The Leading Role of Manufacturing in China’s Regional Economic Growth

Abstract

China has experienced fast industrialization and long-lasting economic growth since the end of the 1970s. However, the manufacturing sector is not evenly distributed over the Chinese territory, which resulted in an unequal economic development across regions. The aim of this article is to measure whether Kaldor’s engine of growth hypothesis holds true in the case of the Chinese regions over 1996–2006. We believe this is an appropriate model for this country given that manufacturing still represents around half of the Chinese gross domestic product. In order to obtain more reliable results, our approach pays attention to the endogeneity of our explanatory variables, heteroscedasticity, and spillovers effects across regional economies by means of spatial econometric techniques and the spatial heteroscedasticity and autocorrelation consistent estimator.

Keywords

increasing returns growth China spatial econometrics

Introduction

China’s long-lasting economic growth since the country’s opening in the late 1980s has attracted a lot of attention among researchers. While Solow’s neoclassical growth model has often been used to measure and identify the sources of per capita income growth across Chinese provinces (see Qian and Smyth 2006; Chow 1993; Chow and Li 2002, among others), we believe that this approach is inappropriate for its lack of consideration for the geographic location and the presence of increasing returns to scale that give Eastern provinces a competitive advantage over the rest of the country. Eastern provinces’ access to the sea guarantees their capacity to trade easily with the rest of the world and their greater than average growth in population, capita income, and foreign direct investment (FDI) reflect localized agglomeration economies there (Kaldor 1970; Krugman 1991; Glaeser et al. 1992). For instance, the average growth rate of the East was 9.4 percent, 6.7 percent in the Central area, and 8.0 percent in the West over 1985–1991 (8.3 percent for the country as a whole, Hansen and Zhang 1996). In addition, the government’s efforts since 1978 have focused on developing this part of the country first (Gallup, Sachs, and Mellinger 1999; Guo 2005).

In this context, the Kaldorian approach, which is directly implied by the endogenous growth theory and thus relies on increasing, not constant as in the necoclassical framework, returns to scale, seems more appropriate to uncover the dynamics of regional growth in China (Kaldor 1966). The second advantage of Kaldor’s laws is their emphasis on the manufacturing sector. In China, manufacturing still represents around half of the country’s gross domestic product (GDP) and 14 percent of its labor force (National Bureau of Statistics [NBS] 2008). This may result from the government’s development strategy that has focused on this sector since 1978. As Lin, Cai, and Zhou (2003) mention, this policy allowed the country to take advantage of its large labor force. With the gradual abolition of the institutional factors that impeded labor mobility until the 1980s, China’s productivity grew by moving labor from low-productivity sectors in rural areas to high-productivity ones in urban areas. According to Fang, Yang, and Meiyan (2009), the significant rural–urban income disparities that resulted from the productivity gap reinforced migration to urban and more developed areas. In addition, the 1979 “open door policy” allowed the country to attract many FDIs and promote exports (Lin, Cai, and Zhou 2003). As a result, the Chinese manufacturing sector has reached an impressive average annual growth rate of 24 percent over 1996–2006, that is, more than 5 percentage points higher than the country’ GDP growth rate (NBS 2008).

While the empirical estimation of Kaldor’s laws at the national level has been fairly popular in the past (see, for instance, Parikii 1978; McCombie 1983; Thirlwall 1983; Bairam 1991; Atesoglu 1993; McCombie and Thirwall 1994; Necmi 1999), estimations at the regional level are less numerous, mostly because of data constraints. Exceptions include McCombie and de Ridder (1984) and Bernat (1996) for the US states, Casetti and Tanaka (1992) for the Japanese regions, León-Ledesma (1999, 2000) for the Spanish regions, Harris and Lau (1998) for the UK regions, while several authors worked on the European Union (EU) regions (Fingleton and McCombie 1998; Pons-Novell and Viladecans-Marsal 1999; Fingleton 2004; Dall’erba, Percoco, and Piras 2008). When it comes to the case of China, Hansen and Zhang (1996) employed a pooled data set of twenty-eight regions over 1985–1991, while Jeon (2006) used both time-series and panel data on twenty-four regions over 1979–2004. Both studies find significant increasing returns to scale. However, they only estimate the unconditional form of the laws, that is, they do not include any of the additional explanatory variables nor the spatial effects we will focus on in our estimations.

In this article, we first propose to extend the traditional estimations of Kaldor’s laws by paying attention to total factor productivity (TFP) as opposed to labor productivity only. This allows us to account for the role of efficiency change and technological change on regional growth, to avoid imposing a priori restrictions on the inputs elasticities, and to relax the assumption of similar production functions across regions. For instance, Shiu and Heshmati (2006) find that the latter component does not hold in the case of China, as they find a standard deviation of labor elasticities across provinces of .07 over 1993–2003. Second, we also extend the usual absolute Kaldor’s laws by introducing some variables capturing the regional levels of technology diffusion, human capital, per capita FDIs, per capita regional development spending, accessibility, and density, as the regional science literature give them an important role on dynamic agglomeration economies and thus localized increasing returns (Fujita and Thisse 2002). By adding them to our specification, we can measure their individual impact on local productivity growth. The same approach has been used by Angeriz, McCombie, and Roberts (2008) for the EU regions. Finally, from an econometric viewpoint, we propose three contributions. The first one consists in adopting a spatial econometric approach that allows us to account for the geographic location as well as the presence of interregional spatial externalities at the origin of increasing returns to scale (Fujita and Thisse 2002; Baldwin and Martin 2004). While some studies have already used this technique on the regional economies of China (Coughlin and Segev 1999; Sandberg 2002; Ying 2004), none of them has focused on an estimation of the Kaldorian framework (or one of its subset, i.e., the Verdoorn’s law). Second, in the absence of a large number of observations (China has thirty-one provinces/metropolitans), we cannot model the presence of heterogeneity (East–West pattern) in the form of spatial regimes and/or group-wise heteroscedasticity as is usually done in the spatial econometric literature (see Dall’erba and Le Gallo 2008, e.g., on the EU regions). Therefore, we use the spatial heteroscedasticity and autocorrelation consistent (SHAC) estimator developed by Kelejian and Prucha (2007) that allows the disturbance process to account for unknown forms of correlation and heteroscedasticity across our units of observation. To our knowledge, only Dettori, Marrocu, and Paci (2009) have used it for an estimation of regional TFP. Finally, with the exception of Angeriz, McCombie, and Roberts (2008) and Dettori, Marrocu, and Paci (2009), the problem of endogeneity has not been treated appropriately in the estimation of Kaldor’s laws. This problem, raised as early as 1983 by Thirlwall (1983) in the context of these laws, comes from the fact that the growth rate of GDP and the growth rate of manufacturing output are intrinsically linked with each other. If not treated accordingly, endogeneity leads to biased estimates. As a result, we will use the three-group method developed and first applied in a spatial context by Fingleton (2003) to build appropriate instruments and test for remaining endogeneity.

