Globalization, Trade & Wages: What Does History tell us about China?

Transcription

1 Globalizaion, Trade & Wages: Wha Does Hisory ell us abou China? Kris James Michener Deparmen of Economics Sana Clara Universiy, Hoover Insiuion & NBER and Se Yan Guanghua School of Managemen Peking Universiy December 2009 Chinese impors and expors grew rapidly during he firs hree decades of he wenieh cenury as China opened up o global rade. Using a new daa se on he facor-inensiy of raded goods a he indusry level, we show ha Chinese expors became more unskilledinensive and impors became more skill-inensive during hese hree decades. The exogenous shock of World War I dramaically raised he price of Chinese expors, increased he demand for Chinese goods overseas, and increased he demand for unskilled workers producing hese goods. These rends coninued even afer he war ended. We show ha he iming of he rise in expor prices is consisen wih he observed decline in he skill premium in China. The skill-unskilled wage raio flaened ou during he 1910s and hen fell by eigh percen during he 1920s. We simulae he price shock of World War I using a general equilibrium facorendowmens model of rade and find evidence consisen wih he observed fall in he skill premium in China during he 1920s. Keywords: Inernaional rade, inequaliy, skill premium, facor endowmens, China JEL Classificaion Numbers: F15, F33, N20, N23, N40 We hank John Brown, Carolyn Evans, Kevin O Rourke, and Alan Taylor as well as seminar paricipans a UC Sana Cruz, Carlos III, and IMT Lucca for helpful commens and suggesions. Michener acknowledges he financial suppor of he Global Fellows Program, Inernaional Insiue, UCLA and he Hoover Insiuion, Sanford Universiy.

2 Globalizaion, Trade, and Wages: Wha Does Hisory ell us abou China? One of he mos conenious issues wih respec o he global growh in rade is is effecs on wages. Much of he debae has focused on how he expansion in rade beween developing and developed counries affecs he wages in he Unied Saes and Europe. The logic of rade heories such as he Heckscher-Ohlin-Vanek (HOV) model and Solper-Samuelson (1941) sugges ha an expansion of rade will raise he wages of skilled workers and lower hose of unskilled workers in developed counries if hey are relaively well endowed wih skilled labor. This has led some commenaors o link he growing wage inequaliy wihin he U.S. o rade wih he res of world; even more specifically, some have suggesed ha as U.S. rade wih China grows, wages for unskilled workers in he U.S. will fall in response. An equally ineresing quesion is how he global rade boom has alered wages in developing counries. For example, facor endowmens also predic ha he skill premium should be falling in rapidly developing counries like China where he sock of unskilled workers is large relaive o developing counries. Using ciy-level daa, Wei and Wu (2001) find evidence ha inequaliy has fallen wihin China, and ha he decline in rural-urban inequaliy have been mos pronounced in areas ha increased heir openness (rade-gdp raios). On he oher hand, oher empirical sudies sugges ha globalizaion has likely increased inequaliy in developing counries in he las hree decades, alhough hese findings depend on counry-specific and ime-specific facors. 1 For example, Wan, Lu, and Chen (2007) presen evidence ha increased FDI and rade have widened inequaliy wihin China more recenly. 1 For a survey, see Goldberg and Pacvcnik (2007). For aricles on he effecs of rade on skill premia in specific counries see, for example, Robbins (1996), Beyer, Rojas, and Vergara (1999), Gasparini (2003), Hanson (2004), and Roberson (2000, 2004). 1

3 Deermining he impac of he curren rade boom on wages in developing and developed counries is complicaed by many facors. Firs, rade has become much more complicaed han wha is ypically described in facor endowmens models, whose origins are a leas 90 years old. Inra-indusry rade, ousourcing, offshoring, and mulinaionals complicae he esing of heir heoreical predicions. More generally, confounding influences make he ask of causal inference exremely challenging for sudies examining curren rade flows. For example, in hinking abou he impac of expanding rade on he wage premium in developed counries oday, empirical researchers may also need o accoun for declining union power, falling minimum wages, increased raes of immigraion of unskilled workers, and greaer skill-biased echnological change. Similarly, analyses ha focus on he impac of rade on wages in developing counries, like China, also face an array of challenging empirical issues ha make idenificaion difficul, including assessing he impacs of echnological change, foreign direc invesmen, and sae inervenion on facor prices. A number of scholars have suggesed ha he lae nineeenh and early wenieh cenuries may be beer periods for esing he empirical predicions of facor endowmens models. Esevadeordal and Taylor (2002) argue ha low barriers o rade (especially in simple manufacured goods and agriculure), more skewed facor endowmens, less rade in differeniaed producs and services, and minimal inra-indusry rade are some characerisics of he firs era of globalizaion ha may make i a good laboraory for esing his class of models. O Rourke and Williamson (1994, 1999) have provided empirical evidence of facor price convergence and oher predicions of hese models during he grea expansion of rade in he lae nineeenh cenury. 2

4 In his paper, we use he lens of hisory o beer undersand how a rapid expansion in rade affecs he skill premium for a developing counry. In paricular, we ask how he skill premium in China responded o an earlier era of globalizaion and explosive growh in rade. We assemble new daa on Chinese rade and wages for he period from 1903 o 1928 and use he exogenous shock of World War I o shed ligh on his quesion. During he firs hree decades of he wenieh cenury, China experienced a remendous growh in rade wih he res of he world. The nominal value of expors sexupled and impors rose roughly by he same amoun (Figure 1). Alhough China had been forcibly opened o rade in he 1840s, he scale of rade was quie small unil he Treay of Shimonoseki was signed and an era of conflic wih foreign powers concluded wih Chinese supplicaion in he Boxer Uprising ( ). The beginning of our sample period hence corresponds o a period of less conflic wih foreign powers and China s enry ino he global rade boom of he lae nineeenh and early wenieh cenuries. Like he curren period of globalizaion, China s economy dramaically opened up o world rade: oal rade as a share of GDP almos ripled during our sample period, an increase ha is roughly comparable wha China has experienced since Wih respec o he effecs ha rade had on wages, he mos imporan change may have occurred in response o World War I an even ha alered global rade paerns and had lasing effecs on Chinese rade. We show ha he price of Chinese expors rose, leading o a surge in Chinese expors. Wages and employmen rose for unskilled workers, which were uilized inensively in he producion of Chinese expors. We marshal new daa on he facor conen of Chinese rade o show ha China s expor boom in he firs hree decades was characerized by a rapid expansion 3

