Made in China Matters: Integration of the Global Labor Market and Global Labor Share Decline

Transcription

1 Made in China Matters: Integration of the Global Labor Market and Global Labor Share Decline Li Daokui 1 and Xu Xiang 2 Modern macro research expends great effort to identify the driving force of increasing global inequality. Of all the factors influencing global decline of labor share and increasing inequality, the impact of developing countries is largely ignored. In this article we show that the integration of the global labor market, specifically, Chinese labor's return to the global labor market, plays a key role in the global labor share decline since the late 1970s. Several key institutional changes in China, including Reform and Opening Up and China joining the WTO, accelerated the so-called Great Doubling of the global labor market. We use regression evidence to show that the integration of Chinese labor is as important to global labor share decline as the global decrease in the relative price of investment goods and structural changes in different economies. We also build a two-country dual economic model to explain the simultaneous decline of labor shares in labor-intensive and capital-intensive countries. Our results have implications for global rebalancing and labor market policies of developed and developing countries and shed new light on the role of China in the world economy. 1. Introduction What s the driving force behind global labor share decline since the late 1970s? In this article we address this question from the perspective of international labor market integration. We explain the decline of the labor share as a result of the increasing trade volume with China. With more imports of labor-intensive products from China, the labor share of China s trade partners faces downward pressure. For advanced countries, increased 1 Tsinghua University 2 Central University of Finance and Economics, correspondence: seanxuxiang@126.com

2 exports of capital-intensive goods further increases capital shares and intensifies global inequality. Most existing literature emphasizes increasing capital-output ratio and efficiency gains in capital-producing sectors to be the center of global labor share movements. While signifying our agreement with a nod, we would like to emphasize the effect of global labor market integration, especially China re-entering global labor market since the late 1970s. Global labor integration's influence on labor share movements is hardly mentioned in existing literature on labor share determinants. In the last 30 years, institutional changes and reforms have broken down the labor market barriers and trade barriers in developing countries. Meanwhile, the size of the global labor supply doubled as China, India, and ex-soviet bloc nations joined the global economy through international trade (Freeman, 2006), This will definitely have a significant impact on global factor shares. Classical trade theory predicts upward labor share in labor-intensive countries and downward labor share in capital-intensive countries, but in reality labor share also declines in labor-intensive countries like China, India and some Southeast Asia?? countries. We believe this inconsistency of theory and reality is due to lack of concerns over the economic structural changes of developing countries, especially those which are higher engaged in international trade. We organize the paper into three parts. We start by documenting the integration of Chinese labor in the global labor market from 1978 and 2013, and focus on the evolution of institutional endowments in the last 35 years. Labor migration accounts for 52.6% of new labor in urban areas and was accelerated by the institutional changes like Hukou reform and China joining the WTO. Export producers favor migrant workers as they require lower wages and fewer social security obligations, and the large pool of rural surplus labor reserve allowed China to sustain its competitive advantages in international trade in the last 30 years and also affects the foreign labor market. Next, we develop a two-country dual economic model that relates labor share declines to the surplus labor reserve in the agricultural sector in China. With this model, we are able to capture labor share declines in both labor-intensive and capital-intensive countries due to the dual economic features of labor-intensive countries. In our model, capital flows into labor-intensive countries when rural labor migrates from the agricultural sector to the modern sector and especially to the processing industry we define. For labor-intensive countries, wages grow slower than output, as the

3 wage requirements of migrant workers are lower than their marginal productivity, resulting in labor share decline. For capital-intensive countries, lower domestic labor demand and the growth in capital-intensive industries raise capital share and reduce labor share. We are able to identify key factors of global labor share determination using this model. We then use panel regression on a country-specific sample to test our theoretical results. Our data comes from various datasets including the KN labor share dataset established by Neiman and Karabarbonious, WDI dataset, WTO statistics database and the yearbooks of our sample countries. Our empirical analysis supports our view that trading with China reduces a country s labor share. According to the regression results, more than 60% of global labor share decline from 1995 to 2011 is due to the growth of China s international trade and labor migration of China in the global labor market. We have also tested other explanations of global labor share. Relative price of investment, capital formation share of total output and structural changes are identified as key explanation variables for the global labor share. We also test the effect of China joining the WTO and this institutional change has amplified the effect of China s increasing trade on global labor share. Our contribution to existing literature on labor shares is twofold. First, we use a two-country dual-economy model to analyze global labor share movements. We provide a simple and straightforward theory to explain labor share decline in both labor-intensive and capital-intensive countries. This framework is the first to consider the influence of economic restructuring of developing countries in global labor share determination. Our framework can be expanded to consider labor market frictions, different types of technology improvements, terms of trade and price markups in the determination of global labor share. Our second contribution is our quantitative analysis on the influence of global labor share determinants, and especially trade with China. We use panel regression to identify key explanatory factors of global labor share decline and test the effect of key institutional changes. Our results support the view that economic factors in the context of economic and political institutions play a very important role in labor share determination and inequality dynamics. Our findings have implications for the future and policy making in both developing and developed countries. First, the labor share decline in the last 30 years was due to the fast growth of international trade with developing countries. Globalization speeds up this process and the labor

4 forces in capital-intensive countries face fierce competition from abroad. On the other hand, the labor force in labor-intensive countries does not enjoy all the benefits of being integrated into the global market. Actually, they only get a small part, as their wages are much lower than their marginal productivity. Labor market institutions in labor-intensive countries like the rural-urban separation in China make it possible for capital owners to have more bargaining power in labor negotiation and enjoy most benefits from international labor integration. The importance of institutions should be emphasized in future policy making related to labor market and wealth redistribution. Next, our findings shed new light on the future movements in the global labor market and factor shares. According to our findings and work of other scholars on the Chinese labor market, labor migration in China is close to its end. However, this doesn t mean the end of global labor market integration and labor share decline. The rise of manufacturing exports in Southeastern Asia and India is fueled by the large amount of surplus labor in these countries. The engine of global labor market integration might be shifting to these countries and cause further global labor share decline. The article is organized as follows. Section II relates our work to the existing literature. In Section III we introduce how China is integrated into the global labor market and examine its implications. In Section IV we build a two-country dual economy model to illustrate labor share determination in both labor-intensive and capital-intensive countries. In Section V we present our regression results on a cross-country sample ( ) and test existing theories on labor share determinants. Section VI concludes. 2. Related literature Most of existing literature can be classified into four categories. The first is research on labor share, which focuses on documenting and explaining global labor share decline. Research on increasing capital shares is also reviewed. The second category of literature mainly investigates global labor market integration, especially the role China plays. The third category examines how domestic institutional changes affect global labor market. The last category related to our research is research on the economic development of dual economies. 2.A Literature on global labor share decline