Therefore, this article intends to fill these gaps by measuring the role of manufacturing in the growth process of thirty-one Chinese regions over 1978–2004 and by paying specific attention to the presence of spatial dependence between regions. This article is structured as follows. The next section provides a brief description of China’s regional economic development process. The section on Basic and Extended Kaldor’s Laws reviews the Kaldor’s laws. The section on Data and ESDA describes the data and tests for the presence of spatial dependence in the variables of interest in Kaldor’s laws. While the section on Empirical Analysis starts with the estimation of Kaldorian laws by ordinary least squares (OLS), it continues with a more sophisticated approach that includes additional variables and takes into account endogeneity, heteroscedasticity, and spatial effects across observations using spatial econometrics and the SHAC estimator of Kelejian and Prucha (2007). The final section concludes.

China’s Regional Economic Development Process

Following the example of other developing countries, China gave priority to the manufacturing sector throughout the development phase which started in 1949; long before the country opened its economy in 1978. According to Lin, Cai, and Zhou (2003), the Chinese development strategy followed two phases: before 1978, the country experienced a comparative advantage-defying (CAD) strategy, which encouraged firms to deviate from the economy’s existing comparative advantages when they entered into an industry or chose a technology. The CAD strategy was designed to develop heavy industry for political and strategic security considerations, even though the capital necessary for heavy industry was scarce. Therefore, this CAD strategy resulted in low-economic growth, distorted economic structure, and low efficiency. However, in 1978, China started to implement a series of policies transforming the country from a central planning system to a market-oriented system, which provided the necessary macroeconomic environment to allow enterprises to perform according to the principles of a market economy (the comparative advantage-following, or CAF, strategy). Because of its choice of focusing on manufacturing to promote growth, China encouraged firms to specialize in labor-intensive activities, which can be found in light industries rather than heavy industries. Besides, following the series of reforms successfully experienced in the agricultural sector, the manufacturing sector also experienced some reforms such as a decentralization of the fiscal system, price deregulation, and exports subsidies. This resulted in a 7.5 percent increase of manufacturing production per year between 1979 and 1982 (McMillan, Whalley, and Zhu 1989; Lin 1992; Wen 1993; Huang and Rozelle 1994; Fan 1997). From 1978 to 2004, the annual average growth rate of manufacturing was more than 2 percent points higher than the overall growth rate of GDP.

In addition, China implemented in 1979 the “open door policy” to absorb FDIs and promote the growth of manufacturing. FDI jumped from virtually zero in 1979 to an amount of US$193.73 billion in 2006. Based on the experience of the four Asian “tiger” regions/countries, FDI in China started in the coastal regions of four cities in Guangdong and Fujian Provinces.¹ The specifics of FDI policies of each city varied, but all aimed at getting more foreign-invested firms to set up enterprises in China.² In 1984, the strategy enlarged to fourteen coastal cities/regions, including Hainan Province. Later in the early 1985, three zones opened to FDI: the Yangtze River delta, the Pearl River delta, and the Zhangzhou–Quanzhou–Xiamen region. The legislation on FDI became more open in 1986. Some policies were implanted to give more freedom of operations to the foreign investment enterprises and granted more tax incentives to attract foreign investments. In the meantime, local governments were also given more authority to review the applications of foreign investors. Another surge of FDI took place in 1992 after all the Chinese regions were allowed to open up for FDI. Today, 90 percent of them are concentrated in the southern and coastal regions, and mainly in³ the manufacturing sector (more than 50 percent of them). This led to a faster growth of the manufacturing sector in those areas than in the rest of China.

The contribution of FDI to economic growth in China has attracted a lot of attention. From a theoretical viewpoint, it is widely agreed that FDI improve the productivity of the host country through the provision of better access to technologies and spillover. (Borensztein, De Gregeorio, and Lee 1998; Markusen and Venables 1999). However, from an empirical point of view, finding robust evidence supporting their existence is not as straightforward. On one hand, Bao et al. (2002) and Démurger (2000) find that FDI are an effective channel for technology transfer that mainly benefited the Chinese eastern provinces. On the other hand, Shiu and Heshmati (2006), Yeung and Mok (2002), and Ng (2006) find a negative impact of FDI on GDP or TFP growth, while Ying (2004) and Zhang (2002) do not find any significant impact. The latter study concludes that the contribution of FDI to China’s technological progress through technology transfer is still not noticeable. The literature review by Görg and Greenaway (2002) indicates that the evidence of FDI generating positive spillover effects in the host country is sometimes weak and negative. This is because the reallocation of resources from domestic to multinational firms might be initially ineffective in productivity terms, as suggested by Aitken and Harrison (1999) and Konings (2001). This is not inconsistent with the idea that competition improves productivity through positive spillovers, since some firms may be affected by the latter while others are affected by the previous effect. Another element that shapes the final outcome is the capacity of the domestic firm to absorb the knowledge available from the multinational one. Domestic firms using very backward production technology and low-skilled workers may be unable to learn from multinationals (Kokko, Tansini, and Zejan 1996; Hale and Long 2006). Because of the differences in the expected outcomes at the firm level, Görg and Greenaway (2002) suggest that studies at the aggregate level may underestimate the true significance of FDI effects. They also add that the time lag used in a study affects the sign of the relationship (Ng 2006, find an optimal lag of three years for FDI to impact TFP growth) and that multinationals may want to guard their firm-specific advantages closely to prevent leakages to domestic firms, therefore no spillover takes place.

Noticing the increasing gap between the East and the West of the country, the Chinese government put forward some regional policies targeted to inland regions, such as the Western Development and the “Three Along” strategy, at the end of the twentieth century (Liu 2009; Zha 1992). Under the current regional development plan, the middle part of the country should see its energy and raw material industries develop while the focus in the western part is on agriculture, forestry, animal husbandry, and transportation. However, the current support for the hinterland and the West has nothing to do with the huge inflow of investments that took place before 1978. Because economic efficiency is still the primary goal of the government, support for economic growth in the East has not stopped and territorial imbalances are likely to widen according to Zha (1992). Therefore, uncovering the provincial economic dynamics in the frame of Kaldor’s laws is the topic of the next sections.

Basic and Extended Kaldor’s Laws

Review of the Basic Kaldor’s laws

Kaldor’s laws were put forward by Kaldor in 1966 and 1967. Based on an econometric analysis of the output, productivity, and employment growth rates of twelve OECD countries over the fifties and sixties, Kaldor states that the manufacturing sector is the main engine of growth. In fact, Kaldor’s ideas were influenced by Young (1928) who emphasized the overall macroeconomic spillover effect of promoting the manufacturing sector. Kaldor also ascribes the concept of dynamic economies of scale to Arrow’s (1962) idea of “learning by doing”, which claims that a greater manufacturing output growth rate leads to a greater productivity growth rate in this sector. Other sectors are not assumed to display economies of scale according to him, maybe because services were not as developed as they currently are.