5 in he producion and sale of unskilled-inensive producs o he res of he world. In he second decade of he 20 h cenury, China s growh in expors of unskilledinensive manufacures, mining, and agriculural producs received an addiional boos in demand from World War I. Whereas he war disruped rade in many oher pars of he world, i led o an expansion in Chinese expors, creaing new markes for Chinese goods ha had previously been served by producers in belligeren counries. Employing uni value daa from Chinese rade saisics, we show ha he prices of key expors rose rapidly in response o his exogenous shock and coninued o rise even afer hosiliies ended. Expors coninued heir upward rajecory as China s producs were inegraed ino he global rade nework. Using new daa on wages for unskilled and skilled workers, we argue ha he growh in Chinese rade and he price shock of World War I largely accoun for he flaening ou of he skill premium in he 1910s and he subsequen 8 percen fall in he skill premium beween 1920 and 1928 (Figure 2). We develop a general equilibrium model of rade (based on facor endowmens) and show how a price shock affecs he skill premium. We hen simulae he model and compare he resuls o he observed daa on changes in he skill premium. We find ha our model can simulae he decline in he skill premium once a shock is aken ino accoun. In he nex secion, we provide some hisorical background on he growh in Chinese rade during he firs hree decades of he wenieh cenury. Secion III describes how we employ he heoreical predicions from facor endowmens models o conduc a facor conen analysis of Chinese rade. Secion IV describes our new rade daabase assembled from he publicaions of he Chinese Mariime Cusoms Service and our mehodology for esimaing he facor conen of rade. Secion V assesses he relaionship beween facor conen and expor growh, calculaes he 4

6 change in expor prices, and relaes our rade daa o changes in he skill premium in China during he firs hree decades of he wenieh cenury. Secion VI presens a general equilibrium model of rade, simulaes a price shock, and compares he resuls from he model o observed changes in he skill premium. The final secion discusses he implicaions of our findings and concludes. II. China s Firs 20 h -Cenury Trade Boom The firs hree decades of he wenieh cenury mark he period when China s rade wih he res of he world began o expand significanly. Unil he 1840s, China was largely a closed, agrarian economy; however, pressure from Grea Briain and oher foreign powers led China o open is economy o inernaional rade. The 1842 Treay of Nanking permied foreigners o rade wih he Chinese in five pors and sipulaed a general five percen ad valorem ariff on almos all goods leaving and enering China. Over he subsequen sixy years, China saw a gradual opening up of rade as China signed reaies wih foreign powers and addiional pors were allowed o ransac wih foreigners. 2 China s defea in he Sino-Japanese war in 1895 ushered in furher changes o Chinese producion and rade. The Treay of Shimonoseki allowed Japanese businesses o inves direcly in China and produce goods and services ha could be sold abroad as well as wihin China. Soon afer he reay was signed, his privilege was exended o oher foreign naions via mos-favored-naion agreemens. Foreign capial financed railroad, elecommunicaions, and shipping enerprises and spurred indusrializaion. 2 During his period, he larges impor commodiies were opium, coon exiles, and peroleum producs (kerosene, gasoline, ec.). Major expors were ea and silk. 5

7 By he early wenieh cenury, he number of ciies open o rade had climbed o 48 ciies. As pors opened up and foreigners were allowed o inves and rade, China ransiioned from a closed o an open economy. Wih no noable rade policy resricions in place, China hen experienced a susained growh in inernaional rade (Figure 1). Cheng (1956) esimaes ha beween 1900 and 1913 he oal value of rade grew wice as much as i had beween 1868 and In he firs 13 years of he wenieh cenury, he value of foreign expors nearly ripled, impors almos quadrupled, and he annual growh rae of rade averaged 7.4 percen. China s rade growh was faser han he world average in he firs hree decades of he wenieh cenury: is share of world rade increased from 1.5 percen around 1898 o 3.44 percen by By he early wenieh cenury, he Chinese economy was exploiing is comparaive advanage in unskilled manufacures. Indicaive of his growh was rade in coon exiles, which became one of he fases growing indusries over he subsequen decades. As we emphasize hroughou his paper, World War I had ransformaive effecs on he Chinese economy. I disruped rade in oher pars of he world and redireced i in he ways ha direcly benefied China, including a large increase in demand for is expors. Alhough China did no experience a dramaic acceleraion in he growh rae of expors during World War I, his fac disguises several imporan effecs he war had on Chinese expors and wages. Many counries experienced a conracion in rade during World War I (Glick and Taylor 2001). Trade growh for China, on he oher hand, was superior o he average counry during his period. Impors were disruped hroughou World War I, bu expors were only emporarily affeced. Afer declining for one year, expors resumed heir upward rajecory. Even during he war, hey coninued o grow a rend raher han below rend (Figure 1). In 6