5 Ever since the work of Kaldor (1957), stability of factors shares has been broadly used in macroeconomic models including Cobb-Douglas production models and growth models. The stability of labor shares of OECD countries has been challenged since the late 1990s, when Blanchard (1997) discovered the properties of the medium run, a period of transition from short-term fluctuations to long-term growth and documented declining labor shares of continental Europe countries since the early 1980s. Krueger (1999), Acemoglu (2002), Blanchard and Giavazzi (2003), Jones (2003), and Bentolila and Saint-Paul (2003) focus on the labor shares of OECD countries and also discover large declines since 1970s. Harrison (2002) and Rodriguez and Jayadev (2010) extend to non-oecd countries and find that labor share decline also existed in developing countries. Work of Bai and Qian (2009a, 2009b) and Li, Liu and Wang (2009) discover significant labor share decline in China since the 1990s. Hofman (2001) finds that the labor shares of Latin American countries kept decreasing from the 1950s to the 1990s. Blanchard and Giavazzi (2003) systematically study how regulation and deregulation in goods and the labor market affect wages, unemployment, and labor share. According to their research, significant decline of the labor share of Western Europe in the 1980s was mainly due to a decrease in the bargaining power of workers and heightened labor market deregulation. Karabarbounis and Neiman (2014) document a 5 percent labor share decline in global corporate labor share from 1975 to 2012, based on country-specific data that includes the labor share data of more than 110 countries. Corporate labor share is used to circumvent measurement difficulties in overall labor share. As a result, they find that 42 of the 59 countries with at least 15 years of records exhibited downward trends in their labor shares. In addition, more than 90% of the labor share decline reflects within-industry declines rather than inter-industry. They believe this result rules out the possibility that increasing trade integration of China or globalization plays a key role in global decline of labor share. Some recent research expanded the labor share literature. Piketty and Zucman (2014) focus on aggregate wealth-to-income ratios and capital-output ratios in the top eight developed countries from 1970 to They attribute global labor share decline to the long-run recovery of asset price and slowdown of productivity and population growth. Acemoglu and Robinson (2014) argue that the role of political and economic institutions is ignored in the work of Marx, Ricardo and Piketty on general laws of capitalism and use economic factors in the context of economic and political institutions to explain histories of inequality in South Africa and

6 Sweden. Bridgman (2014) nets out capital depreciation and production taxes in gross labor share and finds that net labor share does not decline as much as gross labor share in the U.S. and other large countries. We believe the following four kinds of factors have led to global labor share decline in the last 35 years. First, decline in the relative price of investment. Karabarbounis and Neiman (2014) estimate that the decline in the relative price of investment explains roughly half of the global labor share decline. Second, structural changes of economies. Work of Gollin (2002) emphasized that the differences in labor shares between countries mainly derive from within-industry labor shares. Li, Liu and Wang (2009) build a dual-economy model to illustrate labor share movements in the industrialization process of developing countries. Their empirical results support the view that labor share in economic development seems to follow a U-shaped curve, the lowest point of which is USD 6000 per capita PPP (2000 constant). Third, technology. In Hicks (1932) theory, labor share of output, or labor share divided by capital share, is determined by three factors: capital-labor ratio, factor-augmenting technology and elasticity of substitution. A majority of existing estimates indicate a less than one short-run elasticity of substitution (Chirinko, Fazzari and Meyer, 1999, Krusell, Ohanian, Rios-Rull, and Violante, 2000, Antras, 2004, Klump, McAdam and Willman, 2007, Oberfield and Raval, 2014). However, most recent research has suggests the long run elasticity of substitution could be higher than one (Acemoglu, 2002, Karabarbounis and Neiman, 2014), and if this is right, developments in capital-augmenting technology would have significantly influenced global share. The last but not least factor is the integration of the global labor market, and this is the factor we want to emphasize in this paper. The expansion of the global labor force directly reduced the capital-labor ratio and the labor share even in the absence of strong assumption on the elasticity of Substitution. By documenting the influence of China on the global labor market in the last 35 years we want to shed new light on understanding global labor share movements. 2.B Literature on the economic development of dual economies According to the Heckscher-Ohlin and Stolper-Samuelson theories, international trade integration increases abundant factor share. However, China, as the relatively labor-abundant economy, also experiences labor share decline. Corporate labor share of China decreased from 44.6% in 1992 (the earliest data we can get) to 36.6% in 2009, and the overall labor share of China decreased from 59.3% in 1992 to 49% in While we

7 adhere to the H-O and S-S theories, we have to change the assumptions for China to fit the two-country model. China had the largest rural surplus labor in the world when Chinese government announced the reform and opening-up policy in In the following 35 years this big surplus labor pool has fueled Chinese economic growth by keeping wage rates at relatively low levels and by increasing corporate profits, and this is in line with the dual economy theory of economic development. Work of Lewis (1954), Rains and Fei (1961), Jorgenson (1967) and Todaro (1970) provide a new method to study economic development: the dual economy model. Lewis set up the basic dual economy model with two sectors: the traditional agricultural sector and the modern industrial sector. In the agricultural sector marginal productivity is either zero or negative and is not related to wages. In the Lewis model, labor force in the agricultural sector works at a very low wage rate determined by economic institutions. The wage rate of the modern sector is slightly higher than that of the traditional sector, and in that sense this economy has unlimited supplies of labor. In the Lewis model, the key factor of economic development is capital accumulation. The modern sector keeps absorbing surplus labor from the traditional sector as the supply of workers through capital and labor supply is no longer unlimited. This transition ends when surplus labor is exhausted and marginal productivity of both sectors converge??. The dual economic structure disappears and the economy enters the modern economy stage. Rains and Fei emphasize the importance of surplus production from agriculture and further refine the labor transfer process. Jorgenson builds a neoclassical model to study the relationship between dual economic structure and economic development. Work of Todaro focuses on migration from rural areas to cities and his conclusion is that the gap between municipal and rural labor income decides the scale of migration. He also discovers that in some undeveloped countries where marginal productivity in agriculture is positive and the unemployment rate in cities are at fair levels, migration from rural areas still takes place and is accelerating. In this article we build a simple two-country trade model to explain labor share decline in both trade partners. Our key assumption is that the labor-intensive country has dual economy features, and its labor income is not decided by marginal productivity. We observe labor share decline in both countries as trade grows. This model is more suitable for explaining the labor share decline in the last years during which China s trade