More precisely, the first of Kaldor’s laws states that the growth rate of GDP is positively related to the growth rate of the output of the manufacturing sector (equation 1). Note that in his setting, it is the expansion of manufacturing sector that creates GDP growth. As such, the first law is usually called the engine of growth hypothesis:

q_{G D P} = a_{1} + a_{2} q_{m} w i t h a_{2} > 0,

where $q_{G D P}$ is the growth rate of GDP and $q_{m}$ is the growth rate of manufacturing output.

The second law, also called Verdoorn’s law,⁴ states that the manufacturing labor productivity growth rate is positively related to the growth rate of the manufacturing output. Traditionally, the Verdoorn’s law has been estimated as a linear relationship between the exponential growth rate of labor productivity ( $p_{m}$ ) and that of output ( $q_{m}$ ) in the manufacturing sector (McCombie and Roberts 2007):

p_{m} = b_{1} + b_{2} q_{m} w i t h b_{2} > 0,

where $p_{m}$ is the growth rate of labor productivity in manufacturing. The coefficient $b_{2}$ is usually called the Verdoorn’s coefficient, which has often been found around .5 empirically (Gomulka 1983; McCombie and de Ridder 1984; Fingleton and McCombie 1998; Pieper 2003). It implies that one percentage of growth in output will induce .5 percentage increase in productivity growth or, the other way around, a one percentage point in productivity growth induces a two percentage point increase in output growth. Verdoorn’s law implies the presence of increasing returns to scale in the sense that the technical change is endogenously induced by the output growth (Fingleton and McCombie 1998). Note that Kaldor noticed the endogeneity problem that comes from the definition of labor productivity (equation 2). Indeed, since $p_{m}$ is equal to $q_{m} - e_{m}$ (where $e_{m}$ is the growth rate of labor employment in the manufacturing sector), $q_{m}$ appears on both sides of the equation.

The third law states that growth in manufacturing output leads to growth in overall productivity in an economy, which is measured as a positive relationship between labor productivity growth of all productive sectors and manufacturing output growth (equation 4).

p_{} = d_{1} + d_{2} q_{m} w i t h d_{2} > 0,

where $p_{}$ is the aggregate productivity growth. According to Cripps and Tarling (1973), an alternative way to express the third law is:

q_{G D P} = d_{1} + d_{2} e_{m} w i t h d_{2} > 0,

where $e_{m}$ is the growth rate of manufacturing employment. The idea underlying equation (4) is that of a wage differential between high productivity sectors (the manufacturing sector) and low productivity sectors (agriculture). The transferring of labor from the agricultural sector to the manufacturing sector is not assumed to decrease output in agriculture but should increase the manufacturing productivity because of an increase in output in this sector and “learning by doing” effects. Productivity in nonmanufacturing sectors also increases because of the labor absorbed by the manufacturing sector (Kaldor 1968; Cripps and Tarling 1973; Thirlwall 1983; Drakopoulos and Theodossiou 1991).

Thirlwall (1983) highlighted that testing the direct relationship between labor transfer and productivity growth of an economy may be problematic because of difficulties in measuring productivity growth in many activities other than manufacturing. Therefore, Cripps and Tarling (1973), Thirlwall (1983), and Atesoglu (1993) suggest a regression between the growth of GDP and the growth of employment in manufacturing and nonmanufacturing sectors as a way of testing the third law (equation 5):

q_{G D P} = d_{1} + d_{2} e_{m} - d_{3} e_{n m} w i t h d_{2}, d_{3} > 0,

where $e_{m}$ is the growth rate of manufacturing employment and $e_{n m}$ is the growth rate of nonmanufacturing employment.

Extended Versions of Kaldor’s Laws

We propose three extensions compared to the traditional models of Kaldor’s laws. The first one proposes to pay attention to TFP while the second one introduces some variables rarely added, but theoretically relevant, to a Kaldorian framework. The third extension, which consists in modeling interregional spillover effects, will be developed in the next section.

The original Kaldor’s laws have been designed to test the relationship between manufacturing labor productivity growth and manufacturing output growth. However, increasing evidence at the international (Easterly and Levine 2001) and regional level (Shiu and Heshmati 2006; Angeriz, McCombie, and Roberts 2008; Dettori, Marrocu, and Paci 2009) shows that differences in growth rates are mostly explained by TFP, which accounts for efficiency change and technological change, rather than the accumulation of the traditional factors of production. As a result, we will analyze in subsequent sections both the standard Verdoorn’s law and the one based on TFP. This variable is not directly observable, therefore it must be calculated. As is often done in the literature, we use a standard growth accounting methodology to generate the regional values of TFP, where the production function ${t f p}_{i} = q_{i} - α k_{i} - β l_{i}$ varies across regions and where the right-hand side variables are respectively the (log of) the growth rate of output, of capital, and of labor for each region i. $α$ represents the elasticity of output to capital, $β$ the one of output to labor. Under the assumption of constant returns to scale, $α$ equals (1− $β$ ) and is calculated as the capital’s share of total output. They are assumed to differ across regions and over time, but not their sum (see Angeriz, McCombie, and Roberts 2008, for further explanations). More precisely, $α = 0.5 * (α_{t 0} + α_{T})$ where $α_{t 0}$ is the average capital’s share of total output over 1995–97 and $α_{T}$ is the same for 2005–07. $β_{i} = \frac{w_{i} L_{i}}{Y_{i}}$ where $w_{i}$ is the average wage of workers in manufacturing in 2000 prices. These data also come from the Yearbook of the Chinese Statistical Institute.

Because the assumption of constant returns to scale may be too restrictive in the case of the Chinese provinces, we also consider $α$ and $β$ under the assumption of increasing returns to scale. In this case, both coefficients need to be estimated. They differ by region but are constant over our time period. We performed an OLS estimation of the model $q_{i, t} = α k_{i, t} - β l_{i, t} + ϵ$ where $α$ and $β$ are coefficients to estimate, where TFP is the residual (Solow 1956; Abramovitz 1956), and for which we have data for every single year over 1995–2007 in order to increase our number of degrees of freedom.

The second extension consists in following the approach of Fingleton (2001) who assumes that the regional growth rate depends on several additional explanatory variables measured at the initial level to guarantee their exogeneity:

The initial technological gap is measured as the difference between a region’s TFP level in manufacturing and the one of the leading technology region (G). Two different types of initial technology gaps have been measured, one based on the assumption of constant returns to scale, where ${T F P}_{i, t_{0}} = \frac{Q_{i, t_{0}}}{K_{_{i, t_{0}}}^{α} L_{_{i, t_{0}}}^{1 - α}}$ , and one based on the assumption of increasing returns to scale where ${T F P}_{i, t_{0}} = \frac{Q_{i, t_{0}}}{K_{_{i, t_{0}}}^{α} L_{_{i, t_{0}}}^{β}}$ . In the first case, the coastal province of Zhejiang is the technology leader. Under the assumption of increasing returns to scale, where $α$ and $β$ have been estimated econometrically, the leading region is Hebei, also a coastal province. We expect a positive effect of the technology gap on productivity, as the backward regions are more likely to adopt new technologies, because of lower switching or adoption costs, which result in larger net benefit (Abramovitz 1986; León-Ledesma 2002).