8 he en years from 1917 o 1927, hey grew a 7 percen annually. Jus as imporan, he price of is expors began o rise during he war and coninued an upward rajecory hrough he 1920s. Chinese expors o counries like he Unied Saes and Japan increased rapidly. The war redireced rade beween oher counries o China, and allowed i o expand producion and furher specialize according o comparaive advanage. Mos of China s op expors (coal, minerals and mineral producs, raw coon and coon exiles, brisle, and edible oils) received a considerable boos from World War I; moreover, demand from he res of he world did no subside afer he war ended (see Table 1). For example, rade records indicae ha domesic yarn firms (coon exiles) sared o expor in 1913, and by he mid 1920s, had largely displaced impors (Figure 3). Brisle (one of China s op 10 expors) was used o make brushes for machines, guns, and cannons; by 1930, China was supplying 90% of he world s brisle (You, 1990). Anoher op 10 expor, edible oil, also received a considerable boos from World War I. Afer he oubreak of war, he oil-pressing indusry in Europe, swiched o he producion of miliary-relaed producs, leaving a huge gap in demand for edible oil in hese counries. China filled his gap by dramaically increasing is expors o European belligeren counries. III. Theoreical Framework To formulae esable predicions of he effecs of his rade boom on wages and (in Secion VI) simulae he effecs of a price shock on Chinese rade, we draw on insighs from facor endowmens models of rade. The Heckscher-Ohlin-Vanek (HOV) model predics ha when an economy ha is relaively well endowed wih unskilled labor opens up o foreign rade, i will specialize in producing unskilled labor inensive producs for expor. According o he Solper-Samuelson heorem, an 7

9 increase in expor prices will lead o an expansion in rade expands, and wages for unskilled workers will rise relaive o hose of skilled workers. Alhough HOV models and heir relaed heorems ofen absrac from realiy in heir parsimony (wo counries, wo goods, wo facors) and in heir heoreical assumpions (consan reurns o scale, perfec compeiion, idenical producion echnologies, free mobiliy of goods, ec.), hey are neverheless useful for framing how rade based on comparaive advanage affecs facor prices wihin counries and for drawing aenion o he winners and losers in rade. For example, policymakers have used he simple predicions of he 2x2x2 version of he models and he Solper Samueson heorem o explain how increased rade and globalizaion is impacing wages. They also have been used by economiss o consider he exen o which rade is driving increased wage inequaliy in he U.S. (Revenga, 1992; Lawrence and Slaugher, 1993; Leamer, 1998), and similarly, in predicing he consequences ha China s expanding bilaeral rade wih he U.S. will have on U.S. wages (Krugman, 2008; Lawrence, 2008). Despie heir heoreical elegance, esing he predicions of he Solper- Samuelson heorem is challenging. 3 Empirical researchers ofen resric heir aenion o simple versions of facor conen models so ha hey reain clearer heoreical predicions and avoid issues such as facor subsiuabiliy. There are neverheless idenificaion issues ha make esimaing he effecs of rade on wages using modern daa exremely challenging. Skill premiums oday may be driven by a variey of facors ha are difficul o disenangle, including echnology, migraion, and insiuional changes in labor markes. On he oher hand, here are several reasons why facor endowmens models may have more power in explaining he effec ha rade has on wages in hisorical 3 An excepion in he hisorical lieraure is O Rourke, Taylor, and Williamson (1996), which examines he relaionship beween commodiy and facor price convergence across a panel of eleven counries. 8

10 periods including our sudy of China during he firs hree decades of he 20 h cenury. 4 Firs, bulky sandardized commodiies, such as whea and mea, and simple manufacures, like coon goods, were he basis for he global growh in rade in he lae-nineeenh and early-wenieh cenuries (Findlay and O Rourke, 2003). In conras o he differeniaed rade in goods and services oday, differences in facor endowmens may be sufficien for explaining he movemen of raw maerials and simple manufacures across naional boarders (i.e., i may be unnecessary o appeal o newer rade models emphasizing produc differeniaion or he wihin-indusry effecs, such as Meliz (2003)). Second, here were fewer ariff and non-ariff barriers o rade, especially wih respec o agriculural goods and low-skilled manufacured goods wo imporan areas of expor for China during our sample period (Esevadeordal and Taylor, 2002). 5 Third, he effecs of echnological change operaed differenly on developing counries skill premia, like China s, during he firs global rade boom. In conras o oday, when we observe a considerable amoun of skill-biased echnological change, echnological innovaions in China in his earlier era were more of a complemen o han a subsiue for unskilled workers in he producion process. 6 For example, in a fas-growing Chinese expor secor like coon exiles, he inroducion of new machinery (ring spinning), which improved labor produciviy, did no displace he 4 For a review of facors driving skill premium oday in developing counries, see Goldberg and Pavcnik (2007). As noed here, here are likely large differences in how rade operaed in he earlier period of globalizaion in comparison o oday. 5 Much of he lieraure on he recen period of globalizaion has used evidence from changes in rade policy (i.e., ariff liberalizaion) o undersand he impac of rade on wage inequaliy; however, since rade policy is he oucome of poliics, i is an endogenous variable. Several recen sudies have exploied addiional cross-secional (indusry) and ime variaion in he daa in order o deal wih issues of causaliy (Hanson, 2007; Wei and Wu, 2001; Topalova, 2004; Goldberg and Pavcnik, 2005.) In his paper, we exploi an alernaive source of exogenous variaion o idenify he impac of rade on wages World War I which we argue is an exogenous shock for Chinese expors. 6 They were ofen embodied in machines ha could be impored from oher counries. Indeed, his may have been he moivaion for H-O o assume ha he same echnology exised for wo counries in heir model of rade. Since ranspor coss fell dramaically during he era in which hey wroe (and during 9

