor productivities in the next section I develop direct measures ofcomparative advantage showing that factor allocations are indeed becomingincreasingly perverse


distorted price measures Further the measure of comparativeadvantage and the interpretation of the results as a movementtoward regional autarky should be invariant with respect to theeconomic regime be it central planning local planning or freemarket To this end in the next section I turn to the analysis ofagricultural yields factor allocations and the weather

D Agriculture and the Weather

Panel (a) in Figure VI plots the average yield of land sownwith grain against the overall labor intensity of agriculturalproduction in 27 Chinese provinces in 197836 The typical econo-mist would not nd the positive association depicted in the guresurprising The greater the quantity of inputs applied per hectareof agricultural land the greater should be the yield per unit ofthat land With grain the principal agricultural product account-ing for the lionrsquos share of hectares sown it is not surprising thatthe yield of land sown with grain rises with the overall laborintensity of agricultural activity in a province37 Unfortunately by1997 as shown in panel (b) this lsquolsquonaturalrsquorsquo positive association hadcompletely disappeared This development forces one to reexam-ine the logic of panel (a)

A positive association between labor per hectare and yield perhectare might reect causality running from the x axis to the yaxis but it can also reect causality running from the y axis to thex axis If one thinks of all provinces as having the same agricul-tural lsquolsquotechnologyrsquorsquo then the causality runs exclusively from x to yas greater labor inputs per hectare result in greater yields per

36 Following the data I will use in Table V I use the 29 provincial-leveladministrative divisions in place in 1985 including Hainan under Guangdong andChongqing under Sichuan With data on primary employment in Tianjin andZhejiang in 1978 unavailable this leaves the 27 provinces depicted in the gure

37 In 1978 grain accounted for 80 percent of total hectares sown In 1997 theshare of grain was 73 percent Within grain the shares of rice wheat corn andtubers in total hectares sown was 29 percent 24 percent 17 percent and 10percent respectively in 1978 and 28 percent 27 percent 21 percent and 9percent respectively in 1997 It is hard to believe that these small changes in theoverall role of grain and its composition account for the results which follow (Alldata are from the China Statistical Yearbook)

More problematic perhaps is my use of the labor in primary industry (not Ishould emphasize the rural labor force which includes labor engaged in nonpri-mary activities) as a measure of the labor devoted to the cultivation of land It ispossible that an increasing share of primary sector labor is devoted to nonfarmingactivities eg forestry animal husbandry and shing I should note howeverthat in the analysis in Table V I use direct measures of inputs that are moreexclusive to farming eg irrigation and chemical fertilizers The evolvingrelationship between these factors and measures of comparative advantage whichwill be my main emphasis is the same as that of primary labor


hectare However there are likely to be substantial differencesacross provinces with some locales having land and weatherconditions more suitable to the cultivation of grain and agricul-tural products in general These provinces endowed with asuperior lsquolsquotechnologyrsquorsquo will have more productive land raising themarginal product of labor on that land and drawing labor into the

FIGURE VIFactor Intensity and Productivity (27 provinces)


sector In this case the causality will run both from x to y and fromy to x as greater yields lead to a greater labor intensity ofproduction

To explain the orthogonality between factor intensities andyields apparent in panel (b) one merely has to pervert thecausality running from yields to factor intensities by say introduc-ing interregional barriers to trade With barriers to trade thetendency for factors to ow into more productive sectors can beweakened and possibly reversed In a free market economyprovinces with better agricultural yields would experience adecline in the price of grain which would reverse the ow offactors of production into that sector In China with the enduringnational and local controls over the price of grain price mecha-nisms such as this are unlikely to be at play Nevertheless thequantity restrictions implied by price controls and local procure-ment plans will generate the same effect as in provinces withproductive land farm labor faced with limited internal demandmigrates into industrial activity Evidence in favor of a reversal ofthe link between yields and factor intensities is given by panel (c)of Figure VI which shows that the provinces with the mostproductive agricultural sectors in 1978 were the ones thatrelative to the mean change experienced the greatest declines inthe labor intensity of agricultural production With a sufficientperversion of the causality running from yields to factor intensi-ties ( y to x) one can overcome the natural relation between factorintensities and yields (x to y) producing the orthogonal cloud ofpanel (b) in Figure VI

To support the preceding argument about perverse alloca-tions one would like to have a measure of the suitability ofdifferent regions for agricultural production and show that factorallocations have moved against that measure Ideally the mea-sure should be time invariant uninuenced by all of the remark-able changes in regional comparative advantage that most surelyhave accompanied twenty years of reform and growth As such ameasure I choose the weather In Table V I regress a panel ofprovincial grain yields during the period 1985 to 1997 on variousmeasures of agricultural factor inputs and rainfall with provin-cial dummies to control for mean effects on yields Turning rst tothe weather I nd that grain yields rise with moderate (but notexcessive) quantities of rainfall and are negatively related to themonthly variation in that rainfall This seems completely intui-


tive Turning to the measures of other inputs I nd that grainyields rise with the share of agricultural land which is irrigatedand with the application of fertilizer However the total power ofagricultural machinery and quantity of labor per hectare of