The stock of human capital in the region (HK). As often depicted in the endogenous growth theory (Romer 1986; Lucas 1988), the expectation here is that human capital positively affects technological progress.

The level of accessibility (acc) is assumed to be positively correlated with productivity growth. Indeed, Aschauer (1989) and Barro (1990) predict that if public (transportation) infrastructures are an input in the production function, then new public infrastructures increase the marginal product of private capital, fostering capital accumulation and growth.

Density (d) reflects the fact that human capital tends to agglomerate in cities and thus leads to greater productivity. The fact that denser regions are also more productive regions has been verified on the US states by Ciccone and Hall (1996) and on the European regions by Ciccone (2002).(e) The per capita level of FDIs. As mentioned in the China’s Regional Economic Development Process section, the literature offers conflicting predictions on the role of FDIs on local productivity. On one hand, Romer (1993) and Blomström (1986) find that foreign firms facilitate the technology transfer and spillovers to local firms while, on the other hand, Aitken and Harrison (1999) do not find any evidence of a positive technology spillover from foreign firms to domestically owned ones.

The per capita level of regional development spendings (gov) are the sum of several elements: capital construction, funds and loans for the technical updates of enterprises, expenditure for simple construction, expenses for science and technology, expenditure for operating expenses of the department of Industry & Transportation, of the department of Commerce, of the department of Culture, Education, and Health Care, of the department of Sciences, and expenses to support underdeveloped areas. Endogenous growth theory grants public policies an important and positive role in the determination of growth rates in the long run; therefore, we expect a positive impact of government expenses on growth and productivity.

Finally, we also include the spatial lag (

W q_{G D P}

W p_{m}

) to capture the fact that the growth rate of GDP or of manufacturing productivity in one region depends, to some extent, on the growth rate of the same variable in neighboring regions. In fact, as showed by Fingleton (2001), the spatial lag version of Kaldor’s extended first law is obtained from a theoretical model with assumptions that underpin the endogenous growth theory and the New Economic Geography theory. The starting point is a Cobb-Douglas production function determining the level of output in the manufacturing sector. He assumes that the growth rate of technical progress in one region is a function of capital accumulation and technical change taking place in all the other regions, but that their impact decreases with distance. According to Fingleton (2001), under certain assumptions, the spatial lag form of the Verdoorn’s law equals the AK model of the endogenous growth school of thought (Aghion and Howitt 1998). For our case, further details on the definition of the spatial weights matrix and the implications of spatial dependence for our models will be provided in the Data and ESDA section.

With the addition of the variables above, the three Kaldorian laws become:

q_{G D P} = a_{1} + a_{2} q_{m} + a_{3} W q_{GDP} + a_{4} G + a_{5} H K + a_{6} a c c + a_{7} d e n s + a_{8} f d i + a_{9} g o v + ε.

p_{m} = b_{1} + b_{2} q_{m} + b_{3} W p_{m} + b_{4} G + b_{5} H K + b_{6} a c c + b_{7} d e n s + b_{8} f d i + b_{9} g o v + ϵ .

\begin{aligned} q_{G D P} = d_{1} + d_{2} e_{m} - d_{3} e_{n m} + d_{4} W q_{G D P} + d_{5} G + d_{6} H K + d_{7} a c c + d_{8} d e n s + d_{9} f d i \\ + d_{10} g o v + ϵ . \end{aligned}

where $ϵ$ is an error term with the usual properties. However, we will come back to this assumption later. Note that the coefficients $a_{3}, b_{3}, a n d d_{4}$ reflect the presence of interregional spillovers, that is, that productivity growth occurring in region i is affected by the growth of productivity (via technological progress) in surrounding regions. Variables $q_{G D P}, q_{m}, p_{m}, e_{m}, a n d e_{n m}$ are expressed in logarithm. Following the endogenous growth theory, equations (6) to (8) describe specifications in which regional (productivity) growth is positively associated with the diffusion of technology from more advanced (more urbanized and more accessible) regions to less developed regions.

Data and ESDA

Data

The majority of the data used for this study come from the annual editions of China Statistical Yearbook (from NBS 2005, 2008). We exclude the regions of Hong Kong, Macao, and Taiwan of our sample because of data incompatibility, so that we work on a sample of thirty-one regions. Note also that the province of Hannan and the municipality of Chongqing, which separated from the provinces of Guangdong (in 1985) and Sichuan (in 1999), respectively, are treated as separated regions. We subtracted the values of Chongqing from those of Sichuan in 1995 and 1996 based on the 1997 ratio of the corresponding variables between both provinces. Our data set covers the 1996–2006 period, but all the data are measured on a smoothing average of the initial period (1995–97) and the final one (2005–07), so that our results are less sensitive to a sudden change in the value of our variables of interest. All the data are expressed in 2000 Yuan.

In order to compute the manufacturing TFP and its growth, the yearly data on capital stock are the “annual average balance of net value of fixed assets” in the manufacturing sector. The net value gives us directly the value of capital stocks at the time considered, that is, its depreciation since its creation/purchased has already been accounted for. This variable does not exist at the aggregate level. This is why many previous contributions have relied on “Gross fixed capital formation” and the perpetual inventory method to calculate capital stocks at the aggregate level. However, as noted by Zhang (2008), there is no homogeneity in the capital stocks calculated in the literature as the depreciation rates used go from 4 percent to 9.6 percent in the case of the Chinese provinces. Our approach at the sectoral level allows us to avoid this type of assumptions.

Human capital is measured as the average number of college and university students per inhabitants over the period. Accessibility is measured as the product of the average length of highways and railways over the period divided by the regional area. Density is, as usual, the ratio between population and the regional area; FDI and government expenses are in per capita terms, and the calculation of the technology gap has been explained above. All these variables are measured at the initial period (average of the three first years) to guarantee their exogeneity.

Including Space

Compared to some previous estimations of Kaldor’s laws, this article takes into account the possibility of spillover effects among regions during the growth process, as in Bernat (1996), Fingleton and McCombie (1998), Pons-Novell and Viladecans-Marsal (1999), Fingleton and López-Bazo (2003, 2006), or Dall’erba, Percoco, and Piras (2008). As mentioned earlier, taking into account the influence of spatial location in our framework means that the effect of both spatial heterogeneity and spatial dependence needs to be accounted in the regional growth process. We take both effects into account using a spatial lag model with Kelejian and Prucha’s (2007) nonparametric heteroscedasticity and autocorrelation consistent estimator of the variance–covariance matrix in a spatial context (SHAC). Because of an obvious lack of degrees of freedom, it is not possible here to estimate models with multiple regimes, as is commonly done in the spatial econometric literature (see Dall’erba and Le Gallo 2008, for instance).