11 demand for unskilled labor in his indusry (Zhao and Chen, 1997). 7 By conras, he use of primarily uneducaed females grew rapidly in his indusry. Governmen esimaes sugges ha, beween 1912 and 1920, employmen in coon exiles in China grew by 32%, from 228,497 o 301,544, almos all of which was an expansion in unskilled labor (Minisry of Agriculure and Commerce, 1928, pp. 9-11). We would be he firs o acknowledge ha if echnological change is complemenary o unskilled labor, hen some of he observed decline in he skill premium may be due o changes in echnology. 8 I should be emphasized, however, ha he vas majoriy of China s expors during he firs hree decades (in paricular, agriculure, handicraf indusries, and mining) experienced lile echnological progress during our sample period (Wang, 1998; Wrigh, 1984). For example, brisle, one of he leading expors in he firs hree decades of he wenieh cenury, was a ypical labor-inensive handicraf; mos workers employed in is producion were uneducaed and from low-income peasan families (You, 1990). Many of he oher leading expors in our sample were agriculural and mining producs, and he producion of hese producs was unskilled inensive. I hus seems unlikely ha echnological change was he principal force ha drove he precipious decline in he skill premium. As a consequence, we will focus our aenion on accouning for he effecs arising from differences in facor endowmens. Fourh, he growh in he sock of educaed workers in China during he firs hree decades of he wenieh cenury was likely oo small o aler he skill premium par of our sample period), i would have seemed unrealisic o assume ha counries couldn simply have impored echnology (embodied in machines) from oher counries (Feensra and Taylor, 2008). 7 Chinese coon exile facories impored ring-spinning echnology from more indusrialized counries, which used unskilled labor more inensively han he older echnology of mule spinning. The ype of echnology employed also depended on he qualiy of coon used in producion. (Saxonhouse and Wrigh, 1984, 1987). 8 If skill-biased echnological change is concenraed in low-skilled secors, hen i could generae a decline in he skill premium (Leamer, 1998), which would be indisinguishable from a fall in he skill premium induced by a rade shock. 10

12 significanly. Enrollmen raes in secondary schools rose lae in he period and he sock of hese newly educaed workers was likely oo small o have much of an effec on he wages of skilled workers (Xiong, 1990). And as noed above, unlike oday, here appears o have been no surge in demand for skilled workers during his earlier era of globalizaion so ha skill complemenariy is less likely o explain he movemens in skill premium in China. 9 Fifh, alhough workers depared Europe in large numbers and wen o he Americas during he nineeenh and early wenieh cenuries, China s paricipaion in his wave of global migraion was much smaller. Roughly en million Chinese emigraed beween 1840 and 1920, which would ranslae ino an average of a lile less han one hundred housand per year (Ge, Cao, and Wu, 1993, pp.485-6). In relaion o he oal populaion of China, roughly 400 million during ha period, he emigraion would have had a negligible effec on wages in China. Hence, i seems reasonable o assume ha he effecs of emigraion on he wages of unskilled workers during our sample period is likely much more mued han in developing counries oday. Sixh, afer 1894, foreign direc invesmen was permied in China. Scholarship suggess ha FDI may have served as a caalys for China s indusrializaion. I was mos concenraed in he pars of he manufacuring indusry ha had a greaer reliance on advanced machinery. Overall, he oal increase in foreign capial flows during our sample period was small and he rae of growh in FDI was fairly seady hroughou he firs hree decades of he wenieh cenury (Hou, 1965). I exhibis no break or surge around World War I. FDI s effecs on he skill 9 This is a noable difference from wha is observed in developing counries oday, like Argenina, Brazil, Mexico, Chile, Colombia, Hong Kong, and India, where he share of skilled workers wihin indusries has increased dramaically over he pas wo decades (Robbins (1996), Sanchez-Paramo and Schady (2003), Aanasio, Goldberg and Pavcnik (2004), Hsieh and Woo (2005), and Kijima (2006). 11

13 premium are ambiguous: i may have increased he demand for skilled workers as i has oday or, as in he case of he role of echnology in early-wenieh-cenury China, been complemenary o unskilled labor. 10 Sevenh, unlike oday, when insiuions such as unions and minimum or sae wages can impac observed wage raes, here is no evidence ha he Chinese labor marke faced significan regulaion; hence he skill premium was no likely influenced by labor marke insiuions during he firs hree decades of he wenieh cenury. Finally, rade in inermediae producs (i.e. ousourcing, offshoring, and global produc sharing ) was insignifican in he earlier era of globalizaion. IV. Daa and Measuremen of he Facor Conen of Chinese Trade We now examine how he Chinese rade boom of he firs hree decades of he wenieh cenury affeced he skill premium in China. Figure 2 shows ha he skill premium rose during he firs decade of he wenieh cenury (when rade expanded, bu expor prices were relaively consan), bu hen flaened ou and declined as expor prices rose and he expor boom coninued virually uninerruped unil Alhough he ime series graph is broadly consisen wih he view ha he rapid growh in rade may have impaced he wages of skilled and unskilled workers in China, we subjec his hypohesis o more scruiny by considering wheher he facor conen of Chinese rade is consisen wih he predicions of facor endowmens models. Facor endowmen models in he spiri of HOV-SS predic ha, as China opened up o rade wih he res of he world and received a boos when expor prices rose (beginning around World War I), expors of goods ha use relaively more 10 Solper-Samuelson effecs assume ha labor and capial are mobile wihin a counry, bu immobile across borders. However, if we allow for FDI, Rybczinski s Theorem suggess no long-run impac on 12