8 with the rest of the world has increased significantly. 2.C Literature on the global labor market Our work is closely related to the literature on the effects of labor from China, India, and the ex-soviet bloc joining the global labor market, which nearly doubles the global labor force (Freeman, 2006, Roach, 2004). "The great doubling reduced the ratio of capital to labor by 39% as of 2011 and completely changed the supply-demand relationship in the global labor market. Trade played a key role in this process, according to Fenestra (1997). There are two reasons why we want to emphasize the case of China. First, China contributes the most to global labor market integration. Second, exports from China have had the most profound influence on the global labor market compared with other countries over the last 30 years. Autor, Dorn and Hanson (2012) focus on how rising Chinese import competition between 1990 and 2007 affected the U.S. local labor market. They found that rising imports have caused higher unemployment, lower labor force participation, and reduced wages in local labor markets that further result in declining labor share. Advanced countries like the U.S. are not the only ones facing grim labor market competition. The labor share of Mexico decreased from 38.3% in 1993 to 27.6% in 2011, a 10.7% decline in 18 years, and a big part of this decline is due to "made in Mexico" being replaced by made in China. We have observed a reversal of this trend since the late 2000s and will provide more details in the fifth section of this article. 3. Integration of Chinese labor in the global labor market ( ) The integration of Chinese labor in the global labor market is a major driving force of China s fast economic growth in the last 35 years. In 1978 when the reform and opening-up policies were first implemented, China had an urban labor force of 95 million. In 2013, this number had risen to 380 million. More than 150 million or 52.6% of the new urban labor force are immigrants. In the same period (1978 to 2013), the European and American labor force increased by 26 million and 36 million, respectively. In this section we document China's labor market growth and specifically labor migration from 1978 to 2013 and study how institutional changes accelerated the labor migration process. We also study how China s labor

9 migration affects the foreign labor market through trade. 3.A Labor market structure of China: before and after 1978 As late as 1977, China was an agricultural country basically separated from the global market: 82.5% of China s population and 76.8% of China s labor force were farmers in rural areas. By 2013, the rural share of China s population and labor force had declined to 46.3% and 49.1% respectively, and the decline was mainly due to the reform and opening-up polices carried out since the third plenum held in The reform and opening-up had two stages. In the first stage from the late 1970s and early 1980s, decollectivization of agriculture, the opening up of the country to foreign investment, and permission for entrepreneurs to start businesses were the main theme. However, most industry remained state-owned in that period. In the second stage of reform from late 1980s to 1990s, reform measures including privatization and contracting out of much state-owned industry were carried out. The implementation of reform and opening-up policies provides two key new benefits for China s economic growth. First, the new household contract responsibility system allows farmers to leave farmlands for higher income jobs as long as they could finish the contracted agricultural production duty. Before this system is carried out, farmers have to spend 12 months a year on the farmland, even when there is nothing to farm. Now they can set up township enterprises or migrate to the cities to work after they finish their farming duties decided by the contracts. With the development of agricultural technologies, a larger proportion of rural labor aged 16 to 45 works full time in the urban areas and leave farming work to their elders or women staying in rural areas, who also take care of their children. A large group of immigrant workers was formed and it has greatly influenced the Chinese economy and become a social phenomenon. To understand how Chinese labor is being integrated into the global labor market, we need to identify the size of China s labor migration to the urban sectors. There is a good amount of research on this issue, most of which makes estimations according to household surveys and international comparison. Work of Cai (2007) focuses on the age structure and distribution of China s rural labor and estimates that in 2004 China still had a 120 million labor surplus in rural areas. Wang and Zhong (2011) compare China s labor migration with that of Japan, South Korea, and Taiwan; they believe China has passed the so-called "Lewis Turning Point : zero surplus labor. They also emphasize that although there is no surplus labor in

10 Chinese rural areas, the share of agricultural labor in China is still much higher than other countries at the same income level, and there is still room for further labor migration with technology improvements and institutional changes. In this paper we estimate the number of China s labor migration in two ways. First, we used the household survey data of migrant workers and dependency ratios to calculate the gross labor transfer on a yearly basis. Second, we use data from China s yearbooks to calculate the aggregate number of migrant workers in urban areas. a) Gross labor transfer ( ) Gross labor transfer from rural areas to cities consists of two components. The first is migrant workers, the rural labor force working in cities while keeping their families in rural areas. The second is workers in migrant households. The household survey team of the National Bureau of Statistics (NBS) in China announced the number of gross labor transfer in China from 2001 to 2003 but not in the following years; we make our estimation using the number of the migrant population, migrant workers and household dependency ratio. We present our results in the following chart 3.1. In every year between 2001 and 2011, more than 10 million rural laborers went into cities to work. The number of migrant workers went up prior to the 2007 global financial crisis and then started to fall. In 2010 and 2011, the number of new migrant workers went up again to more than 10 million. The number of laborers in migrant households also kept rising before the financial crisis, but stagnated after it. The peak in gross labor transfer took place in 2010, when 19 million new migrants joined the urban labor force. One possible explanation for the recovery of the labor transfer was that jobs lost during the financial crisis were replaced by new jobs created in the real estate industry boom that started in b) Aggregate labor migration ( ) Before 2000, China s National Bureau of Statistics did not account for migrant workers. To better understand the labor transfer in China, we calculate the aggregate number of labor transfer using yearbook data from 1980 to Three sets of data in China s yearbooks are categorized as labor statistics: the total amount of the labor force; rural and urban working population; and working population of three major industries. The gap between rural labor and labor in the agricultural sector or the gap between urban labor and labor in the industry and service sector is the number of