Are these spatial effects relevant in the Chinese regional case? Figure 1 below clearly indicates that the regions around Beijing as well as Guangdong (in the South), Qinghai and Xizang (Tibet; in the South–West) display a higher growth rate of GDP than the rest of the country. This implies potential spatial autocorrelation among the provinces surrounding Beijing and among the other provinces. The pattern may also indicate the presence of spatial heterogeneity because the Beijing and its neighbors perform better than the rest of the country. This distribution is a bit different for manufacturing output growth as most of the peripheral regions (whether on the coast or bordering foreign countries) display the highest growth levels while the central regions do not seem to perform as well (Figure 2).

Figure 1.

GDP growth rate over 1996–2006 (in log).

Figure 2.

Manufacturing output growth rate over 1996–2006 (in log).

Testing the presence of spatial autocorrelation is often conducted by computing Moran’s I statistics (Cliff and Ord 1981; Anselin 1988). However, a spatial weights matrix must be chosen first. In this study, we construct different weights matrices based on geographical distance.⁵ The spatial weights matrix Queen represents Queen contiguity, K4 is the four-nearest neighbors matrix, while D(950) is the great-circle distance with a cutoff of 950 miles, above which the interactions are assumed to be negligible.

The values of Moran’s I for seven variables, namely the growth rate of regional GDP, of manufacturing output, manufacturing labor productivity, TFP (with constant and increasing returns), manufacturing employment and nonmanufacturing employment, are reported in Table 1. They indicate a positive and significant (at 5 percent) spatial autocorrelation for only four of our variables: labor productivity growth (with K4 only), TFP growth with constant returns (with K4 only), manufacturing employment growth (with all matrices), and nonmanufacturing employment growth rate (with queen and K4). Therefore, the pattern of regional economic growth in China is not totally random, more especially when it comes to employment growth. In other words, regions with similar characteristics tend to be spatially clustered.⁶

Table 1.

Standardized Value of Moran’s I Statistics.

	Queen	K4	D(950)
GDP growth rate	.040 (.240)	.116 (.089)	.011 (.180)
Manufacturing output growth rate	−.015 (.569)	.105 (.106)	−.032 (.550)
Manufacturing labor productivity growth rate	.176 (.034)	.111 (.168)	−.012 (.182)
Manufacturing TFP productivity growth rate (with increasing returns to scale)	.071 (.179)	−.045 (.191)	−.010 (.171)
Manufacturing TFP productivity growth rate (with constant returns to scale)	.176 (.034)	.111 (.101)	−.001 (.747)
Manufacturing employment growth rate	.179 (.037)	.168 (.037)	−.031 (.037)
Non-manufacturing employment growth rate	.131 (.050)	.114 (.047)	−.005 (.066)

Note. p values given in parentheses. Inference is based on the permutation approach with 999 permutations.

Traditionally, the choice of the appropriate form of spatial dependence (spatial lag or spatial error model) depends on the results of a classical “specific to general” search approach outlined in Anselin and Florax (1995) using tests described in Anselin et al. (1996). We decided not to adopt this strategy here for two reasons. First, as we have argued above, the spatial lag model can be theoretically founded, following Fingleton (2000, 2001). Second, the properties of the “specific to general” approach are unknown in the presence of endogeneity of some variables, which is the case here as we show below. In addition, the LM tests of this approach are maximum-likelihood based tests, whereas we use instrumental variable (IV) estimation methods in this article (see the section on Endogeneity).

Accordingly, we use a spatial lag model, as specified in models 6–8. Moreover, we take into account the fact that the residual terms of models 6–8 may contain unmodelled factors. One strategy for dealing with this unmodelled residual organization would be to specify a parametric error process such as a spatial autoregressive error process or a spatial moving average process (Fingleton and Le Gallo 2008). However, the properties of the estimators obtained would be affected if the error process is misspecified. An alternative strategy consists in using nonparametric consistent SHAC, Spatial Heteroscedasticity and Autocorrelation Estimation, since it allows for unknown forms of correlation and heteroscedasticity across the units of observation, as suggested by Kelejian and Prucha (2007). They derive a robust estimation of the variance–covariance matrix in the presence of unknown forms of correlation and heteroscedasticity of the error terms of a general model including right-hand side exogenous and endogenous variables and show that this estimator performs well in a variety of situations.⁷

Empirical Analysis

We start this section with a basic OLS-SHAC estimation without considering the spatial lag component of equations (6), (7), and (8) displayed above. We use the results as a benchmark for comparison with estimation results that include the spatial lag variable and the potential presence of endogeneity on the right-hand side.

OLS-SHAC Estimations

We start with the OLS results of Kaldor’s laws, as displayed in Table 2. Because inference on parameters can be affected by unknown forms of heteroscedasticity and spatial correlation between spatial units, we use the SHAC estimator. Specifically, Kelejian and Prucha (2007) assume that the $(N \times 1)$ disturbance vectors $ϵ$ of equations (6)–(8) are generated as follows: $ϵ = R ξ$ , where R is an $(N \times N)$ nonstochastic matrix whose elements are not known. This disturbance process allows for general patterns of correlation and heteroscedasticity. The asymptotic distribution of the corresponding OLS estimators imply the variance–covariance matrix $Ψ = n^{- 1} Z^{'} Σ Z$ , where $Σ = (σ_{i j})$ denotes the variance–covariance matrix of $ϵ$ . Kelejian and Prucha show that the SHAC estimator for its (r, s)th element is:

{\hat{Ψ}}_{r s} = n^{- 1} \sum_{i = 1}^{n} \sum_{j = 1}^{n} x_{i r} x_{j s} {\hat{ϵ}}_{i} {\hat{ϵ}}_{j} K (d_{i j}^{*} / d_{n}),

where $x_{i r}$ is the ith element of the rth explanatory variable; ${\hat{ϵ}}_{i}$ is the ith element of the OLS residual vector; $d_{i j}$ is the distance between unit i and unit j; $d_{n}$ is the bandwith, and K(.) is the Kernel function with the usual properties. Here, we have used the Parzen kernel, with a bandwidth set to the first decile, the first quantile, and the median of the distance distribution. The results obtained are quite robust to the choice of bandwidth. We present in Table 2 those obtained with the median.⁸

Table 2.

OLS-SHAC Estimation Results.

		Second law
Dependent variable	First law GDP growth*	Labor productivity growth*	Total factor productivity growth		Third law GDP growth*
Dependent variable	First law GDP growth*	Labor productivity growth*	Constant returns	Increasing returns	Third law GDP growth*
Constant	.356 (.000)	1.560 (.000)	.109 (.608)	−.887 (.848)	1.400 (.000)
$q_{m}$	.650 (.000)	.202 (.004)	.369 (.001)	1.765 (.556)
$e_{m}$					.530 (.000)
$e_{n m}$					−.371 (.076)
G	.257 (.061)	−.744 (.014)	−.277 (.071)	.197 (.000)	−.239 (.003)
HK	.002 (.011)	.005 (.000)	.001 (.187)	.002 (.922)	.005 (.000)
Acc	1.65.10⁻ ⁵ (.379)	−1.12.10⁻⁴ (.000)	−2.51.10⁻ ⁵ (.269)	2.35.10⁻ ⁴ (.457)	−3.2.10⁻ ⁵ (.186)
Dens	−9.14.10⁻ ⁵ (.157)	5.59.10⁻⁴ (.000)	−2.32.10⁻ ⁵ (.808)	−.003 (.064)	1.86.10⁻⁴ (.000)
FDI	−2.16.10⁻⁴ (.000)	−9.08.10⁻ ⁶ (.908)	3.22.10⁻⁴ (.000)	.005 (.002)	−2.03.10⁻⁴ (.007)
Gov	5.85.10⁻⁴ (.001)	−.001 (.000)	−7.66.10⁻⁴(.000)	−.007 (.221)	−4.52.10⁻ ⁵ (.767)
Adj.R ²	.452	.4515	.4663	.037	.5411

Note. p values given in parentheses.