14 unskilled labor (he abundan facor in China in comparison o skilled labor) in he producion process will increase. The demand for unskilled labor will rise as he economy expors more. China will also begin o impor more goods ha are produced wih relaively more skilled labor, hus reducing he domesic demand for skilled labor. As long as he supply curves for labor are no perfecly elasic, he shifs in demand for skilled and unskilled workers will cause he wages of skilled workers o fall relaive o unskilled workers. Hence, facor endowmen models predic ha China s rade boom will cause he skill premium o fall. A. Daa To examine he facor conen of Chinese rade, we assemble new deailed esimaes of expors and impors a he indusry level from China Mariime Cusoms (hereafer CMC ) rade publicaions. CMC was likely he only bureaucraic organizaion in China ha operaed wihou inerrupion (due o wars or funding shorages) from 1858 o Alhough i repored o he Chinese governmen, is op adminisraion as well as mid-level managers and echnocras were largely foreigners iniially Briish ciizens, bu laer on also Japanese and Americans. CMC s primary asks were collecing cusoms revenue and recording and publishing daa on foreign rade; however, i evenually expanded is operaions o include collecing revenues from domesic rade, adminisering he posal sysem, developing inland and coasal waerways, and represening China a inernaional fairs. CMC s geographical reach grew from jus foureen saions in he 1860s o nearly fify during he 1920s, covering no only he coasal regions bu also inland ciies. facor prices. 13

15 CMC published 160 volumes of deailed rade saisics, which span roughly 90 years of commercial ransacions ( ). The rade records were colleced a he por level and include informaion on he quaniies and he values of all commodiies passing hrough each reay por. If aggregaed, hey also provide a deailed picure of China s rade wih he res of he world. In comparison o oher economic or demographic daa on China during his period, he qualiy and deail of he CMC rade daa is excepional and rivals he rade publicaions of he advanced naions of he lae nineeenh cenury. 11 CMC rade saisics were primarily published a he por level. As a resul, he unis of measuremen and currency someimes varied across pors and over ime. We herefore sandardized he measuremen and currency unis and hen aggregaed he produc-level daa o he naional level. B. Measuremen To assess he influence of he expansion in rade on he skill premium in China, we examine he facor conen of Chinese expors. We keep our facor conen analysis simple and consider differences in producion based only on labor characerisics wheher workers were skilled or unskilled. Alhough his is clearly a simplificaion, i enables us o ake advanage of our deailed rade daa o learn somehing abou China during a period of Chinese hisory when lile informaion on oher firm or indusry characerisics exiss. Afer creaing a daabase of quaniies and values of all he raded commodiies using he CMC rade publicaions, we classified expors and impors based on economic aciviy and skill inensiy. To do his, we firs chose a sandard 11 Trade saisics repor ypes and desinaions of rade, so ha we are able o ensure ha rade is no double couned. Values repored are F.O.B. 14

16 classificaion sysem so ha we could sysemize he aggregaes of economic aciviy and hen measure facor conen by indusry group. This is especially imporan because our daabase covers Chinese rade for all he reay pors over hree decades, he nomenclaure of raded goods someimes changed, and he individual cusomhouses someimes colleced rade saisics using heir own naming sysems. Since he Sandard Indusrial Classificaion Sysem (SIC) was no adoped in major indusrial surveys or censuses, i has limied usefulness for he wide variey of commodiies in our Chinese rade daa. Insead, we use he Index of Occupaions and Indusries from he 1950 U.S. Census of Populaion ( IND1950 ). The basic conen of he occupaional and indusrial classificaion was largely derived from earlier censuses, in paricular, he 1940 Census. Since IND1950 is somewha rerospecive in design, i provides a consisen se of indusry codes ha is broad enough o capure he rade being conduced by China beween 1903 and We nex classified indusries according o skill inensiy. Since here is no agreed upon mehodology for deermining he facor conen of producs or indusries, empirical sudies use a variey of approaches o proxy facor conen. One approach is o rank indusries according o average wages. If workers are paid heir marginal producs (as would prevail in compeiive markes for facors and goods), hen, on average, higher paying indusries ough o reflec higher average produciviy or skill. Anoher way o proxy for skill inensiy is o rank indusries by average educaion levels. A hird approach is o calculae he share of producion workers (relaive o non-producion workers) in each indusry. Because U.S. hisorical census daa offer broad indusry coverage and deailed informaion on educaion and wages, our saring poin for classifying skill inensiy for Chinese indusries was o creae benchmarks based on he 1940 U.S. Census. (As a 15

17 robusness check, we consider an alernaive measure of skill inensiy a he indusry level, based on a more limied Chinese survey, laer in he paper.) The 1940 U.S. Census is he firs census ha provided informaion on an individual s educaion and earnings. The 1940 Census records each individual s highes level of educaional aainmen, ranging from no educaion o five or more years of pos-secondary educaion. In our analysis, we classify workers wih nine or more years of educaion as skilled workers. We coun he numbers of skilled and unskilled employees in he 1940 Census for each indusry (using he IND1950 classificaion described above) and hen calculae he fracion of workers in each indusry ha had nine or more years of educaion. The 1940 Census also records each individual s annual wage, allowing us o aggregae hese daa and obain average indusry wages (again based on IND1950 indusry classificaion). Hence, using U.S. Census daa, we are able o obain informaion on skill inensiy a he indusry level based on (1) educaion and (2) wages. We repor indusry rankings using hese wo merics in Tables 2 and 3. The Spearman rank correlaion coefficien for he wo merics is 0.746, and is significanly differen from 0. V. Analyzing he Effecs of he Chinese Trade Boom on Wages A. Facor Conen Analysis A firs es in he spiri of facor endowmens models is o examine wheher oal expors were becoming more unskilled-inensive in heir composiion over he course of our sample period. Using he daa on indusry averages for wages and educaion, we classify he indusrial secors ino wo broad groups, unskilled and 16