11 aggregate labor migration. In the following figure we draw the time series of aggregate labor migration and its year-on-year differences. From figure 3.1 above we can clearly distinguish three phases of labor transfer from 1980 to The first phase was from 1980 to 1996 during which labor transfer grows at high speed. During this 17-year period, aggregate labor transfer grows by 11% on a y-o-y basis, and the average growth rate of GDP and trade are 10.2% and 13.5% respectively, both of which are significantly higher than before The average growth rate for the agricultural, industrial and service industry are 5.1%, 12.3% and 11.5%. For the first time in Chinese history, growth in the latter two industries exceed that of the agricultural sector, and the high growth in both sectors is fueled by the huge amount of rural surplus labor. The second phase was from 1997 to 2002 when aggregate labor migration started to slow down. Both internal and external reasons led to the decline. The external reason is the Asian financial crisis in late 1997 that reduced foreign demand for Chinese exports. The internal reason was the SOE reform started in 1997 that laid off a large number of urban workers. According to official data, more than 21 million SOE workers were laid off between? 1998 and 2001, and a large number of them competed with migrant workers for work opportunities. In this phase of labor transfer, the average GDP growth fell to 8.4% and growth in the industrial sector fell to 8.4%, 4.9% lower that of the earlier phase. Growth in trade is still 13.5%, and one likely explanation is that laid-off workers rather than immigrant workers filled the new job vacancies created by exporters. In the second phase, the wage rate grew much slower than GDP growth. The third phase was from 2003 to the present when aggregate labor migration sped up again. After China joined the WTO in 2001, a large number of manufacturing jobs were created and international trade grew at top speed. From 2003 to 2013, China s GDP grew by 10.2% per year, while total trade volume grew by 18.9%, 5.4% higher than previous periods, and made more contributions to China s GDP growth compared with consumption and investment. Why is the integration of Chinese labor into the global labor market more important than that of Japanese labor and Korean labor in their takeoff period? Size of labor migration matters. Japan, whose industrialization started in the Meiji Restoration of 1868, also sees labor migration in its economic recovery after WWII. However, the size of Japan s labor migration is quite small compared with that of China. The peak of Japan s labor migration, which occurred in 1967, was less than 250 thousand. The

12 case of South Korea was quite different. The economic rise of South Korea was fueled by a relatively large labor migration. According to the work of Kim (2014), the agricultural share of the labor force of South Korea decreases from 63% in 1963 to 5% in In the mid-1960s export oriented industrial enterprises in South Korea started to expand and at the same time labor migration started to accelerate. The labor migration in South Korea reached its peak in 1975 at 800 thousand, and then started to fall. From 1960 to 1980, 6 million rural laborers transferred to urban areas, quite a large scale compared with other countries. However, China s aggregate labor migration from 1980 to 2000 was more than 120 million, 20 times South Korea s total. In fact, China s labor migration is unprecedented in history and has had a profound influence on the global labor market. 3.C Institutional endowments of China s labor market The fact that China possesses a large amount of rural surplus labor does not automatically generate economic development and burgeoning trade. Indonesia and Pakistan have the 4 th and 6 th largest populations in the world, respectively, both having a large amount of rural surplus labor. But both economies have been growing at a steady speed and labor immigration didn t occur on a large scale. The reasons why labor started to immigrate after the late 1980s are the internal conditions and more importantly, the institutional endowments of modern China. The key institutional endowments in the Chinese economy are China s household registration system, also known as the Hukou system. Every resident in China has a household registration record that identifies him or her as a resident of a certain area and includes identifying information such as name, parents, spouse, and date of birth. The Hukou system limits population migration from rural areas to urban areas. A rural worker seeking to move to urban areas to take up non-agricultural work would have to apply through the relevant bureaucracies, and the number of workers allowed to make such moves is tightly controlled. Before 1978, the central planning system and Hukou system made it nearly impossible for rural workers to leave farmlands and created a unique dualistic structure between rural and urban areas. The reform to the Hukou system started in the early 1980s to meet rising labor demands in the cities, especially in the industrial sector, allowing rural labor to start flowing into urban sectors at a slow pace. Due to the discrepancy between the slow pace of hukou reform and gradually increasing urban labor demand, the migration of China s surplus

13 labor wasn t completed immediately after the reform and opening-up of China in This process took more than 30 years and continues today. It allows China to maintain its favorable position in labor costs in the mid-run, while the Hukou system itself further lowers the pricing power of labor in the modern sector. Residents with rural Hukou can only go to school in rural areas and don t enjoy the social securities in cities even if they work in cities. Therefore, employers can further depress wage rates in labor negotiations, since they assume migrant workers raise their children in rural areas and have no social security demands. The advantages in labor costs were directly reflected in good prices and allowed China to develop its economy through export. The "Chinese-style" labor transfer is a key reason why China s export trade can greatly influence the global economy and labor share. 3.D Features of China s migrant workers The following three features distinguish China s migrant workers from their urban cohorts and result in export-oriented companies' favoring migrant workers. First, migrant workers are less educated compared with urban workers. Most of the migrant workers can only do simple manual work like assembling IPhones or weaving in traditional industries or service sectors. According to the labor statistics collected by the National Bureau of Statistics (NBS) of China, more than 80% of China s current migrant workers never went to high school. The large number of migrant workers causes fierce competition in simple manufacturing and processing jobs and lowers average wage rates for those jobs. However, these wage rates are already significantly higher than the average agricultural income. Thus, migrant workers are still willing to work in cities at this wage level. Second, migrant workers are highly mobile among workplaces. Because of the restrictions of the Hukou system, migrant workers can t buy houses or receive any benefit from urban social securities, and they move frequently around different cities and provinces to seek higher income jobs. According to a survey by CASS, every year more than 60% of the migrant workers change their workplace at least once and more than 35% will move between cities and provinces. Most of the moves take place after the Chinese traditional spring festival. Third, location of labor migration and job selection are greatly influenced by regional economic development and macroeconomic policies. The labor migration of China can be separated into two periods. From the 1980s to the early 2000s, the major form of labor migration was inter-province.. Labor demand in coastal provinces rose due to fast-growing export

