【原创】China in World Industrialization
XuYi(徐毅)1and Bas van Leeuwen (巴斯·范鲁文)2
1Guangxi Normal University,Guilin, China
2International Institute ofSocial History, Utrecht University,theNetherlands
Abstract:Combiningthe sectoral accounting method of the Systemof NationalAccounts (SNA) withnew statistical materials from the United Nations, as wellas historical research into various countries around the world, this paperarrives at an estimate of value added of Chinese and world industries between1850 and 2012. In doing so, we not only modify past work byPaul Bairoch, butalso arrive at significantly different conclusions about China’s position inworld industrialization.
Key words: China, industrialization, value added, world
JEL:L60, N60, O55
1. Introduction
Aspart of the process of world industrialization, the British IndustrialRevolution of the late 18th century spread across the Europeancontinent, the Americas and Asia later in the 19th century.Countries that embarked upon the path of industrialization began to record and accountfortheir industrial process; leading eventually to the estimation of complexindustrial aggregates. Despite the growing quantity of industrial productionestimates at that time, not much progress was made in theirquality andcomparability (Wagenfuhr, 1933). With the creation of the System of NationalAccounts (SNA) in the 1930s, scholars around the world slowly began to use thissystem, thus enhancing international comparability, to account for thelong-term industrial performance of various countries.
Basedon the SNA’s standard definition, the accounting of industrialization wasgenerally based on three core indicators, i.e. a benchmark of total valueadded, the growth rate (or index) of industrial output, which could be linkedto the benchmark to obtain a time series of value added, and population (toarrive at per capita estimates).In his two monographs published in the 1940sand 1950s, the German economist Hoffmann (Hoffmann, 1965) attempted to usethese three indicators instead of being confined to the previous incomparableoutput indicators to account for the industrial growth of the United Kingdombetween 1700 and 1950 and Germany between 1850 and 1960. In doing so, Hoffmannmade a great contribution to the quantitative study of world industrialization;to this day, his works remains among the classical literature in the Europeanhistory of industrialization.
It was not until the last decade of the Qing Dynasty (1644-1912) thatChina began to conduct economic surveys using modern statistical methodologies.After the fallof the Qing Dynasty, the Ministry of Industry and Commerce of theBeiyang Government (renamed “Ministry of Agriculture and Commerce” in 1912) setthe statistical formats and charters for the survey of agriculture, industryand commerce through a ministerial decree in November 1911, the first year ofthe period of the Republic of China (1911-1949). Sincethen, a large number of economic surveyswas carried out on a nationwide basis.The survey of agriculture, industry and commerce started in 1911 and continueduntil 1921, resulting in the “Statistical Statements of Industry and Commerce” fornine years. In these statistical statements, two chapters of Industry and Mining were dedicated to recording the information of relevantenterprises (those employing more than seven persons), labor and variousindustrial and mining productions. Despite their valuable information, theseearliest Chinese industrial statistics had numerous mistakes and omissions dueto the problematic economic, social, and military conditions at that time andare therefore infrequently used by economic historians of today (Chang, 1965).
In 1937, Lieu,D.K.(1937) published theReport of China’s Industrial Survey, which was “far morecomprehensive and precise than any previous industrial statistical survey andeven better than the industrial survey programs carried out by the UnitedKingdom and the United States.” It became a major source of data forfuturescholarsattempting to estimate China’s industrial output value in the 1930s. On thebasis of this survey, Lieu,D.K. creatively followed the value added methodology toestimate China’s industrial growth in the 1930s, later published in English as China’s National Income from 1931 to 1936(Liu, 1946). Later, Ou,P.S. et al. provided an evenmore comprehensive estimate of the value added of China’s machinery industryand handicraft industry in 1933 in China’sNational Income: 1933 published in 1947. In two of his papers published inthe same year, Ou,Pao-san.et al.adjusted theindustrial output value of 1933, mainly the value added of the machineryindustry and, in addition, created an industrial value added series coveringthe years 1929, 1931-1936 and 1946 (Ou,P.S. et al., 1947). In 1965, Liu, Ta-chung and Kung-chia Yehreworked and modifiedthese previous estimationsin their bookEconomy of Mainland China: National Income and Economic Development, 1933-1959to arrive at the new estimate of China’s industrial output value in 1933, whichbecame hugely influential in Western academia (Liu &Yeh, 1965). Nevertheless,due to limitations of information and methodology, no Chinese scholar has evertraced back the accounting of China’s industrial value added, other than forone or two individual sectors, to the 19th century, not to mention calculatethe share and position of China’s industry in the world total during differentperiods of time.