18 skilled. We denoe indusries where he fracion of workers wih nine or more years of educaion exceeded 0.48 as skilled indusries. We divide he daa a his value since he wo indusries above and below his cuoff seemed mos dissimilar in erms of labor force characerisics (moor vehicles and moor vehicle equipmen versus glass and glass producs). In a similar way, we also divided he indusry daa ino skilled and unskilled using log wages. Indusries wih log wage values greaer han are classified as skilled, which pus meal mining and poery producing as he wo indusries on he dividing line. 12 Figures 4 and 5 display he composiion of expors and impors in erms of facor conen. In 1903, when he magniude of foreign rade was fairly small and China was relaively closed, mos of is impors and expors were composed of unskilled-inensive producs. However, as Chinese rade grew in imporance over he nex 25 years, we see significan movemens in he raios for boh expors and impors. Using eiher wages or educaion as our measure of skill, expors became more unskilled-inensive over he enire sample period rising from abou 0.92 o 0.99 for he educaion raio and from roughly 0.8 o roughly 0.9 for he wage raio (Figure 4). In conras, he share of unskilled impors declines subsanially. Using he measure based on educaion, he share of unskilled expors falls form 0.88 o 0.75 (Figure 4, Panel A). In he same vein, he fracion of impors ha are skillinensive increases from 0.11 o 0.25 over he sample period (Figure 6). The rend owards more unskilled inensive expors and more skill-inensive impors is paricularly pronounced afer Anoher way of assessing he general facor conen of rade is o examine he deailed indusry daa. In Figures 6-9, we display he value of expors and impors for each indusry on he y-axis and is corresponding skill inensiy on he x-axis for four 12 The resuls repored laer in he paper do no appear ha sensiive o changing hese cuoffs. 17

19 years: 1903, 1913, 1919, and Skill inensiy increases as we move in a righward direcion along he x-axis. We presen his evidence for skill inensiy based on educaion (Panel A of Figures 6-9) and log wages (Panel B of Figures 6-9). The figures show ha expors are largely clusered a he lower levels of skill inensiy whereas impors dominae he highes values of skill inensiy. These characerisics of he figures are even more eviden by These resuls consiue srong evidence ha rade was fundamenally responding in ways ha are consisen wih facor endowmens models of rade. Even if we found no evidence ha he ha unskilled-inensive expors were rising over our sample period, he facor conen daa could sill be consisen wih he predicions of facor endowmens models if i were rue ha, overall, Chinese expors grew faser for unskilled-inensive indusries han for skill-inensive indusries. This would indicae ha he expansion in expors shown in Figure 1 was driven by unskilled-inensive expors. (Similarly, skill-inensive impors should grow faser han hose for unskilled indusries.) To es his alernaive predicion, we divide he daa ino en indusry groups based on skill inensiy and hen compue he growh raes for each decile. Tracking he growh in expors of individual indusries migh be preferable, bu i is complicaed by he fac ha some indusries lack daa for our whole sample period. Using indusry groupings allows us o examine indusries of similar skill inensiy and follow hem over he enire sample period. We weigh he deciles by heir share of he oal value of expors, and hen plo he growh rae of expors for each group relaive o is skill inensiy. Figures graph he average annual growh raes for expors and impors from 1903 o 1928 where indusry groups are ordered by skill inensiy (using eiher 18

20 he wage or educaion classificaions as indicaed on he graph). The figures show ha he fases growing deciles for expors ended o be hose wih he lowes skill inensiy. Indeed, since such a large preponderance of expors use unskilled labor inensively, he visual impac of he graph is diminished since here is no way o include all he zero values of indusries (no expored) ha would be more skill inensive. On he oher hand, he fases growing deciles for impors ended o be he mos skill inensive. B. Robusness Check In our empirical analysis, we have implicily drawn on an assumpion of he HOV model ha echnology is he same across counries so ha we could derive skill inensiies for Chinese indusries using 1940 U.S. census benchmarks. However, because he assumpion of common echnology may no have held in pracice and because our U.S. skill-inensiy benchmarks are based on daa from 1940, we explore wheher he facor conen analysis is robus o an alernaive classificaion scheme. To carry ou our robusness check, i is necessary o idenify an alernaive survey for deriving skill inensiy a he indusry level. Unforunaely, survey daa a he individual, occupaional, or indusry level for China during he firs hree decades of he wenieh cenury are very scarce. The 1928 Shanghai Census is he earlies survey wih sufficien deail o derive esimaes of skill inensiy a he indusry level. I was adminisered by he Bureau of Indusry, Agriculure, and Commerce of he Greaer Shanghai Municipaliy and included informaion on facory names and addresses, ownership, capializaion, number of workers, wages, raw maerials, and power uilizaion. The survey records only he highes and lowes wage raes by indusry; afer closer inspecion, we 19

21 deermined ha he minimum and maximum wages in each indusry were insufficien o generae reliable average indusry wage raes. Insead, using daa found in he survey, we compued he capial-labor raio o evaluae skill inensiy of indusries under he assumpion ha indusries which use more labor are, on average, characerized by lower-skilled workers. To compue his alernaive measure of facory inensiy, we firs classified he raded goods using he classificaion sysem adoped in he 1928 Shanghai Survey, and hen calculaed he capial-labor raio by dividing he oal physical capial (value of physical capial sock) by he number of workers in each indusry. We hen ranked he raded goods indusries according o heir capial-labor raios (Table 4). Figures 12 and 13 show ha, using he 1928 Shanghai survey produces resuls ha are similar o hose based on he U.S. skill-inensiy benchmarks. Growh raes for Chinese expors were highes for unskilled indusry deciles and growh raes for impors were highes for he skilled deciles. C. Evidence on he Price of Expors Facor endowmen heories sugges ha changes in produc prices are he mechanism ha alers rade flows and facor prices. Therefore, a necessary condiion for rade o impac he skill premium is a change in he price of expors. We have hus far shown ha China became more specialized in producing and exporing unskilled inensive commodiies. Assuming he marginal produc of workers has no changed, if expor prices rise subsanially, he skill premium in China will evenually fall, since Chinese expors use relaively more unskilled labor in heir producion. 20