14 manufacturing industries, and most migrant laborers went to these provinces, including Guangdong, Fujian, etc. Ever since the 2000s, intra-province labor migration has accelerated and increasing demand in central and western provinces changed the direction of labor migration. "The Western China Development Plan has created huge labor demand and many manufacturing industries have relocated their factories to the western areas to reduce costs. We expect this east-to-west labor migration trend to continue as a result of China's current macro policies orientation. 4. Theoretical model In this part we build a basic two-country dual economy model to interpret how labor shares are determined. We are able to show how labor shares of both capital-intensive countries and labor-intensive countries decline simultaneously due to dual economy features and more importantly rural surplus labor in developing countries. In our model, we assume two countries produce exactly same kind of products except for the joint product of both countries, but have different resource endowment and technology endowment. 4.A Baseline model- classical trade model The two countries in our baseline model are identical except for their resource endowment and technology endowment. In our baseline model there is no dual economy, and both countries only have one production sector: the modern manufacturing sector. Entrepreneurs provide capital and get investment returns that they can use to consume or reinvest. Workers provide labor and earn wages for consumption. We assume the two countries have C-D production function, which later can be relaxed to CES production function: (4-1,4-2) A stands for production technology in which we combine labor-augmenting and capital-augmenting technologies. K stands for capital, N stands for labor. In our baseline model, we assume that capital can flow between the two countries while labor cannot, and capital return is equal to margin capital output. Capital is transferred between these two countries so that capital

15 returns in both countries are the same for domestic entrepreneurs. Home biases are not included in the baseline model but can be added without difficulty. The above relationship is expressed in the following equation. The equation can also be express in this form. (4-3) (4-4) From the above equation we can see that at constant technology levels countries with higher labor endowment will attract capital from abroad, which decreases domestic capital return. This process ends when the capital return of both countries is equal. 4.B Labor share determination in the dual economy model In this section we introduce dual economic features into the baseline model and see how labor share is determined. In both countries we introduce a traditional agricultural sector, also known as the less developed sector that does not participate in international trade. The following features of the traditional sector differ from the modern sector. First, in this sector capital input is constant so that the only two inputs that could affect output are technologies and labor. Second, when the labor input of the traditional sector reaches a certain cutoff, marginal productivity drops to zero. Any labor input in excess of that level is surplus labor. The labor input level is governed by technology, population of the economy, capital input, and farmland size, the last three of which are exogenous. The production function of this sector is as follows: (4-5,4-6) We assume that one country, C, has surplus labor, which could be transferred to the modern sector in the middle run. The other country, A, is more developed, with zero surplus labor in its traditional sector. Only significant technology development can create new surplus labor by decreasing the necessary amount of labor for agricultural production. We assume labor income is equal to marginal productivity in both sectors at T=0.

16 (4-7,4-8) Institutional change is the trigger for labor migration in our model. With key institutional changes being implemented, like China s Hukou reform in the late 1970s, rural surplus laborers are able to move to urban areas for manufacturing job opportunities and provide dynamics for the growth of the modern sector. Before labor migration takes place, labor share of both countries is determined by the marginal productivity of both industries. (4-9) (4-10) According to the labor share equation above, we can easily conclude that labor share is determined by three factors in the dual economy model: labor share of both the traditional and model sector, and the relative share of traditional sector output in total output. The relative output share is closely related to the capital-output ratio, which is regarded as a determining factor for labor share in recent literature. Recent research on how labor share of the individual sector affects overall labor share is less common. By decomposing labor share, we find two benefits. First, this model captures the impact of economic restructuring and transition on the global labor share. The transition from the traditional sector to the modern sector accompanying large-scale labor migration is a key reason for the decline in the global labor share in the last 40 years. However, its influence is barely captured in existing literature. Second, this model solves the problem H-O models failed to answer: why did the labor share decline in both capital-intensive and labor-intensive countries? Now we are moving to T=1. Because of the existence of rural surplus labor and key institutional change that allows surplus labor to immigrant to urban areas, marginal capital productivity of C increases, capital is either raised from domestic or flows from abroad to country C, and the additional capital either goes into the modern sector or a new industry which we want to

17 emphasize here: processing trade. In our model trade relies on joint work between both countries. Country A provides the capital components of the joint project while country C provides the labor components, an analogy of globalization in the real world. Advanced countries export intermediate products to labor-intensive developing countries like China, India etc.; and surplus labor reserves in those countries allow domestic entrepreneurs to earn extra profits by keeping labor costs low. The wage rate paid to country C s labor in the processing trade business is a fraction of the marginal productivity of the modern sector. Country A produces intermediate good p: (4-11) Intermediate good p is processed in China and becomes the same final product as the product in the modern sector: (4-12) From the new production function we can get the new labor share for both countries in labor migration periods: (4-13) (4-14) The traditional sector has higher labor share due to its labor-intensive nature. The transition of the economy from the traditional sector to the modern sector will reduce labor share for developing countries. The processing trade business sector also has its influence on the labor share of both countries. We compare the labor shares of both countries at T=1 with their T=0 labor shares separately. For country C labor migration increases the share of the modern sector that is also the low labor share sector in our model, and the share of processing trade. According to our analysis, the labor share of the processing trade business is even lower than the modern sector as domestic entrepreneurs have labor cost advantages due to the rural surplus labor reserve. Country C s labor share decreases as a whole

18 compared with that in T=0. In terms of country A, domestic modern manufacturing suffers from the rise of country C s labor share, but emerging multinationals and fast-growing processing trade allows country A to focus on its competitive advantage -- capital-intensive industries -- and to outsource the relative low value-added production process to country A. In a word, capital flows from the modern sector to the processing trade sector whose labor return is even lower. Both countries labor share declines, but due to a different mechanism. 5. Empirical approach In this section we use panel regression from an international sample to test our theoretical results. We basically reach three conclusions from our empirical results. First, the integration of Chinese labor through trade could roughly 60%-68% of the global labor share decline from 1995 to The effect could be smaller if we look at a longer period such as Second, the relative price of investment, capital formation share of total output and economic structures also have a strong influence on labor share. Third, from our robustness analysis, we find that several key institutional changes, such as China joining the WTO in 2001, amplify the effect of trading volume with China on labor share movements. 5.A Data sources and measurement In this subsection we provide a brief summary on our data construction and measurement. The dependent variable we focus on is the labor share of different countries. In the SNA system of national accounting, economic activities are categorized into three sectors: the corporate sector, household sector and government sector. The corporate sector includes both financial and non-financial corporates. The overall labor share of the economy is defined as the labor compensation in three sectors divided by GDP, which means we can also calculate the labor share of separate sectors by dividing labor compensation in separate sectors by the sector output. a) Two sample periods The time period of our empirical analysis is from 1980 to 2011, for which we have complete trade data of China with its trade partners. National Bureau of Statistics (NBS) and China's Customs Administration have published detailed trade information since 1995, including China s trade