TABLE VPRODUCTIVITY FACTOR ALLOCATIONS AND THE WEATHER

(ANNUAL OBSERVATIONS 29 PROVINCES 1985 TO 1997)

(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield

With provincial dummies

Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944

Without provincial dummies

Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

Yield 5 ln(grain outputgrain sown area) Irri 5 ln(irrigated areatotal sown area) Fert 5 ln(chemicalfertilizers total sown area) Power 5 ln(total power of agricultural machinery (kw)total sown area) Labor 5ln(employment in primary industrytotal sown area) Prec 5 average monthly rainfall (cm) Prec2 5 Precsquared CVPrec 5 coefficient of variation of monthly rainfall Lshare 5 primary employmenttotalemployment OutputC 5 ln(grain output per capita) Weather 5 mean value of weather variables multipliedby coefficients in column (1) Wtime 5 Weather p (year 2 1985) Signicant at the 10 percent levelSignicant at the 1 percent level T-statistics are in parentheses The sample size increases slightly betweencolumn (1) and columns (2)ndash(8) as using the mean value of weather I do not need observations on weathervariables in each period


agricultural land have no signicant impact on yields Theseresults mirror panel (b) of Figure VI38 I interpret them asreecting the negative response of these factor allocations toomitted variables which from year to year favor yields in oneregion or another

Using the coefficients from my regression of yields on rainfallI derive a synthetic measure of provincial weather-dependentcomparative advantage in agriculture (lsquolsquoweatherrsquorsquo) as the meanannual value of the combined effects of precipitation precipitationsquared and the variation of precipitation39 In columns (2)through (7) I then regress various measures of factor inputs onthis measure interacted with time (lsquolsquoWtimersquorsquo) With provincialdummies included in the regression (as in the upper panel of thetable) the coefficient on Wtime indicates the association betweenthe provincial trends of the dependent variable and the favorable-ness of the weather40 As the table indicates over time provinceswith better weather withdrew factors of production from agricul-ture The coefficients on the application of irrigation fertilizerpower (machinery) and labor are all highly signicantly negativeas is the coefficient on the share of the labor force in agricultureThe output per capita of grain products also moved againstpatterns of comparative advantage (column (7)) The magnitude ofthe coefficients indicate that a standard deviation increase in thevalue of weather (026) was associated with about 1 percent perannum slower growth in the application of fertilizer and in theoutput of grains per capita and about 12 percent per annumslower growth in the application of irrigation power and laborExcluding Wtime from the regression one nds that the meanprovincial trend growth of irrigation fertilizer power labor andoutput per capita was 10 70 42 00 and 10 percent per annum

38 I should emphasize that because of the absence of weather data forearlier years the analysis in the table begins in 1985 By that time the relationbetween labor intensity and yields had already begun to deteriorate

39 Strictly speaking this is a measure of absolute advantage However ifabsolute advantage in the nonagricultural segments of the economy is orthogonalto the weather as it plausibly might be then subject to random error lsquolsquoweatherrsquorsquo isa measure of comparative advantage as well

40 To see this write the regression as

(1) Yit 5 Btt 1 BwWi t 1 a i 1 e it

where Y represents the dependent variable W is my measure of comparativeadvantage the a rsquos are the provincial dummies i denotes the province and t(running from 0 to T ) time Partialing out the provincial dummies one nds that

(2) Yit 2 Yi 5 Bt(t 2 T2) 1 BwWi(t 2 T2) 1 e it


respectively With regard to the share of the labor force inagriculture the only variable not measured in lns a standarddeviation increase in the favorableness of the weather wasassociated with a 2 percent faster decline in the labor share or a24 percent fall over the entire sample period During the sameperiod the mean provincial share of labor in agriculture fell from59 to 49 In sum the effects are both statistically and empiricallysignicant

As a test of robustness in the lower panel of Table V I run thesame regressions without provincial dummies including both themean value of the weather and its interaction with time Withoutprovincial dummies the regressions in columns (2) through (7) nolonger examine the provincial trends in the dependent variablebut instead indicate the trend in the cross-sectional associationbetween the level of the dependent variable and the quality ofprovincial weather On the face of it the results now appearconsiderably weaker as most of the coefficients on Wtime areinsignicant However as the reader can see the coefficients onWtime are actually identical to those in the upper panel In abalanced panel the orthogonalization of an observational unitmean interacted with time ie a variable such as Wtime withrespect to either unit dummies or the observational unit meanitself yields the exact same residuals as independent regressors41