In Westernacademia, in 1982Belgianeconomist Paul Bairoch published his estimate ofthe industrial (mainly manufacturing) output of major countries around theworld forthe nine separate years (1750, 1800, 1860, 1980,1900, 1913, 1928, 1953 and 1980). His major conclusions were as follows: by themiddle of the 18th century, China remained a major industrial powerin the world, while Europe overtook China somewhere between the 18thand19th century; yet Asian countries, including China, began to catchup in the second half of the 20th century (Bairoch, 1982). Bairoch’squantitative study on the world industrialization process aroused significantattentionin international academia and became important literature being widely cited byscholars around the world. Nevertheless, Bairoch’s quantitative study, particularly his estimate before 1952, still has significant room for improvement, asmany of his estimates were purely established on simple assumptions and theinformation and estimation method he employed were not explained in detail (Bairoch,1979, 1982). Yet, over the past decade international academia have seen anemergence of new quantitative results of studies on the industrialization ofvarious countries, including a recent study on the industry of China’s QingDynasty. Thus, this paper attempts to utilize these various new results toadjust the grand vision of world industrialization revealed to us byPaul Bairoch 30 years ago and, more importantly, to re-evaluate the positionand role of China over the long course of world industrialization.
2. Concept, Methodology and Data
2.1 General data
Since there is nocommon definition of industry, in the SNA it is defined as the share ofnational income that does not belong to agriculture or services. Hence, thespecific differences of manufacturing process and technology, i.e., the differences between the machinery industryand handicraft industry, are not considered in the accounting of industrial valueadded (United Nations, 2008). In accounting for industrialization using valueadded, previous scholars generally calculated the value added of the modern industryand handicraft industry in combination to arrive at the gross industrial valueadded; the same is true for the accounting method in this paper. In addition,Paul Bairoch only accounted for the growth of the manufacturing industry,while we estimate the value added of the manufacturing, mining and constructionsectors respectively following the SNA’s definition of industry. It should befurther noted that we follow the currently most reliable and popularinternational currency settlement unit in international academia, i.e., the 2005 International dollars, as thecurrency unit in estimating the industrial value added of various countries.According to the level of data availability, we collect and reconstruct thelong-term series of industrial value added of various countries for thefollowing 6 main sources.
First,we directly extract the industrial value added series of 212 countries denotedin the constant US dollar price of 2005 from 1970 to 2012 in the UN database ofnational accounts. Second, for the period from 1950 to 1970, we extract theindustrial value added indices of most countries from the Yearbook of National Accounts (United Nations, 1958-1976) and the United Nations Statistical Yearbook(United Nations, 1958-1968). It should be noted that the data series for thepre-1970 period are not expressed in the 2005 International dollarsbut aregiven in constant prices in local currencies. Since these constant pricebenchmark may have different price ratio’s among the various industrial sectorsthan 2005, this may slightly affect the growth rates of industrial output butavoids the more serious “Gerschenkron effect” as described below.
These pre-1950 constant price series than need to be converted into the 2005 Intldollar to link with the value added series after 1970.As pointed out before, inlinking the data series with different sectoral weights, we try to avoid the“Gerschenkron effect”, i.e. by overweighing fast growing time series the growthrate will be overestimated and, consequently, industrial production will beunderestimated the further back in time one goes. If, for example, we have twotime-series in constant prices for 1950-2000. One of them grows fast and theother slow. In 2000 the fast growing series has a value of 7 and the slowgrowing one has a weight of 3, i.e. weights of 70% and 30%. In 1950 the fastgrowing one has a weight of 1 and the slow growing one of 2. If we were to takeour 2000 weights, the industrial output in 1950 would be 1*0.7+3*0.3=1.6. Henceindustry grows from 1.6 to 10 between 1950 and 2000. However, actual growth isfrom 3 to 10. Hence, the most desirable long-run series of value added arecreated through changing weights of different sectors due to changing pricesand output in each year, rather than relying on the prices and sectoral weightsof a certain year. However, due to limited information in the study of economichistory, it is unlikely to modify the prices and weights of each year and it isonly possible to do so for a few benchmark years with concentrated price andoutput data. As mentioned above, we attempt to link the value added series withdifferent sectoral rates, thus minimizing the “Gerschenkron effect.”