19 volume with every one of its trade partners. The trade data before 1995 is not complete and comes from different sources, which is also not reliable. For the period before 1995, we use a smaller sample that includes China s four major trade partners (Hong Kong, U.S., Japan and Germany) to conduct a simple correlation analysis, and analyze the sample using panel regression. b) Small sample correlation test ( ) The small sample we use contains the data of four economies: U.S., Hong Kong, Japan and Germany, and they rank 1st, 2nd, 3rd and 6 th, respectively, in Chinese trade volume ranking in In the sample period we only have the overall labor share data of these four economies and a few years trade data, both of which have limited the empirical method we could use. For this reason, we draw the scatter diagram of the four economies' relative trade volume and their labor shares and its fitting curve. We made exponential adjustments to raw data to make sure trade data is comparable to labor share. In figure 5.1 we show the scatter diagram and fitting curve. We can easily see a negative correlation relationship between trade relations with China and labor share, with a coefficient around We are increasingly convinced of a relatively strong relationship between trading with countries labor share and their trade relationship with China. In the next subsection we are going to present the regression results from the bigger sample from 1995 to c) Dependent variable: corporate labor share We use a similar approach to Neiman and Karabarbounis (2013) and focus on the labor share of the corporate sector as our dependent variable. We get three benefits by using this approach. First by using corporate labor share we avoid the accounting issues in overall labor share. Work of Gollin (2002) pointed out that overall labor share accounting methods vary between countries and are mostly biased. On the other hand, corporate labor share is calculated the same way across countries and is more appropriate for our empirical analysis. Second, using corporate sector labor share also allows us to avoid problems with the government sector, such as how to interpret production function for the government and how to define labor compensation in government, making cross-country comparison easier. Third, from the perspective of data processing, corporate labor share is

20 more stationary and reliable for empirical analysis. We also replicate our qualitative analysis using overall labor share data for comparison purposes. The labor share data we use is from KN dataset, 2014 version. (We would like to thank Neiman and Karabarbounis for documenting and publicizing this dataset that has been a great support for scholars doing labor share research.) Some of our independent variables including the capital formation share of corporate output also come from the same dataset, which we have noted in the following sections. d) Sample description For the period, we chose the countries that have at least 8 years of labor share data to do our panel regression. There are 66 countries of this kind, 37 from Europe, 11 from Asia, 8 from the American region, 8 from Africa and 8 from Oceania. Countries with different economic institutions and at different development levels are included in our model. Most countries in our sample experienced labor share decline: 46 countries experienced labor share decline, while 20 experienced labor share increases. Of those 53 countries where the trends were statistically significant at the 5% level, 44 experienced labor share decline, and 7 experienced labor share increase. Of the world s 8 biggest economies, 7 experienced labor share decline, the only exception being Great Britain. 5.B Explanatory variable variables If we can get the number of workers that serve in the export business for a certain country, it will be a better variable to estimate the effect of China s labor integration. However this data doesn t exist and we have to use other data to measure China s labor integration into the global market. In this article we use trade data and FDI data. 1) The first key explanatory variable in our regression is the share of trade with China in an economy s total output. We use this variable to capture how China s labor integration into the global market affects the labor share of China s trade partners. We get the trade volume data between China and the countries in our sample from the trade yearbooks issued by China s NBS and customs administration. Import volume, export volume and the total trade volume, respectively, are divided by GDP and then put in our regression. 2) Relative price of investment is also a key explanatory variable in our panel regression. We use the PWT price data and WDI price data documented by Neiman and Karabarbonious to measure the relative price.

21 3) Capital-output ratio is also regarded as an important explanatory variable in labor share movements. Work of Bentolila and Saint-Paul identifies a one-one relationship between labor share and capital-output ratio for OECD countries, and uses industrial data to test their results. It is hard for us to get reliable time series data on the capital-output ratio for our sample period, and we use the fixed investment share of corporate output and capital formation share of total output in our regression to represent movements in capital-output ratios. 4) In our model we also consider the effect of economic development. Economic development comes with the expansion of the financial sector and industrial structural change. The expansion of the financial sector and especially the expansion of social credit provide funds to enterprises to further invest and recruit more labor for development. According to the work of Li, Liu and Wang, growth in the wage rate is lower than that of normal GDP when surplus labor exists in this case, and will further decrease labor share. In our regression analysis we use the total credit divided by GDP in each country to evaluate the effect of financial development on labor share. Industrial structural change also has a major influence on a country s labor share. The economic development process for most countries is also the process of restructuring. Industrialization marks the start of economic transition from agriculture to manufacturing. Development in the service sector needs to take place for the economy to further develop. In this process the labor share will usually decline and then rise again (Li, Liu and Wang, 2009, Bai and Qian, 2010). We put in our regression the share of agricultural share of total output and service share of total output to capture how transition from agriculture to industry and industry to service sector influence labor share. 5) We also introduce two variables to describe local labor market conditions of those countries. We use the proportion of the labor force in the population to measure trends in local labor market and the potential growth of the labor force. The second variable we use is FDI to China share of a country s GDP. Countries that invest a larger share of its capital in China have a higher possibility of dampening the domestic labor market. When the entrepreneurs have higher bargaining power in labor bargaining, labor share experiences a decreasing trend. According to the analysis above, we expect a negative correlation between FDI to China and labor share. 5.C Baseline Regression