Consequently the only reason why the coefficients on Wtime areinsignicant in the lower panel is that without provincial dum-mies the t of the regression worsens considerably leading to theconclusion that there is considerably more lsquolsquoerrorrsquorsquo variance in thedependent variable and consequently much less condence inany parameter estimate It seems reasonable to recognize that themean provincial value of the dependent variables in columns(2)ndash(7) might vary for reasons other than the weather and thatprovincial dummies are an acceptable means of controlling forsuch effects Nevertheless I include the results in the lower panelto show the reader how the results I have emphasized can beweakened

41 In other words after partialing out the effect of Wi the right-hand sidevariables are the same as in the preceding footnote The left-hand side variable ofcourse is different but the additional variation is only across i categories (the tdimension variation within each i category is unchanged) As the only variation ofthe orthogonalized right-hand side variables is across t (their mean within each icategory is zero) the coefficients are unchanged I should note that the regressionsin column (4) are unbalanced (I am missing an observation on power for oneprovince) and hence the coefficients differ slightly between the upper and lowerpanels of the table


Column (8) of Table V reruns the regression of column (1)with Wtime instead of the annual weather variables As thiscolumn shows relative to average trends provinces with goodweather experienced declining yields This result is not particu-larly surprising As columns (2)ndash(6) showed the trend growth ofall of the factor inputs I have measured has varied negatively withthe quality of the weather It seems likely that this negativecorrelation extends to other unmeasured complementary inputsIf so the estimated impact of the weather on yields woulddiminish over time Column (8) however does highlight theweaknesses of the methodology I have pursued Omitted factorinputs bias the coefficients on rainfall while omitted measures ofcomparative advantage bias the coefficients on other factor inputs(eg power and labor) thereby further distorting the coefficientson rainfall Consequently my synthetic measure of the lsquolsquoweatherrsquorsquois subject to considerable error and bias As a sensitivity test Ihave reestimated column (1) using the data for 1985 alone orusing the data for 1985ndash1989 and the results (for the whole table)are much the same I have no means with the data at my disposalto even begin to address problems of endogeneity and bias but theresults do not seem sensitive to reasonable respecications

As a nal defense the unrepentent skeptic might reinterpretthe preceding results as merely showing that Chinese agriculturalproduction during the reform period has become less dependentupon the weather the simple manifestation of technologicalprogress that has undermined historical sources of comparativeadvantage in agriculture Denied the possibility that even theweather has remained an enduring source of comparative advan-tage in agriculture I am forced to return to the simple correla-tions that motivated the analysis of Table V Figure VII graphs therelationship between grain yields and the labor intensity ofagricultural production in fteen Chinese provinces across theentire postwar era42 As the gure shows from 1952 to 1978 therewas consistently a signicant positive correlation between thesetwo variables By 1997 that correlation was gone For almost 30years with all of the excesses and inefficiencies of central plan-ning there existed a completely sensible relation between laborinputs and land yields During the reform periodrsquos lsquolsquomovement to

42 The fteen provinces are Inner Mongolia Liaoning Jilin HeilongjiangShanghai Anhui Jiangxi Henan Hunan Sichuan Guizhou Yunnan QinghaiNingxia and Xinjiang Together these provinces accounted for 515 percent ofgrain output in 1978


marketrsquorsquo this relation disappeared It is difficult to argue that thecomplete elimination of a correlation between factor intensitiesand yields is consistent with the rationalization of production andfactor allocations in response to market incentives It is quiteconsistent however with an economy where the regional struc-ture of production through local bureaucratic control has been

FIGURE VIIFactor Intensity and Productivity (15 provinces)


distorted away from patterns of comparative advantage be thesefounded in the weather or other sources

IV CONCLUSION

The central proposition of this paper is seemingly unbeliev-able One is asked to accept that twenty years of economic reformin the Peoplersquos Republic which have brought extraordinaryeconomic growth and burgeoning international openness haveresulted in a fragmented internal market with efdoms controlledby local officials whose economic and political ties to protectedindustry resemble those of the Latin American economies of pastdecades It seems plausible that the endogenous response ofactors to the rent-seeking opportunities created by gradualistreform could give rise to new distortions whose lifespan farexceeds that of the rents which motivated their arrival Neverthe-less it seems hard to believe that this paper drawing itsinspiration from a broad reading of popular Chinese media hashit upon a creditable example

The central proposition of this paper is surprisingly sup-ported by a wide variety of data Anecdotal stories of industrialduplication are supported by data on the structure of GDP whichshow widespread convergence in the structure of productionduring the reform period despite what is reputed to have alreadybeen extensive and presumably excess diversication under theplan The interregional variation of prices rose during the 1980sand then uctuated wildly during the 1990s patterns compatiblewith anecdotes of interregional trade wars that are periodicallysuppressed by central intervention While regional relative out-puts have converged regional relative factor allocations and laborproductivities have bizarrely diverged which undermines at-tempts to explain the output movements as driven by free marketforces and patterns of comparative advantage Finally the natu-ral link between labor intensity and agricultural yields presenteven under central planning completely vanishes during thereform period while the correlation between agricultural factorallocations and the quality of the weather weakens as provinceswith better weather conditions shed factors of production Al-though individual pieces of this evidence can be explained awayin their totality they are difficult to dismiss There is everyindication that the economy of the Peoplersquos Republic whileopening up internationally has become fragmented internally