Third, as for the years before 1950, we rely onthe studies of international academia on the industrial value added ofindividual countries. Over the past few decades, international academia hasyielded a host of study results on the industrial value added of countries inEurope, Latin America and Asia before 1950, as shown in Table 1:
Fourth, for those countries whose industrial valueadded cannot be directly obtained, we try to extract the industrial outputindices of these countries from the InternationalHistorical Statistics compiled by Mitchel (Mitchel, 2007). Following thecommon practice of international academia, it is assumed that the ratio betweenindustrial value added and gross output of these countries before 1950 isconstant (Hoffmann, 1955; Allen,2000; van Zanden& van Leeuwen, 2012), andwe directly use the industrial output indices of these countries as the valueadded indices to link with their industrial value added series after the 1950s.
Fifth,for countries with still less information such as Poland, Romania and Bulgariain Eastern Europe, we can only find relevant indices that reflect industrialproduction in the InternationalHistorical Statistics, such as the population index that reflects thedevelopment of the construction sector, the output index of certain mineralproducts (such as coal and petroleum), and the raw materials index orconsumption index of certain manufacturing industries (such as clothing,cotton, wool and beer). As mentioned before, the available data of industrial valueadded for various countries of Eastern Europe after 1950 suggest that theproportions of various industries are as follows: the construction sectoraccounts for 20%, mining 10% and manufacturing 70%. We assume that suchproportions of industrial sectors also apply to various countries of EasternEurope before 1950 and thus arrive at their industrial output indices based onvarious indices of certain sectors. Then, using the same method, the industrialoutput indices of various countries of Eastern Europe are linked with theirindustrial value added series after 1950.
Sixth,for those countries for which no data at all were available, we assumed that theirgrowth rate was equal to an economically comparable country, i.e. comparable interms of size as well as economic structure. Since the larger industrializedcountries were all included, the share of countries with no data was relativelysmall. It can be seen in Table 2 that, with the exception of 1850, the share ofnational industries whose value added is directly estimated in world grossindustrial value added exceed 80% throughout the years after 1880, whichensures the reliability of our estimation.
2.2 Chinese Industrial Value Added
Here we turn in more detail to the calculation of Chinese industrial value added. Forthe post-1970 period our sources of information are the same as those for othercountries during the same period of time; the data from 1950 to 1970 are takenfrom the study by Maddison and Wu Xiaoying (Maddison & Wu, 2008). Theperiod from 1850 to 1949 will be divided into two parts for discussion: the Qing Dynasty (1644-1912) and the periodof the Republicof China(1912-1949).
Chang, Chung-li andLiu, Ti conducted a quantitative study on the outputvalue of the handicraft industry in the Qing Dynasty (Chang, 1962; Liu Ti,2010). However, due to the lack of data, their estimation was confined to arelatively specific study on a few handicraft sectors; the majority of otherhandicraft sectors were generally combined into one category of “other sectors”to estimate their output value, which greatly affected the overall reliabilityof the estimation results. Thus, we are first faced with the question of how todefine the industrial sectors of the Qing Dynasty, which are still divided intomanufacturing, mining and construction sectors as mentioned above.
As regards the classification of the manufacturing industry, we have arrived at arelatively complete classification by comparing the classification of China’sindustry in the Qing Dynasty in 1933 by Ou,P.S.andLiu, T.C. with the classification by Peng,Zeyi and Xu,Jianqinget al.(Peng,Zeyi, 1957; Li,Shaoqiang, Xu,Jianqing, 2004), i.e., by the nature of products, the total industryis divided into 12 specific sectors including food, textiles, clothing, timberprocessing, paper making and printing, metal products, earth and stone,leather, means of transport, chemicals, accessories and apparatuses, as well asmiscellaneous goods; each sector normally includes more than one sub-sector.
Our study thus follows the division of 14 sectors (including 12 industrial sectors,handicraft, and mining and construction) to respectively estimate their outputvalues (value added) and aggregate all the results to arrive at the value addedofindustry. Based on the availability of data, in a recent paper we estimate the valueadded of each specific industrial sector for eight years, i.e. 1661, 1685,1724, 1776, 1812, 1850, 1887 and 1911 (denoted by the price of 1933)(Xu,Shi ,van Leeuwen, Ni, Zhang & Ma 2016). From these benchmarks we use 1850, 1887and 1911 in this paper thus necessitating us to elaborate on their constructionin more detail.