22 In our baseline regression, we use fixed-effect panel regression to analyze the relationship between labor share and trade with China. In addition to the variables we mention above, we also put in our model a time dummy variable i.year to control for the time fixed effect. Table 5.1 below presents a descriptive summary for these variables. Our baseline regression results are presented in Tables 5.2 and 5.3. Reg(1) and Reg(2) in the first two tables take into account the time fixed effect, while the following regressions are without this fixed effect. Different investment prices (PWT price and WDI price) and different measures of industrial structure (share of agriculture and service sector and share of industrial sector) are separately put in our regression for robustness test. Table x shows the results using corporate sector labor share as dependent variable, and table y shows the results using overall labor share as dependent variable. Because we focus on cross-country variations in international trade, Reg(1) in table x gives us benchmark results. In various specifications, the coefficient for total trade volume with China is highly significant and negative. This means, for any country in our model, increasing its trade volume with China will significantly decrease its labor share in the corporate sector. But by how much? China s international trade accounts for 12.1% of the world s total trade volume, which means that 12% of a typical country s total trade is done with China. The average China trade share increased from 1.6% to 3.5% from 1995 to Combining the above data with our coefficient, our conclusion is that in the 17 years between 1995 and 2011, 1.5%-1.7% of the corporate sector labor share decline can be explained by the expansion of China s trade. That is to say, in this period, 60%~68% of the global labor share in the corporate sector is due to the growth of China s international trade and labor migration of China in the global labor market. We can also identify the effect of investment prices on labor share by the results of our panel regression. The view of Neiman and Karabarbonouis is that the decline in the relative price of investment goods cause labor share to decline. The positive coefficient of PWT prices leads us to agree with the view of Neiman and Karabarbonouis. However, the impact of price decline of investment goods is not comparable to that of the share of China s trade. Using the coefficient of PWT prices and the yearly movements in PWT prices, the relative investment price decline could explain 0.63% in the global labor share decline from 1995 to 2011, or 14% of corporate labor share decline, much smaller than the effect of increasing

23 trade with China. Here let s assume Neiman and Karabarbonouis conclusion that 50% of the global labor share decline from 1980 to 2011 is correct, and we can get two important lemmas. First, the effect of the relative price of investment in the first 15 years after 1980 is much greater than the next 17 years ( ), where expansion in trading with China plays a larger role. Second, expansion in trading with China and the decline in the relative price of investment goods can explain 74%~82% of the total labor share decline, and are the two most important determinants of labor share decline. As for other control variables, the signs before their coefficients are also in line with our expectations. The coefficient of capital formation share of total output is negative and significant. The impact of capital formation share on labor share is relatively small. When the capital formation share increased by 1%, labor share decreased by 0.1%. The increase in capital formation share of total output could explain roughly 2% of the entire labor share decline. Industrial structural changes on labor shares also have a significant influence on labor share. In the above section, we expect the agriculture to industrial sector transition will decrease labor share, and the industrial to service sector transition will increase labor share. The regression results support our view and the industrial to service sector transition is significant. We replicate our regression using the developing countries in our sample as a smaller new sample, and the coefficient of agriculture share of total output is positive and also significant. The coefficient of the proportion of labor force in the population and credit expansion are both significant and positive. At the same time, FDI to China share of GDP is not a significant variable to explain labor share movements. We reach the following five conclusions from the comparison of corporate labor share panel regression and overall labor share panel regression. 1. Trade volume with China also has a negative pulling effect on overall labor share, but the coefficient and significance are both smaller than the corporate sector case. Low data quality and different methods in calculating overall labor share could be a key reason. In the period from 1995 to 2011, overall labor share declines of 0.9%, 28.5% could be explained by an increase in trade volume with China. 2. The coefficient of PWT prices and WDI prices are insignificant in the overall labor share regression. In addition to the data quality issue we discussed above, informal sector and self-employment are relatively

24 insensitive to the relative price of capital goods. Trade volume with China has much more explanatory power in this regression. 3. The coefficient of capital formation share of total output/ GDP is positive and in most cases significant. This is contrary to the results in the corporate labor share regression. 4. As we expected, economic transition from the agricultural to industrial sector leads labor share to go down, and the industrial to service sector transition has a positive coefficient. 5. Proportion of labor force in total population and FDI still have no significant effect on overall labor share movements, and the coefficient of credit expansion is positive and significant. 5.D Variable substitution In the baseline regression, we use the share of trade volume with China in total output as our key explanatory variable. In this subsection we use either share of exports or imports with China in total output to substitute share of trade volume. From a classical perspective, imports and exports should have contrary effects on labor share. An increase in exports will create domestic labor demand and in turn increase wage rates for a labor market with no rural surplus labor. Labor share will increase if the elasticity of substation between labor and capital is bigger than 1 or if full employment is achieved. An increase in imports and especially in labor-intensive products will bring shocks to the domestic labor market, increase the unemployment rate and dampen wage rates, as explained in the work of Autor, Dorn and Hanson (2012). Labor share will decrease in this case. Different from the theoretical results from the classical trade theories, the coefficients of import and export trade share are both negative. The key to understand this result is a special kind of trade we emphasized in our theoretical model: the processing trade business. According to statistics from China s Customs Administration, a large part of China s imports are either raw materials for processing trade or intermediate goods. In 2000, the share of these two kinds of goods accounted for more than 50% of China s total imports. These goods that are either resource-intensive or capital-intensive are assembled or processed in China and then exported to other countries. The domestic labor markets of the outsourcing countries were hurt by this process and the more jobs that are offshored or outsourced, the lower the domestic labor share is.

25 5.E Dummy variables and subsample We use dummies to illustrate the effect of geographical differences. In terms of the geopolitical influence on labor share, we add the dummy variable of European countries, Asian countries, African countries, Latin American countries and Oceanian countries to existing corporate labor share regression. The results are presented in the following table 5.6. The coefficient of European, Asian and African dummies are significant in our regression, and the European dummies are the only ones with positive coefficients. Labor shares of the European countries are significantly higher than those of other continents while Asian and African labor shares are significantly lower. The geographical differences mainly come from the level of economic developments according to the other research we quoted in the literature review, and to examine their theories using our data we separate our sample into two subsamples based on economic development levels: OECD countries and non-oecd countries. By applying the exact same empirical methods to both subsamples we get the first two columns in table X. Although the coefficients of share of trading volume with China in total output for OECD countries and non-oecd countries are pretty close, the coefficients for OECD countries are much more significant. On the other hand, the coefficient of relative price of investment goods is significant for non-oecd countries. The major reason behind this divergence is the trade relationship between China and these countries. Six of China s top 8 trade partners are OECD countries or regions with at least one OECD country, and the other two are Hong Kong and AESAN. The average trading volume share of our OECD sample countries is 9.2 times that of non-oecd sample countries and we believe this is why the coefficients in our OECD subsample are more significant. 5.F Robustness test We conduct robustness test to test the time and space robustness of our empirical results. In terms of time robustness, we want to check whether a key institutional change - China joining WTO in can alter our results significantly. Although we have controlled the time fixed effect before, we could miss the impact of this event. We do subsample regression for the time period 2002 to In terms of space robustness, we also do subsample regression for OECD countries to test its robustness. Table 5.7 shows our estimation results.