Fundamentally this paper argues that the so-called lsquolsquoliberal-izationrsquorsquo and lsquolsquotransformationrsquorsquo of the Peoplersquos Republic over thepast twenty years is perhaps best characterized as a process ofdevolution Although the central government has released controlover prices outputs and enterprise budgets these functions havebeen taken up albeit in a less systematic fashion by localgovernments Thus China has moved from having one centralplan to having many mutually competitive central plans It is nothard to believe that the control and incentive problems thatplague planning are more easily managed at the provincial andcounty level than they are at the continental level Consequentlywhatever the welfare costs of the new distortions it is notsurprising that on net the devolution of power has greatlyincreased welfare as witnessed by the extraordinary growth ofthe past two decades43 An economic system characterized bygovernment control even local government control over pricesoutput and investment however is not typically viewed as beingconducive to long-term prosperity It remains to be seen whetherthe central government of the Peoplersquos Republic can wrest backenough power from its regions enforcing the moral equivalentof the interstate commerce clause of the U S Constitutionrestricting the power of local governments to control the interre-gional movement of goods and factors without however reestab-lishing inefficient central control over the same This would allowit to bring out the most virtuous aspects of interregional local

43 Surprisingly it is actually possible to argue that overall improvements inefficiency during the reform period have not been that high The Chinese nationalaccounts are somewhat unusual in that the statistical authorities depend uponrms to report both the nominal and real ie ination-adjusted value of outputAs noted by Ruoen [1995] and Woo [1995] it is likely that most enterprises do notmake proper adjustments for ination The Chinese statistical authorities dohowever separately collect price statistics If one uses these price indices to deatethe nominal value of output nonagricultural GDP growth during the reformperiod is found to be about 25 percent slower per annum than officially reportedCrude total factor productivity calculations then suggest that taking into accountrising participation rates improved educational levels and the transfer of labor outof agriculture total factor productivity growth in the nonagricultural economyduring the reform period has been about 14 percent per annum respectable butby no means as extraordinary and unprecedented as the output growth reported inofficial statistics Thus it is possible that the reforms and local governmentcontrol while expediting a movement of (possibly slack) resources out of agricul-ture and continuing to encourage valuable investment in physical and humancapital have not been especially successful in raising the efficiency with whichresources are used in the nonagricultural sector This conclusion howeverdepends upon the adjustment of the Chinese output deators If one excepts theofficially reported deators then post-reform TFP growth in the nonagriculturalsector is spectacular ie at least 3 percent per annum for twenty years Thesecalculations are discussed further in Young [2000]


government competition while restricting its more perniciousmanifestations

APPENDIX SOURCES

Table I and Figure I In Chinese data government revenue bysector is computed net of enterprise losses Historically pricesubsidies were counted as negative revenue while debt issues andpayments were counted as part of revenue and expenditurerespectively Price subsidies were shifted from revenue to expendi-ture in 1986 Beginning in 1994 debt and interest were no longerincluded under revenue and expenditure Further in that year allcapital construction nanced by foreign borrowing was removedfrom expenditure Using the data on various budgetary items inthe China Statistical Yearbook and the China Public FinanceYearbook I construct a consistent series which underlies the datain Table I and Figure I using the following denitions (1) pricesubsidies are counted as part of expenditure not negative reve-nue (2) debt issues and interest payments are excluded fromrevenue and expenditure (I could not separate out interestpayments alone) (3) capital formation nanced by foreign borrow-ing is included in expenditure and (4) revenue by sector continuesto be computed net of enterprise losses

An important issue is the treatment of lsquolsquoextra-budgetaryrsquorsquorevenue which grew from 31 percent of revenue in 1978 (using myseries which matches the most recent official denition of reve-nue) to 111 percent in 1992 Until 1992 the denition of extra-budgetary revenue included the various surcharges and leviesraised by government departments as well as the many fundsretained by enterprises but earmarked for specic use egwelfare bonus and capital construction and technical updatingfunds Although local governments tap these funds [Wong 1991a]it seems excessive to consider the full value of these funds as beingunder government budgetary control In 1993 Chinese statisticsstopped counting the funds of state-owned enterprises and agen-cies as part of extra-budgetary revenue lowering extra-budgetaryrevenue to 33 percent of the value of budgetary revenue in thatyear In any case since I have not been able to nd data onextra-budgetary funds by sector of origin or a historical break-down of nonstate enterprise extra-budgetary funds into local andcentral categories I do not include extra-budgetary funds in theanalysis of Table I and Figure I