*The results of the first, second, and third laws are presented with the case of the initial technology gap based on constant returns to scale. Under the assumption of increasing returns, the only notable change is that, in the first law, the coefficient on initial technology gap is significant and negative.

The results confirm the three Kaldorian laws. In the first law, estimation results indicate that a one percentage point increase above the sample mean in the manufacturing output growth results in around .65 percentage point increase in GDP growth. In the second law, the coefficient on the level of increasing returns is .450 and .375 when we focus on the growth of manufacturing TFP (under the assumption of constant returns) or only labor productivity. This is below the usual level of .5, which has been found in several studies measuring the Verdoorn’s law (McCombie and Roberts 2007; Angeriz, McCombie, and Roberts 2008; McCombie, Pugno, and Bruno 2002). Kaldor’s third law is also supported since employment growth in the manufacturing sector positively and significantly impacts GDP growth and the one on employment growth in the nonmanufacturing sector is, as the theory predicts, significant (at 8 percent) and negative. The gap has a significant and negative effect on GDP growth in the third law and labor productivity growth in the second law. These results reflect the absence of catching up in the regions with a low initial level of productivity. This may explain the increase in regional inequalities experienced in China. However, we note that this coefficient is positive and significant for TFP growth with increasing returns in the second law. It may be that the presence of spillover effects in this sector rather than in nonmanufacturing sector allowed places with low initial levels of technology to catch up. Human capital also significantly and positively affects labor productivity and GDP growth, which proves, as the endogenous theory suggests, that investment in human capital is growth enhancing. Accessibility has a significant, negative, and small impact on labor productivity only. Density impacts positively labor productivity and GDP growth (in the third law only), which is what the endogenous growth school of thought would predict. The impact of FDI depends on the specification one looks at. While always small, the coefficient is negative on GDP growth but positive on TFP growth.⁹ This may reflect the fact that technology spillovers took place in the manufacturing sector but not at the aggregate level, and confirm the changes in the sign of the coefficient associated to the initial technology gap. This may be due to the fact that manufacturing receives a larger share of FDI than any other sector, actually seven times larger than real estate, the second largest sector for FDI (Coughlin and Segev 1999; NBS 2005). When it comes to FDI’s impact on GDP growth, as mentioned in the section on China’s Regional Economic Development Process above, Görg and Greenaway (2002) find in their literature review that the outcome is ambiguous. Another explanation often raised in the case of China is the complexity and inflexibility of policies and bureaucratic procedures which impose higher transaction costs on foreign-funded enterprises and thereby limit their effect on GDP (Shiu and Heshmati 2006; Graham and Wada 2001). According to Graham and Wada (2001), these barriers to entry are a way to protect the inefficient and vulnerable state-owned enterprises in nonmanufacturing sectors. Finally, we note that government expenditures have had a small but negative impact on manufacturing productivity and a positive one of GDP growth (first law). Initial government expenditures did not go to the provinces that performed very well in terms of productivity (except for Xinjiang, Hainan, Tianjin, Beijing, and Qinghai). This reflects that government expenditures mainly had a goal of redistribution across provinces, even if some highly productive provinces (those noted above) also received some support for efficiency purposes.

Including Spatial Dependence

We now estimate the spatial lag version of models 6–8 in order to measure the extent to which the economic dynamics of one region depends on the dynamics of its neighbors. As the spatial lag is an endogenous variable, we use the IV method, with the explanatory variables and their first-order spatial lags as instruments (Kelejian and Prucha 1998). Inference with SHAC is still based on equation (5), but using the total set of instruments and IV residuals instead of the explanatory variables and the OLS residuals. The results are displayed in Table 3 (columns 1, 3, 4, and 5), where the spatial lag is based on the queen matrix.

Table 3.

IV-SHAC Estimation Results; Spatial Lag Models 6, 7, and 8.

	First law		Second law		Third law
	GDP growth*	GDP growth	Labor productivity growth*	Total factor productivity growth*	GDP growth*
Dependent variable	1− $q_{m}$ exogenous	2− $q_{m}$ endogenous	3	4—Constant returns to scale	5
Constant	.065 (.674)	.065 (.676)	1.972 (.000)	.105 (.641)	1.121 (.000)
$q_{m}$	.624 (.000)	.637 (.000)	.216 (.103)	.368 (.000)
$e_{m}$					.464 (.000)
$e_{n m}$					−.269 (.190)
G	.101 (.368)	.116 (.346)	−.584 (.110)	−.279 (.150)	−.346 (.000)
HK	7.43.10⁻⁴ (.579)	8.27.10⁻⁴(.515)	.007 (.000)	.001 (.294)	.003 (.002)
Acc	4.59.10⁻⁵ (.094)	4.46.10⁻⁵ (.076)	−1.43.10⁻⁴(.000)	−2.49.10⁻⁵ (.303)	−4.97.10⁻⁶ (.848)
dens	−1.02.10⁻⁴(.019)	−1.03.10⁻⁴(.015)	5.38.10⁻⁴(.000)	−2.16.10⁻⁵ (.855)	1.53.10⁻⁴ (.025)
FDI	−2.03.10⁻⁴(.000)	−2.06.10⁻⁴(.000)	−7.16.10⁻⁵ (.643)	3.22.10⁻⁴(.001)	−1.76.10⁻⁴ (.058)
Gov	5.34.10⁻⁴(.002)	5.43.10⁻⁴(.001)	−.001 (.000)	−7.67.10⁻⁴ (.000)	−5.93.10⁻⁵ (.427)
Spatial lag (queen)	.187 (.010)	.177 (.013)	−.196 (.237)	.005 (.972)	.202 (.421)
Sq. corr.	.4432	.4112	.5621	.5987	.4874

Note. p values given in parentheses. Sq. Corr. is the squared correlation between predicted values and actual values. Similar results are obtained when the initial gap is calculated relatively to the manufacturing labor productivity difference.

*The results are presented with the case of the initial technology gap measured under the assumption of constant returns to scale. When measured under the alternative assumption of increasing returns, the only changes are the following ones: in the first and third laws, the technology gap becomes nonsignificant, whereas density becomes positive and significant in the third law.