26 Both of our robustness tests give us positive results. The significance and signs of our major regression results do not change, and the trade volume share with China in total output has a more significant effect on overall labor share. The unification process on labor share calculation could be a key reason. The OECD subsample panel regression gives us similar results and the only major difference from our baseline regression is that the fixed investment share of corporate output becomes insignificant no matter what labor share measure we use. 5.G Regression using labor migration data To study the direct effect of the labor migration in China on global labor share, we also use the share of Chinese migrant labor in total labor force for different years as an independent variable to explain movements in other countries labor shares. The Chinese migrant labor force shares we use are identical for different countries for the same year, as we don t have the labor data for migrant workers working on goods exported to a specific country. The migrant labor force share turns out to have a significant negative effect on other countries labor shares even though we have control over the time fixed effects and individual fixed effects. We conduct a number of robustness tests and they are all positive regarding a significant negative relationship between China s labor migration and global labor share. More exhibitions of the regression results will be seen in later versions of this paper. 6. Conclusion In this article we provide a new explanation for the global labor share decline, which is "the great doubling of the world labor market, and especially the integration of Chinese labor into the global market. Dual economic structure is the key to understand labor share declines in both labor-intensive and capital-intensive countries, and the empirical results support our view. We use cross-country panel regression to estimate the coefficient of trade volume share of total output and conclude that the growing trade with China explains 60% of the decline in the global labor share from 1995 to We also pay attention to key institutional changes such as China joining the WTO, and these changes affect the global labor share movements as we expected. Our results call for more attention to be paid to the development side of research on global factor shares. In many of the world s developing countries, labor market frictions and institutional barriers such as the

27 rural-urban barrier we see in China play an important role in labor share determination. Institutional changes of this kind also have a global influence through trade and global investment. We hope this article can generate new frameworks for development economics and have some implications for the future trends in labor share movements and labor migration in other developing countries. Tsinghua University Central University of Finance and Economics

28 References Acemoglu, D., Directed Technical Change, Review of Economic Studies, 69 (2002), , Labor- and Capital-Augmenting Technical Change, Journal of European Economic Association, 1 (2003), Antras, P., Is the U.S. Aggregate Production Function Cobb-Douglas? New Estimates of the Elasticity of Substitution, B.E. Journal of Macroeconomics, 4 (2004). David, H., Dorn, D., & Hanson, G. H. (2012). The China syndrome: Local labor market effects of import competition in the United States (No. w18054). National Bureau of Economic Research. Azmat, G., A. Manning, and J. Van Reenen, Privatization and the Decline of the Labour s Share: International Evidence from Network Industries, Economica, 79 (2012), Bentolila, S., and G. Saint-Paul, Explaining Movements in the Labor Share, B.E. Journal of Macroeconomics, 3 (2003). Blanchard, O. J., Nordhaus, W. D., & Phelps, E. S. (1997). The medium run. Brookings Papers on Economic Activity, Blanchard,O.and F.Giavazzi (2003) macroeconomic effects of regulation and in goods and labor markets The Quarterly Journal of Economics 118(3): Chirinko, R., [sigma]: The Long and Short of It, Journal of Macroeconomics, 30 (2008), Feenstra, R. C., Z. Li and M. Yu, 2013, Exports and Credit Constraints Under Incomplete Information: Theory and Applications to China, Review of Economics and Statistics, forth coming. Fei, C. H., & Rains, G. (1961). A Theory of Economic Development. American Economic Review, 51(4). Freeman, R. (2006). The great doubling: The challenge of the new global labor market. Draft, Harvard University. Grossman, G. M. and H. Elhanan, 1991, Trade, Knowledge Spillovers and Growth, European Economic Review, Elsevier, Vol. 35(2-3), pp. 517~526. Gollin, Douglas. "Getting income shares right." Journal of political Economy (2002): Harris, J. R., & Todaro, M. P. (1970). Migration, unemployment and development: a two-sector analysis. The American economic review,

29 Harrison, A. E. (2002). Has Globalization Eroded Labor s Share? Some Cross-Country Evidence, UC Berkeley, Mimeo: 46. Harrison, A., Has Globalization Eroded Labor s Share? Some Cross-Country Evidence, Working Paper, University of California at Berkeley, Hsieh, C.-T., and P. Klenow, Relative Prices and Relative Prosperity, American Economic Review, 97 (2007), Jones, C., Growth, Capital Shares, and a New Perspective on Production Functions, Working Paper, Stanford University, Jorgenson, D. W. (1967). Surplus agricultural labour and the development of a dual economy. Oxford economic papers, John Knight, Quheng Deng, Shi Li. The Puzzle of Migrant Labour Shortage and Rural Labour Surplus in China [J]. China Economic Review, 2011, 22(4) Kaldor, N., A Model of Economic Growth, Economic Journal, 67, no. 268 (1957), Karabarbounis, L., and B. Neiman, Declining Labor Shares and the Global Rise in Corporate Saving, NBER Working Paper No , Kuijs, Louis(2005) Investment and Saving in China. World Bank Policy Research Paper. Lardy N. China: Toward a consumption-driven growth path [J]. Peterson Institute for International Economics Working Paper No. PB06-6, Lewis, W. A. (1954). Unlimited Supplies of Labour. Manchester school. Roach, S. (2004). How global labour arbitrage will shape the world economy. Global Agenda. Rodriguez, F., and A. Jayadev, The Declining Labor Share of Income, Human Development Reports Research Paper 2010/36, 2010.

30 表 3.1 Labor migration estimation(in millions) Year Total Labor Migration New Migrant Workers Migrant Households N.A. N.A N.A. N.A N.A. N.A

31 Figure 3.1 Number of labor migration ( )