I should note that extra-budgetary funds are not the soleproblem with Chinese budgetary data While villages have noformal budgetary authority (and are not counted in the consoli-dated budget) they have developed their own revenues andexpenditures [Wong 1997 pp 198ndash199] At the same time otherlevels of government have shifted both personnel and capitalexpenditures off-budget [Wong Heady and Woo 1995 pp 129ndash130] In sum it is clear that during the reform period the officialbudgetary data account for an increasingly small share of totalpublic nance However trends during the reform period are notmy emphasis I use the data in Table I and Figure I merely toestablish the important revenue role of industry at the beginningof the reforms and emphasize the meaningless character ofdistinctions between lsquolsquolocalrsquorsquo and lsquolsquocentralrsquorsquo revenue in a systemwhere virtually all revenue was collected by local authorities

Table II and Figures II III and IV Each province in Chinapublishes a statistical yearbook The data on provincial GDP bysector come from the annual issues of these yearbooks Nationalincome by sector is compiled from the Hsueh Li and Liu [1993] ACompilation of Historical Statistics [State Statistical Bureau1990] and the annual issues of the China Statistical Yearbook

I should note that until 1988 China was divided administra-tively into 29 provinces provincial level cities and autonomousterritories In 1988 Hainan formerly a part of Guangdongbecame a separate province while in 1997 Chongqing city wasremoved from Sichuan Revised historical data separating Hainanand Guangdong are generally available although similar dataseparating Chongqing from Sichuan are not So to maintainconsistency I add Chongqing back into Sichuan in 1997 butotherwise work with the 30 provincial boundaries in place in 1996

Table III and Figure V The annual data come from the1988ndash1992 and 1994 issues of China Price Statistics Yearbook (thepublication does not appear to have been produced in 1993) whilethe monthly data come from the 490 through 699 issues of ChinaPrice (magazine) Each issue of the Price Statistics Yearbookprovides data on the current and previous year so the dataoverlap from publication year to publication year Since manylocales frequently go unreported but these vary from year to yearI construct the largest possible data set by taking the union of theoverlapping yearbooks treating the most recent publication asmore accurate in cases where the two publications provideinconsistent data on the same locale The precise grade of product


reported for each locale (eg rst grade soya beans or secondgrade soya beans) varies from year to year However it also variessubstantially across locales within years Consequently combin-ing years does not substantially increase this source of variation44

For the retail price data the rst (1988) Price Statistics Yearbookprovides data on average prices in state commercial establish-ments and in some cases average prices in all commercialestablishments I use the data on all commercial establishmentswhenever these are available Excluding the data on state commer-cial establishments from the sample only strengthens the results(ie the positive trend in retail prices is greater) Although it isnot specied the data in later issues of the Yearbook based upon acomparison with the data in the 1988 volume appear to cover allcommercial establishments

Regarding the monthly data the data for 696 are an exactduplicate of the data for 596 so I exclude them from the samplePart of the printed data for 193 and 293 are hopelessly jumbledand these also are excluded from the sample The monthly datawere actually collected every ten days but China Price magazinereports only one date within the month (I have been unable to getthe other price reports) For the period 390 to 1192 the data arefor the fth of the month for the period 1292 to 1093 they coverthe fteenth of the month while for the period 1093 to 599 theycorrespond to the twenty-fth of the month (112593 is unre-ported) I insert separate time dummies for the two dates (15 and25) in 1093

Finally I should note that I eliminate all products with lessthan ve price observations from the sample to avoid having theresults driven by standard deviations calculated off of tinysamples I also eliminate the annual data on Peoplersquos Dailynewspaper and Mass Movies magazine as the reported prices ofthese with a few rare exceptions are identical everywhere

Table IV The Chinese GDP data are the same as those used inTable II ie are drawn from the provincial yearbooks TheChinese employment data are compiled from the provincialyearbooks A Compilation of Historical Statistics [State StatisticalBureau 1990] and Hsueh Li and Liu [1993] I take the provincial

44 I have tried however to identify substantial changes in the universe ofproducts covered and when these occur use them to dene new product typesThus for example in the monthly data there is a shift from reporting the prices ofordinary our to those of enriched our In the annual data these are consideredsubstantially different products (ie are reported separately) and I treat them assuch in the analysis of the monthly data


yearbooks as more accurate where these disagree with the othertwo sources I should note that the provincial employment num-bers differ slightly from those published in national sources suchas the China Statistical Yearbook Similarly the provincial num-bers are slightly different than the total of the subprovincialestimates (also reported in the provincial yearbooks) Usuallyhigher level Chinese data are simply the sum total of lower levelestimates This is not the case for employment data where eachlevel uses its own surveys and estimation procedures The differ-ences however are quite small