22 Hsiao s (1974) expor price index shows ha expor prices grew markedly afer 1913 (Figure 14). 13 Consisen wih oher global sudies of rade, prices of raded goods rose during World War I. Moreover, i appears ha in China s case, he posiive shock in demand was large enough o significanly move prices upward afer a decade of lile change. Afer a brief cessaion afer he war, prices resumed heir upward rajecory. During he period when he skill premium was flaening ou and falling, expor prices roughly doubled. Rising sales and prices led o growing revenue for exporing firms, providing considerable scope for raising he wages of unskilled workers ha were used inensively in heir producion. We used he uni value daa conained in our daabase o compue he growh raes in prices for China s en mos imporan expors (based on value). Figure 15 shows posiive average annual growh raes in uni values for hese major expors over he period 1903 o For nine of he en chief expors, he growh raes were faser afer 1914 (Panel B). In addiion, he commodiies shown in his figure ha used unskilled labor more inensively (agriculural goods and coon yarn) experienced paricularly srong raes of growh in heir prices. Figure 16 shows ha prices in coon yarn grew by 275 percen afer Coon spinning was widely considered a ypical unskilled-inensive manufacuring indusry. China had been a large imporer of coon yarn, bu by he beginning in he wenieh cenury, a domesic coon exile indusry began o compee wih foreign producs. The indusry grew rapidly, and by he mid-1920s, coon-spinning expors exceeded impors. As he coon spinning indusry expanded, i drew in large numbers of unskilled workers. Real wages in coon exiles increased by more han 50% during he war, remained a hose levels afer he war ended, and hen coninued an upward 13 The expor prices shown in Figure 14 are he prices in silver. The expor prices relaive o he price of gold show similar qualiaive resuls: he price increased by 28 percen from 1903 o 1913, bu by

23 rajecory ino he 1920s (Liu 1936). Alhough addiional indusry-level wage daa are quie scarce, we were able o obain figures for real wages for wo oher key expors. Real wages in he silk indusry grew by more han 150% during he war and hen, afer flaening ou a he conclusion of he war, coninued heir upward rajecory. Similarly, real wages for coal workers (based on Kailun, one of he larges mines in China a ha ime) grew by more han 25% during he 1920s. 14 These daa sugges ha wages were rising in he expor secor in response o higher produc prices. Alhough i is impossible o rule ou alernaive explanaions, he movemens in he skill premium daa seem mos consisen wih a rade shock raher han eiher a dramaic rise in he quaniy of skilled workers in he 1920s or very rapid echnological change. The limied daa ha are available on Chinese educaion raes sugges ha he expansion of skilled workers due o increased opporuniies was small and likely came oo lae in our sample period o accoun for he observed decline in he skill premium. As noed earlier, some echnological advances in China appear o have been ongoing in indusries like coon exiles, and hey ended o raise he demand for unskilled workers raher han displace hem. I is herefore possible ha some of he boos in rade we observe during our sample period is associaed wih echnological change in manufacuring, which perhaps resuled from he foreign direc invesmen ino China from Japan and oher counries a he beginning of he 20 h cenury. Neverheless, in coon exiles and oher mechanized manufacuring indusries in China, he pace of echnological change was incremenal. There were no sudden surges ha occurred in he 1910s and 1920s ha would be consisen wih he raher abrup reversal in he wage premium over hese decades. If anyhing, he pace of echnological change was slower during he war period due o he difficuly of percen from 1914 o 1919, and by 60 percen from 1914 o

24 imporing new machinery from wesern counries engaged in comba. The reversal in he upward rend and flaening ou of he skill premium ha occurred during his decade would herefore be inconsisen wih echnology being he main driver. Moreover, in he vas majoriy of oher secors of he economy, paricularly agriculure, handicraf indusries, and mining, here was lile echnological progress during he firs hree decades of he wenieh cenury. VI. Model and Simulaion A. Benchmark model To beer undersand rade s conribuion o he change in he skill premium in he 1910s and 1920s, we now model a price shock o Chinese expors in a generalequilibrium, facor-endowmens framework. We hen simulae his model using daa on he Chinese economy o examine wheher our empirical findings are plausible and consisen wih heory. We firs consider a closed economy model as a benchmark. In he subsequen subsecion, we exend he model o an open economy subjec o price shocks. Suppose here are wo inermediae goods and one final good. Prices of inermediae goods are denoed as 1 p and 2 p and f p is he price of he final good. We se he second inermediae good as a numeraire and hence normalize 2 p o be 1. Oupu for he final good, f Y, is produced using a sandard CES specificaion where ε 1 ε 1 ε f 1 ε 2 ε ε 1 [ γ( ) (1 γ)( ) ] Y = Y + Y (1) where 1 Y and 2 Y are inermediae goods, and ε is he elasiciy of subsiuion beween wo inermediae goods. By profi maximizaion, 14 We were unable o locae daa for real wages in he coal indusry prior o 1919 or for oher key 23

25 Y Y γ 1 ( ) 1 ε = γ p. (2) Thus we have p f 1 ε p ε ε γ ε = ( γ ( ) + (1 ) ). (3) The producion for inermediae good 1 uses unskilled labor, 1 L, and land, T, wih he shares δ and 1 δ respecively. Land has a fixed supply, so we normalize he supply of land o be 1. Hence we specify he producion funcion of inermediae good 1 as he following: Y = A ( L ) δ, (4) where 0< δ < 1. For simpliciy, we assume he echnology used o produce inermediae good 1, 1 A, is consan. Assuming labor markes are perfecly compeiive, he wage for unskilled labor will be equal o is average produc: w = p A ( L ) δ. (5) On he oher hand, he producion of inermediae good 2 uses boh skilled and unskilled labor and is specified as a consan reurns o scale producion funcion: Y = A ( H ) ( L ), (6) α 2 1 α where 2 H and 2 L are skilled and unskilled labor, respecively. The shares of hese wo inpus are α and 1 α respecively. Again assuming perfecly compeiive labor markes, he wage for skilled labor, marginal producs: H w, and ha of unskilled labor, L,2 w are heir w = α A ( H ) ( L ), (7) H 2 α α α w = (1 α ) A ( H ) ( L ). (8) L,2 2 α 2 expor indusries so ha we could carry ou a more deailed saisical analysis in he cross secion. 24