According to the SNA, the general equation for the calculation of output value based onproduction method is as follows: value added= output×price-value of intermediateinput. However, as the historical information of industrial production in theQingDynasty may not always be in the appropriate format, a more eclectic approachis used to arrive at value added estimates. This is done by taking the relatively “solid” benchmark of 1933 as pointof departure and then using indices that reflect industrial output by sectorand linking those to 1933. Under the assumption that the ratio betweenproduction and value added remains constant by sector, this gives us industrialvalue added for each of our benchmark years in 1933 prices.
As mentioned before, for 1933 Ou, P.S.et al.createda benchmark on industrial value added, which was then adjusted by Liu, Ta-chungand Kung-chia Yeh. Hence, we used the updated estimates from Liu, Ta-chungand Kung-chia Yehwith the exception of two industries. First,the data of the construction sector are based on the estimation results updated byK. C. Yehin 1979(Yeh, 1979). The other is the mining industry: as the estimateby Ou, P.S. et al.in 1947 is more detailed and comprehensive, wehave adopted their estimation results forthe mining industry. On this basis, ournext quest was to find data that reflected sectoral industrial production andcould be linked to the 1933 benchmark. We did this in 5 steps:
(1)Direct calculation. The most important aspect of this method is to directly orindirectly acquire the output data of a certain product, i.e. the complete orrelatively complete output data of a certain industry or sector can be founddirectly. For the mining, silk and ceramic sectors, as well as industrialsectors such as weapon manufacturing, minting, banknote printing and papermaking, we have acquired relatively complete industrial data through extensivecollection of historic information; the output of the printing industry can beestimated with relatively satisfactory results by calculating the number ofbook titles that have been printed. Another option is that despite the lack ofcomplete output records, sufficient clues of industry development indices canbe acquired from relevant historic information, such as labor productivity, thenumber of factories, the number of craftsmen, taxation, consumption, import andexport volume, as well as the composition of a certain sector. By consolidatingsuch information, we may also straightforwardlyestimate output data (Xu Yi,2014; Xu Yi, Zhang Zipeng, 2015).
(2)Estimation based on depreciation rate. For some sectors such as shipbuilding,although existing historical information does not provide any output data,such information as the quantity, average length of service and number ofnormal maintenance of ships can be roughly acquired from various sources. Based on such information, we are able to roughlyestimate the normal quantity of replacement for a year (i.e., production output) based on the depreciationrate. This method can be regarded as a special application of the first method.
(3) Estimation based on the consumption of rawmaterials. Although output-related records for some sectors cannot be directlyobtained from historical information, we can nevertheless gauge itstrend by looking at the consumption of raw materials. These sectors include,for instance, logging, food processing, cotton textiles, silk reeling, woolweaving, basket weaving and construction sector, whose raw materials are mainlyvarious types of agricultural and forestry products. Through the estimates ofgross output in such agricultural sectors as food, cash crops and forestry byprevious studies, we may roughly estimate their quantities of consumption asraw materials for the relevant industrial sectors and further estimate theannual output based on the input-output ratio for various sectors (which isrelatively stable under traditional technology). Of course, sometimes we mayalso use this estimation method the other way around, i.e., with given output of silk weaving, theinput-output ratio for silk reeling and silk weaving is used to derive raw silkoutput, i.e., the output of the silk reeling sector.
In existing historical information, there is barely any direct recordof the output value of the construction sector. However, there are a largenumber of income and spending accounts for the construction of various types ofhousing, commercial facilities, irrigation works, roads, bridges, temples and memorialarchways. These accounts record in detail the composition of materials forvarious construction projects and the amount of spending, from which we mayknow that the output value of construction sector in the Qing Dynasty consisted of four partsincluding timber, bricks and tiles, other materials and labor cost. Based onthe regression analysis of many construction sector samples, we have discovereda significant proportionate correlation between spending on timber materialsand the total cost in various construction projects. Hence, we may estimate theoutput value of the construction sector based on the estimated output value of loggingas a sub-sector that provides raw materials to theconstruction sector.