The United States data come from the Bureau of EconomicAnalysisrsquo disk lsquolsquoGross State Product by Industry 1977ndash1994rsquorsquo andCD-Rom lsquolsquoState Personal Income 1969ndash95rsquorsquo

Table V and Figures VI and VII The primary employmentdata are the same as those used in Table IV Data from the ChinaStatistical Yearbook on monthly rainfall in a major city in eachprovince are used to calculate the mean monthly lsquolsquoprovincialrsquorsquorainfall and its coefficient of variation (although a few extra citiesappear in the Yearbook over time the historical issues of theYearbook only present data on rainfall in one city in each provinceso the choice of city is not an issue) All of the remaining data arecompiled from the China Statistical Yearbook A Compilation ofHistorical Statistics [State Statistical Bureau 1990] and HsuehLi and Liu [1993] As the Yearbook data for the mid-1980s do notseparate Hainan from Guangdong or Chongqing from Sichuan inthe analysis I use the 29 provincial denitions in place in 1985adding Hainan to Guangdong and Chongqing to Sichuan asnecessary to generate a consistent series

GRADUATE SCHOOL OF BUSINESS UNIVERSITY OF CHICAGO

REFERENCES

Burns John P lsquolsquoChinarsquos Nomenklatura Systemrsquorsquo Problems of Communism XXXVI(SeptemberndashOctober 1987) 36ndash51

Chen Yong Ju A Study of Interregional Market Blockades in China (ZhongguoDiqujian Shichang Fengsuo Wenti Yanjiu) (Fuzhou Fujian Peoplersquos Publish-ers 1994)

China Price (Zhongguo Wujia) monthly issueslsquolsquoChinarsquos Feuding Regionsrsquorsquo The Economist April 20 1996Chinese Academy of Social Sciences lsquolsquoResearch into the development of Chinarsquos

Internal Marketrsquorsquo Finance and Trade Economics (Caimao Jingji) No 7 (1993)35ndash40

Chinese Economic Studies XXVI (Fall 1993) entire issue on local protectionismDewatripont Mathias and Gerard Roland lsquolsquoThe Design of Reform Packages under

Uncertaintyrsquorsquo American Economic Review LXXXV (December 1995) 1207ndash1223


Falvey Rodney E lsquolsquoComparative Advantage in a Multi-Factor Worldrsquorsquo Interna-tional Economic Review XXII (June 1981) 401ndash413

Far Eastern Economic Review various issuesFord J L lsquolsquoOn the Equivalence of the Classical and the Factor Models in

Explaining International Trade Patternsrsquorsquo The Manchester School XXXV(May 1967) 185ndash198

Foreign Broadcasting Information Service Daily Reports Peoplersquos Republic ofChina

Gelb Alan Gary Jefferson and Inderjit Singh lsquolsquoCan Communist EconomiesTransform Incrementally The Experience of Chinarsquorsquo in Olivier Jean Blan-chard and Stanley Fischer eds NBER Macroeconomics Annual 1993 (Cam-bridge MA MIT Press 1993)

Greene William H Econometric Analysis second edition (New York MacmillanPublishing Co 1993)

Hsueh Tien-tung Qiang Li and Shucheng Liu Chinarsquos Provincial Statistics1949ndash1989 (Boulder CO Westview Press 1993)

Jefferson Gary H and Thomas G Rawski lsquolsquoEnterprise Reform in ChineseIndustryrsquorsquo Journal of Economic Perspectives XIII (Spring 1994) 47ndash70

Kumar Anjali China Internal Market Development and Regulation (WashingtonDC The World Bank 1994a) lsquolsquoChinarsquos Reform Internal Trade and MarketingrsquorsquoPacic Review VII (1994b)323ndash339

Li David lsquolsquoCosts and Benets of Government Control During the TransitionEvidence from Chinarsquos State Enterprisesrsquorsquo manuscript University of Michi-gan 1996

Manion Melanie lsquolsquoThe Cadre Management System Post-Mao The AppointmentPromotion Transfer and Removal of Party and State Leadersrsquorsquo ChinaQuarterly CII (June 1985) 203ndash233

Martinelli Cesar and Mariano Tommasi lsquolsquoSequencing of Economic Reforms in thePresence of Political Constraintsrsquorsquo Working Paper No 100 University ofMaryland Center for Institutional Reform and the Informal Sector 1994

Mascaro Juan The Upanishads Translations from the Sanskrit (Harmonds-worth UK Penguin Books 1965)

Murphy KevinAndrei Shleiferand Robert W Vishny lsquolsquoThe Transition to a MarketEconomy Pitfalls of Partial Reformrsquorsquo Quarterly Journal of Economics CVII(August 1992) 889ndash906