26 Since unskilled labor can move freely among all he secors, wages for unskilled labor will equalize across all he secors of he economy so ha w = w = w. (9) L 1 L,2 Hence, we define he skill premium as s H 2 w α L = =. (10) L w 1 α H We assume ha he echnology used for he producion of inermediae good 1 grows slower han ha in inermediae good 2. We make his assumpion o characerize he fac ha he produciviy of he indusrial secor improves rapidly hrough modernizaion (perhaps even hrough he imporaion of foreign echnologies). Wihou losing generaliy, we simplify he model by assuming ha A 1 is consan over ime, while A 2 grows a a posiive rae of g, i.e. 2 2 A = 1 (1 + + g) A. (11) In our model, here are wo ypes of consumers ha opimize over heir lifeimes. Consumer i, i { L, H}, maximizes lifeime uiliy as specified by a CRRA uiliy funcion: = 0 i 1 σ ( c ) 1 β. (12) 1 σ Consumers can borrow and lend a ineres rae, r, in any period, which is exogenous o our model. Therefore, he budge consrain of consumer i, i { L, H}, is: pc w (1 ) (1 r) i i. (13) = 0 + r = 0 + To opimize uiliy, he consumer hus maximizes equaion (12) subjec o (13). 25

27 We also assume he oal supply of skilled workers is fixed a H and he oal supply of unskilled labor is fixed a L so ha he marke-clearing condiions for he labor marke are: H = H, (14) L = L + L. (15) 1 2 We define a compeiive equilibrium as follows: Definiion of Compeiive Equilibrium: A compeiive equilibrium consiss of L H f 1 2 consumpion { c, c }, producion of each good { Y, Y, Y }, labor supply { H, L }, labor demand { 1, 2 L H L L, H }, wages { w, w }, prices of final good and wo inermediae goods f 1 2 {,, } p p p, and a law of moion g for he sae variable, echnology, such ha: (1) consumpion is he soluion o he uiliy maximizaion problem; (2) labor demand and wages saisfy equaions (5), (7), (8) and (9); (3) labor supply is fixed; (4) labor supply and labor demand saisfy equaions (14) and (15); (5) he law of moion g saisfies equaion (11); (6) he price of inermediae good saisfies equaions (2) and (9); (7) he price of final good saisfies equaion (3); and (8) he producion of each good saisfies equaions (1), (4), and (6). B. An Open Economy Model wih Price Shocks We now exend he model o describe an open economy wih exogenous rade shocks. Trade shocks are modeled as exogenous increases in he prices of he final good. Price shocks induce he home counry o impor more inermediae goods from 26

28 abroad o produce and expor more of he final good. To relae he open economy version of he model o our empirical analysis of China, we assume he home counry is abundan in unskilled labor and scarce in skilled labor. Following facor endowmen heories, we assume he home counry herefore has a comparaive advanage in producing inermediae good 1 and a comparaive disadvanage in producing inermediae good 2. I will herefore impor inermediae good 2 in order o raise he producion capaciy. Based on his framework, he marke for he final good can be described by: ε 1 ε 1 ε f 1 ε 2 ε ε 1 ( γ( ) (1 γ)( ) ) Y + EX = Y + Y + IM, (16) where expored, f Y is domesic consumpion of he final good, EX is he amoun of final good 2 Y is domesic producion of inermediae good 2, and of inermediae good 2 impored. IM is he quaniy We assume ha he home counry is a small economy and is hus a price aker in he marke for boh he final good and inermediae good 2. Therefore, f p and 2 p are exogenous and deermined by global supply and demand. Since we have no modeled he capial accoun, we assume balance of rade for each period so ha: p EX = p IM. (17) f 2 C. Analysis and Simulaion of he Skill Premium We now examine how he skill premium evolves in he wo versions of he model. In he benchmark closed-economy model, since he oal facor produciviy (TFP) of inermediae good 2 grows over ime while he TFP of inermediae good 1 is consan, unskilled labor will migrae ou of he producion of inermediae good 1 and ino he producion of inermediae good 2 unil he asympoic seady sae is reached. 27

29 As described by equaion (10), he skill premium will herefore rise unil i becomes fla afer reaching he asympoic seady sae. If we concepualize inermediae good 1 as describing producion in he agriculural secor, and inermediae good 2 as describing he indusrial secor, he benchmark model describes he rising skill premium ha is ypically associaed wih indusrializaion and faser TFP growh in he indusrial secor. In he open economy model, he skill premium evolves differenly. A persisen increase in he price of he final good induces he home counry o produce and expor more of he final good, using inermediae goods produced domesically (by boh skilled and unskilled labor) as well as some ha are impored from abroad. Since he home counry has a comparaive advanage in producing inermediae good 1, i will impor inermediae good 2. In conras o he benchmark model, he open economy model will slow down he migraion of unskilled labor from inermediae good 1 o inermediae good 2 since unskilled labor is needed in order o produce more of he final good for he global marke. If he price shocks are sufficienly large and persisen, unskilled labor will move back ino he producion of inermediae good 1, hus driving down he home counry s skill premium. The producion of inermediae good 2 in he home counry shrinks, and he home counry impors inermediae good 2 o produce he final good. We can show his by solving he close form of he model. Equaion (11) shows ha he skill premium depends solely on 2 L, he amoun of unskilled labor used in producing inermediae good 2. Hence, we need o solve he relaionship beween price shocks and unskilled employmen in he secor of inermediae good 2. Assuming profi maximizaion and he producion funcion for he final good shown in equaion (16): we obain he following opimal condiion 28