(4)Usingconsumption to reflect output. For some sectors such as the manufacturing ofvehicles, metal products, chemicals, accessories, apparatuses and miscellaneousgoods, although it is difficult to acquire output-related data from existinghistorical information, these sectors are highly relatedto the daily life of the people. As a result, daily consumption records of suchproducts exist extensively in historical information. In this situation, we replaceproduction output with consumption in order to estimate output value.
(5)The output of certain industries or sectors is estimated with the method ofproportionate growth rates. Some sub-sectors are highly correlated with overallindustrial development. For instance, timber products and rattan, bamboo andwicker products are dependent on the development of logging; the manufacturingof metal products is dependent on the development of metal mining; leathermanufacturing is dependent on the development of animal husbandry inagriculture; the manufacturing of clothing products is dependent on thedevelopment of the textile industry; the manufacturing of paper products isclosely related to the development of paper making; the manufacturing of bricksand tiles, lime glass and other earth and stone products is highly correlatedwith the development of the construction sector, etc. For these sub-sectors, wehave reasons to assume that they maintained growth by the same proportion intandem with related industries during the period from 1850 to 1933. Thus, itshould be a viable option to use the ratio between the outputs or output valuesof these sub-sectors and related industries in 1933 to derive their output oroutput value in 1850.
Because statistical data are fragmentary due to the impact of war in 1850 and 1880,interpolations are necessary mainly for the years between 1880 and 1911 forwhich data are lacking. We first utilize the same method and similar data tocreate the annual value added series for such sectors as food, cotton and silkweaving (two major sub-sectors of textiles), printing (a sub-sector of papermaking and printing) and mining; then, the shares of these sub-sectors in theirrespective industrial sectors are used to derive the industrial value added forthe years with missing data.
Inregard to the industry during the period of the Republicof China, Ou.P.S.employed variousindices to derive the annual industrial value added series between 1931 and1936 (including manufacturing, mining and construction) and the value added of1946based on the estimates of 1933. Following the same indices, we estimate theannual industrial value added series from 1931 to 1936 and the value added in1946 based on the estimates of Liu, Ta-chung and Kung-chia Yehfor1933. Some scholars adjusted the nine rounds of agricultural and commercialsurvey statistics during theBeiyang Government period (1912-1928) and estimatedthe annual gross output of the manufacturing industry from 1912 to 1930 (GuanQuan, 2011). Furthermore, using the price information of industrial goods from1912 to 1930, we convert the gross output values estimated by previous scholarsinto gross output, assuming that the ratio between gross output and value addedis constant, and directly link the industrial output indices with adjustedmanufacturing value added series after 1931. Moreover, we separately establishthe annual value added series for China’s mining and construction sectors from1912 to 1930 to link with the adjusted value added series of these sectorsafter 1931.
3. Estimation Results: China and the World
In this part, the results of our estimation willbe presented. First, we identify the differences of our and Paul Bairoch’sestimation of China’s manufacturing value added. As can be seen from Table 3, for the same year, the value added estimated byPaul Bairochis much higherthan our result and higher by an average of morethan threetimes, which obviously overstates China’s industrial performance inmodern times. We may draw three general conclusions.
First, Figure 1presents a new estimate of China’s and the world'sgross industrial value added. As can be seen from this chart, China’sindustrial development tendency is highly correlated with the rest of theworld. Here, our discussion is made from the perspective of industrialglobalization for different stages: in Stage 1 from 1850 to 1910, Europe andthe United States experienced mass industrialization and world industrial grossoutput value grew almost six fold, while China’s industrial gross output valueonly doubled. In Stage 2 from 1910 to1980, industrialization began to spread across Asia, Africa and Latin America,and world gross industrial output value increased by 12 times while China’sgrew by 10 times. With the end of the Cold War and the opening up of China inStage 3 from 1980 to 2012, world industrial gross output value only increasedby two times, while China’s grew by 30 times.
Third,our new estimates of industrial value added for various countries havetransformed the macro pattern of world industrial layout developedbyPaul Bairoch. With the three major industrial countries of the UK, theUS and China, Table 7 shows the different positions of the three countries inthe world industry in Paul Bairoch’s and our estimates. Except for the UK,major differences exist in our andPaul Bairoch’s estimates of China’sshare in world industrial output value. For China, in particular, asPaul Bairoch overestimated modern China’s industrial output value, hisestimate is at least double the level presented in this text.