Naughton Barry lsquolsquoFalse Starts and Second Wind Financial Reforms in ChinarsquosIndustrial Systemrsquorsquo in Elizabeth J Perry and Christine Wong eds ThePolitical Economy of Reform in Post-Mao China (Cambridge Council on EastAsian Studies Harvard University 1985) lsquolsquoIndustrial Policy during the Cultural Revolution Military PreparationDecentralization and Leaps Forwardrsquorsquo in William A Joseph Christine P WWong and David Zweig eds New Perspectives on the Cultural Revolution(Cambridge Council on East Asian Studies Harvard University 1991)

Peoplersquos Republic of China China Public Finance Editorial Board China PublicFinance Yearbook (Zhongguo Caizheng Nianjian) (Beijing China PublicFinance MagazineAgency annual issues)

Peoplersquos Republic of China Public Finance Department Chinarsquos Public FinanceStatistics (Zhongguo Caizheng Tongji) 1950ndash1985 (China Public FinanceEconomics Publishing House)

Peoplersquos Republic of China State Statistical Bureau A Compilation of HistoricalStatistical Material for the Provinces Autonomous Territories and DirectlyAdministered Cities 1949ndash1989 (Quanguo ge Sheng Zizhiqu Zhixiashi LishiTongji Ziliao Huibian) (Beijing China Statistical Publishing House 1990) China Price Statistics Yearbook (Zhongguo Wujia Tongji Nianjian) (BeijingChina Statistical Publishing House annual issues) China Statistical Yearbook (Zhongguo Tongji Nianjian) (Beijing ChinaStatistical Publishing House annual issues)

Perkins Dwight Heald lsquolsquoReforming Chinarsquos Economic Systemrsquorsquo Journal of Eco-nomic Literature XXVI (June 1988) 601ndash645

Ruoen Ren lsquolsquoChinarsquos Economic Performance in International Perspectiversquorsquo manu-script OECD Development Centre October 1995


Song Xinzhong ed Study of the Reform of Chinarsquos Public Finance System(Zhongguo Caizheng Tizhi Gaige Yanjiu) (Beijing Zhongguo Caizheng JingjiChuban She 1992)

Unger Jonathon lsquolsquoThe Struggle to Dictate Chinarsquos Administration The Conict ofBranches vs Areas vs Reformrsquorsquo Australian Journal of Chinese Affairs 0 (July1987) 15ndash45

Wong Christine P W lsquolsquoMaterial Allocation and Decentralization Impact of theLocal Sector on Industrial Reformrsquorsquo in Elizabeth J Perry and Christine Wongeds The Political Economy of Reform in Post-Mao China (Cambridge Councilon East Asian Studies Harvard University 1985) lsquolsquoBetween Plan and Market The Role of the Local Sector in Post-Mao ChinarsquorsquoJournal of Comparative Economics XI (1987) 385ndash398 lsquolsquoInterpreting Rural Industrial Growth in the Post-Mao Periodrsquorsquo ModernChina XIV (January 1988) 3ndash30 lsquolsquoCentral-Local Relations in an Era of Fiscal Decline The Paradox of FiscalDecentralization in Post-Mao Chinarsquorsquo China Quarterly CXXVIII (December1991a) 691ndash715 lsquolsquoThe Maoist lsquoModelrsquo Reconsidered Local Self-Reliance and its Financing ofRural Industrializationrsquorsquo in William A Joseph Christine P W Wong andDavid Zweig eds New Perspectives on the Cultural Revolution (CambridgeCouncil on East Asian Studies Harvard University 1991b) lsquolsquoFiscal Reform and Local Industrializationrsquorsquo Modern China XVIII (April1992) 197ndash227 ed Financing Local Government in the Peoplersquos Republic of China (HongKong Oxford University Press 1997)

Wong Christine P W Christopher Heady and Wing T Woo Fiscal Managementand Economic Reform in the Peoplersquos Republic of China (Hong Kong OxfordUniversity Press 1995)

Woo Wing Thye lsquolsquoChinese Economic Growth Sources and Prospectsrsquorsquo manuscriptUniversity of California Davis October 1995

Young Alwyn lsquolsquoGold into Base Metals Productivity Growth in the PeoplersquosRepublic of China during the Reform Periodrsquorsquo NBER Working Paper No 7856August 2000


I INTRODUCTION

Analyses of the transition to market in centrally plannedeconomies have identied two potential pitfalls in the process ofincremental economic reform First the removal of government-mandated controls in one part of the economy in the context ofcontinued distortions elsewhere can for standard second-bestreasons lead to a fall in output [Murphy Shleifer and Vishny1992] Second political opposition may develop along the path ofreform stalling the reform process or reversing it altogether[Dewatripont and Roland 1995 Martinelli and Tommasi 1994] Inthis paper I use the reform experience of the Peoplersquos Republic ofChina to highlight a third potential pitfall Incremental reformreleases segments of the economy from centralized control whilemaintaining for a prolonged period many of the distortions of thecentral plan The freed segments of the economy nd it protableto exploit the rent-seeking opportunities implicit in the remainingdistortions of the economy Their attempts to capture and thenprotect these rents leads to the creation of new distortions In thissense the reform process can be derailed while some distortionsare eliminated moving the lsquolsquotrainrsquorsquo of the economy to marketothers are added drawing it off on unexpected tangents