The spatial lag component is positive and significant only in the first law: GDP growth of one region is spatially dependent on the GDP growth of its neighbors. Compared to the results in Table 2, we find again that the three laws are confirmed and the associated coefficients are very similar. Coefficient results for the other explanatory variables are very similar to those found in Table 2. The only notable change is the significant and negative coefficient of density on GDP growth in the first law. This is surprising considering the predictions of the endogenous growth theory and that it keeps its positive impact on GDP growth in the third law. In order to investigate this further, we now consider the endogeneity problem in the first law.

Endogeneity

Traditionally, the first law has been charged of “spuriousness” (Thirlwall 1983; Wells and Thirlwall 2003) because manufacturing output is a significant part of the dependent variable (GDP). In the case of China, manufacturing was indeed 53 percent of the country’s GDP in 2005. In order to avoid the possibility of spuriousness, Wells and Thirlwall (2003) recommend two-sided tests of Kaldor’s first law. While the first test (equation 10) consists in regressing GDP growth on the excess of the growth of manufacturing output over the growth of nonmanufacturing output, the second test (equation 11) looks at the relationship between the growth of nonmanufacturing output and the growth of manufacturing output. The results, estimated both by OLS-SHAC and IV-SHAC (when the spatial lag is included), appear in Table 4 below:

q_{G D P} = h_{0} + h_{1} (q_{m} - q_{n m}) + h_{2} G + h_{3} H K + h_{4} a c c + h_{5} d e n s + h_{6} f d i + h_{7} g o v + ϵ .

q_{n m} = k_{0} + k_{1} q_{m} + k_{2} G + k_{3} H K + k_{4} a c c + k_{5} d e n s + k_{6} f d i + k_{7} g o v + ϵ .

Both equations (10) and (11) confirm the role of manufacturing output growth behind GDP growth. The coefficient of the excess of manufacturing growth over nonmanufacturing growth is positive and significant. We also note the positive (but very small) impact of government expenses on GDP growth while FDI impacts this latter negatively. Both variables impact nonmanufacturing output growth in the same way. This confirms previous results on the first law. Finally, we find again a positive impact of the spatial lag coefficient on GDP growth.

Table 4.

OLS-SHAC and IV-SHAC Estimation Results; Models 10 and 11.

Dependent variable	GDP growth	GDP growth	Nonmanufacturing output growth	Nonmanufacturing output growth
Estimation method	OLS-SHAC	IV-SHAC	OLS-SHAC	IV-SHAC
Constant	1.167 (.000)	.532 (.052)	.629 (.000)	.371 (.087)
$q_{m}$ − $q_{n m}$	.350 (.011)	.356 (.002)
$q_{m}$			.373 (.002)	.348 (.002)
Gap	.085 (.713)	−.184 (.405)	.248 (.103)	.104 (.533)
Human capital	.003 (.117)	−2.88.10⁻⁵ (.990)	.002 (.023)	.001 (.049)
Accessibility	1.82.10⁻⁵ (.483)	7.67.10⁻⁵ (.054)	3.17.10⁻⁵ (.193)	5.95.10⁻⁵ (.109)
Density	−8.4.10⁻⁵ (.514)	−1.12.10⁻⁴ (.208)	−1.01.10⁻⁵ (.138)	−1.25.10⁻⁴ (.031)
FDI	−2.22.10⁻⁴ (.000)	−2.11.10^-4 (.000)	−2.85.10⁻⁴ (.000)	−2.81.10⁻⁴ (.000)
Gov	6.38.10⁻⁴(.008)	5.71.10⁻⁴ (.012) (.001)	7.93.10⁻⁴ (.000)	7.59.10⁻⁴ (.000)
Spatial lag (queen)		.363		.172 (.090)
Adj. R ²	.342		.575
Sq. corr		.311		.587

We are concerned with two other issues related to endogeneity: first, endogeneity may be present in the second and third laws also. Second, endogeneity needs to be tested for. Therefore, we have dealt with this issue by constructing instruments for $q_{m}$ , $e_{m}$ , and $e_{n m}$ using the three-group method advocated by Kennedy (1992) in the context of measurement errors and used in a spatial context by Fingleton (2003). For each of these variables, we construct an additional variable that takes values 1, 0, and −1 according to whether the values are in the top, middle, or bottom third of their ranking, ranging from one to thirty-one. We have also constructed the spatial lag of the −1, 0, 1 variables described above. Using these instruments, we have computed the Hausman tests. It appears that the null hypothesis of exogeneity of $q_{m}$ is rejected for the first law (p value of .001) but that it cannot be rejected at 5 percent for the second law (p values of .152 and .144 depending on the definition of the dependent variable). On the other hand, this test does not reject the null hypothesis of exogeneity of $e_{m}$ and $e_{n m}$ at 5 percent (p value of .081) in the third law. As a final robustness test, we have therefore reestimated equation (6) using IV-SHAC by considering that both the spatial lag and $q_{m}$ are endogenous. The results are displayed in column 2 of Table 3. The coefficient results on the endogenous factor and the additional explanatory variables are qualitatively similar to the ones obtained previously.

Conclusions

This study validates Kaldor’s laws in the regional economies of China. The growth of GDP is indeed positively correlated with manufacturing output growth (the first law) and employment growth in the same sector (the third law). There is also support for the second law, also called Verdoorn’s law: the coefficient estimate of .2 (for labor productivity growth) or of .36 (for TFP growth) is below the .5 which is usually found in the literature but is consistently significant. As a result, the strong emphasis on the manufacturing sector has now proven to accelerate the growth of GDP and living standards in regions where this sector is concentrated. Our results also reveal the problem of increasing disparities across regions for two reasons. First, low productivity regions at the initial period do not display any sign of greater than average GDP growth. Second, the results of the Verdoorn’s law indicate that increasing returns to scale are significant and thus justify the accumulation process that is taking place in the more advanced regions increases their gap with the rest of the country. Overall, our results indicate it is time for the Chinese authorities to seriously tackle the problem of unequal distribution of activities across regions. Several other studies which are based on the traditional β-convergence estimation of regional growth also conclude to increasing disparities in China (see Jian, Sachs, and Warner 1996; Fujita and Hu 2001; Wang and Ge 2004, among others).

The other explanatory variables we have added to our analysis allow us to draw some important conclusions. First, we find that spatial dependence is significant in the case of the first law, indicating potential regional development policies should not focus on one region in particular but pay attention to interregional linkages. Second, our results confirm that density and human capital are growth enhancing while accessibility acts negatively on labor productivity growth. Third, the coefficients associated with FDI and government expenses vary with the specification one looks at. FDI act negatively on GDP growth but positively on manufacturing productivity growth. It may reflect the fact that technology spillovers took place in the manufacturing sector but not at the aggregate level. This may come from the much larger share of FDI in this sector than in any other sector, or from all the institutional and bureaucratic barriers to entries the local authorities put forward to protect inefficient state-owned enterprises in nonmanufacturing sectors. As a result, measures favoring FDI in tertiary activities and removing the barriers above should allow the rest of the economy to benefit from better access to technologies and spillover. When it comes to government expenses, their impact is very small but positive on GDP growth but negative on manufacturing productivity growth which indicates that they were able to support the economy of the central and western provinces which, for redistribution purposes, were the main target. While this is good news from an equality point of view, the Chinese government has made clear that efficiency is its main priority. As such, development efforts will keep taking place in the East where most of the country’s growth takes place. We then wonder how many years it will take for the East–West gap to reduce. Because of the presence of increasing returns to scale we have highlighted in this study, we are pessimistic about the possibility of this scenario, more especially because government expenses do influence this outcome but only to a very small extent.