As early as 1986 the government of Zhejiang province beganto restrict the sale of silk cocoons to other parts of China Eachharvest season the government bureaucracy from the provinciallevel all the way down to village units and including police andmilitia forces was mobilized to ensure that Zhejiangrsquos silk farm-ers sold their silk cocoons locally for local processing or directexport Under the plan the prices of raw materials such as silkwere kept articially low while the prices of nal goods such assilk fabric were kept high During the 1980s with the relaxationof central controls on rural industry locally controlled silk-processing industries developed rapidly in Zhejiang a traditionalproducer of silk cocoons to capture the rents implied by thearticial price wedge The trade embargos of the mid-1980sonward appeared as other provinces tried to purchase silk co-coons at above planned procurement prices directly from Zhe-jiang farmers bidding away the rents the Zhejiang governmentcould acquire from local processing or direct export1 In the rsthalf of 1988 silk factories in Shanghai located next door toZhejiang and Jiangsu the principal domestic producers of silk

1 See Chen [1994] preface pp 1ndash2 and text pp 7ndash8











II STORIES








1952 036 021 024 005 000 0131965 073 007 007 009 000 0041978 075 003 012 007 000 0031995 044 006 033 004 002 012

















































III DATA





A Outputs





oio

jSij 2 Sj o

io

j(Sij 2 Sj)2




FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)






Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES













































II STORIES








1952 036 021 024 005 000 0131965 073 007 007 009 000 0041978 075 003 012 007 000 0031995 044 006 033 004 002 012

















































III DATA





A Outputs





oio

jSij 2 Sj o

io

j(Sij 2 Sj)2




FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)






Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES


















































































III DATA





A Outputs





oio

jSij 2 Sj o

io

j(Sij 2 Sj)2




FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)






Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES





































A Outputs





oio

jSij 2 Sj o

io

j(Sij 2 Sj)2




FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)






Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES




































FIG

UR

EII

Con

verg

ence

inth

eC

ompo

siti

onof

Out

put

(dev

iati

onfr

omav

erag

epr

oduc

tion

stru

ctur

ein

curr

ent

pric

es)






Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES








































Absolute Squared



1978 691 294 397 700 373 327 100 053 047 104 068 0371997 523 228 294 468 278 190 052 025 027 036 022 0147897 076 078 074 067 075 058 052 047 058 035 033 038


Absolute Squared



1978 715 295 419 688 342 346 105 051 054 098 056 0421997 537 196 341 406 199 207 057 023 034 030 014 0177897 075 066 081 059 058 060 055 045 064 031 024 039






oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES







































oio

j30 p wi Sij 2 Sj













B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES












































B Prices










Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES







































Retailprices


prices

Industrialmaterials

prices






















lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES





















































lnPPQP

PSQSln

PPQPLP

PSQSLSln

LP

LSln

PPQPLP

PSQSLS ln

LP

LS


1978 071 012 056 0011997 053 015 073 2 017


1977 118 031 063 0121994 072 032 037 001


1952 154 044 085 0131965 098 016 064 0091978 072 016 063 2 0041997 058 020 068 2 015







































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES







































































(1)Yield

(2)Irri

(3)Fert

(4)Power

(5)Labor

(6)Lshare

(7)OutputC

(8)Yield


Year 009 019 091 049 006 2 004 027 026(31) (83) (332) (205) (37) (76) (114) (59)

Irri 309 199(63) (45)

Fert 214 155(62) (43)

Power 2 043 2 080(10) (19)

Labor 2 091 2 148(15) (24)

Prec 021(26)

Prec2 2 001(27)

CVPrec 2 040(19)

Wtime 2 233 2 549 2 194 2 155 2 080 2 426 2 260(48) (94) (38) (44) (64) (85) (56)

R2 942 958 954 970 975 984 922 944


Year 019 091 050 006 2 004 027(18) (107) (40) (07) (10) (33)

Weather 2 540 124 2 327 734 790 224(03) (96) (18) (53) (12) (18)

Wtime 2 233 2 549 2 198 2 155 2 080 2 426(10) (30) (08) (08) (09) (25)

R2 027 522 136 171 029 031N 374 377 377 376 377 377 377 376

























IV CONCLUSION








APPENDIX SOURCES




















REFERENCES


























































IV CONCLUSION








APPENDIX SOURCES




















REFERENCES








































APPENDIX SOURCES




















REFERENCES




















































REFERENCES




































quarterly journal of economics - stanford …web.stanford.edu/~klenow/the razor's edge.pdf ·...

Documents