Overall, while some of our results confirm our expectations and some bring new insights, more especially those focusing into the role of FDI, government expenses, and spatial spillover effects, we believe our results are more reliable than previous studies because we have used the newly developed SHAC estimator and have paid attention to both spatial autocorrelation and the endogeneity of our explanatory variables.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abramovitz

1956. “Resource and Output Trends in the United States since 1870.” American Economic Review Papers and Proceedings 46:5–23.

Abramovitz

1986. “Catching Up, Forging Ahead, and Falling Behind.” Journal of Economic History 43:385–406.

Aghion

Howitt

. 1998. Endogenous Growth Theory. Cambridge, MA: MIT Press.

Aitken

Harrison

. 1999. “Do Domestics Firms Benefit from Foreign Direct Investments? Evidence from Venezuela.” American Economic Review 89:605–18.

Angeriz

McCombie

Roberts

. 2008. “New Estimates of Returns to Scale and Spatial Spillovers for EU Regional Manufacturing, 1986-2002.” International Regional Science Review 31:62–87.

Anselin

1988. Spatial Econometrics: Methods and Models. Dordrecht, Netherlands: Kluwer Academic.

Anselin

Bera

Florax

R. J. G. M.

Yoon

. 1996. “Simple Diagnostic Tests for Spatial Dependence.” Regional Science and Urban Economics 26:77–104.

Anselin

Florax

R. J. G. M.

. 1995. “Small Sample Properties of Tests for Spatial Dependence in Regression Models.” In New Directions in Spatial Econometrics, edited by Anselin

Florax

R. J. G. M.

, 21–74. Berlin, Germany: Springer-Verlag.

Arrow

K. J.

1962. “The Economic Implications of Learning by Doing.” Review of Economic Studies 29:155–73.

10.

Aschauer

, d. 1989. “Is Public Infrastructure Productive?” Journal of Monetary Economics 23:177–200.

11.

Atesoglu

H. S.

1993. “Manufacturing and Economic Growth in the United States.” Applied Economics 25:67–69.

12.

Bairam

1991. “Economic Growth and Kaldor’s Law: The Case of Turkey, 1925-1978.” Applied Economics 23:277–80.

13.

Baldwin

Martin

. 2004. “Agglomeration and Regional Growth.” In Handbook or Regional and Urban Economics, edited by Henderson

Thisse

, 2672–711. Amsterdam, The Netherlands: Elsevier.

14.

Bao

Chang

G. H.

Sachs

J. D.

Woo

W. T.

. 2002. “Geographic Factors and China's Regional Developments under Market Reforms, 1979-1998.” China Economic Review 13:89–111.

15.

Barro

R. J

. 1990. “Government Spending in a Simple Model of Endogenous Growth.” Journal of Political Economy 98:103–25.

16.

Bernat

1996. “Does Manufacturing Matter? A Spatial Econometric View of Kaldor’s Laws.” Journal of Regional Science 36:463–77.

17.

Blomström

1986. “Foreign Investment and Productive Efficiency: The Case of Mexico.” Journal of Industrial Economics 35:97–110.

18.

Borensztein

De Gregeorio

Lee

. 1998. “How Does Foreign Direct Investment Affect Economic Growth?” Journal of International Economics 45:115–35.

19.

Casetti

Tanaka

. 1992. “The Spatial Dynamics of Japanese Manufacturing Productivity: An Empirical Analysis by Expanded Verdoorn Equations.” Papers in Regional Science 71:1–13.

20.

Chow

1993. “Capital Formation and Economic Growth in China.” Quarterly Journal of Economics 108:809–842.

21.

Chow

. 2002. “China’s Economic Growth: 1952-2001.” Economic Development and Cultural Change 15:247–56.

22.

Ciccone

2002. “Agglomeration Effects in Europe.” European Economic Review 46:213–27.

23.

Ciccone

Hall

R. E.

. 1996. “Productivity and the Density of Economic Activity.” American Economic Review 86:54–70.

24.

Cliff

Ord

. 1973. Spatial Autocorrelation. London, UK: Pion.

25.

Coughlin

C. C.

Segev

. 1999. “Foreign Direct Investment in China: A Spatial Econometric Study.” Working Paper 1999-001A, The Federal Reserve Bank of St. Louis.

26.

Cripps

T. F.

Tarling

A. P.

. 1973. Growth in Advanced Capitalist Economies, 1950-1970. Cambridge: Cambridge University Press.

27.

Dall’erba

2005. “Productivity Convergence and Spatial Dependence among Spanish Regions.” Journal of Geographical Systems 7:207–27.

28.

Dall’erba

Le Gallo

. 2008. “Regional Convergence and the Impact of European Structural Funds Over 1989-1999: A Spatial Econometric Analysis.” Papers in Regional Science 87:219–44.

29.

Dall’erba

Percoco

Piras

. 2008. “The European Regional Growth Process Revisited.” Spatial Economic Analysis 3:7–25.

30.

Démurger

2000. Economic Opening and Growth in China. Paris, France: OECD Development Centre Studies.

31.

Dettori

Marrocu

Paci

. 2009. Total Factor Productivity, Intangible Assets and Spatial Dependence in the European Regions. Mimeo.Cagliari, Italy: CRENOS, Center for North South Economic Research.

32.

Drakopoulos

S. A.

Theodossiou

. 1991. “Kaldor’s Approach to Greek Economic Growth.” Applied Economics 23:1683–89.

33.

Easterly

Levine

. 2001. “It's Not Factor Accumulation: Stylized Facts and Growth Models.” World Bank Economic Review 15:177–219.

34.

Fan

1997. “Effects of Technological Change and Institutional Reform on Production Growth in Chinese Agriculture.” American Journal of Agricultural Economics 73:265–75.

35.

Fang

Yang

Meiyan

. 2009. “Migration and Labor Mobility in China.” Human Development Research Paper 2009:09.

36.

Fingleton

2000. “Spatial Econometrics, Economic Geography, Dynamics and Equilibrium: A ‘Third Way'?” Environment and Planning A 32:1481–98.

37.

Fingleton

2001. “Equilibrium and Economic Growth: Spatial Econometric Models and Simulations.” Journal of Regional Science 41:117–47.

38.

Fingleton

2003. “Models and Simulations of GDP per Inhabitant across Europe’s Regions: A Preliminary View.” In European Regional Growth, edited by Fingleton

, 11–54. Berlin, Germany: Springer-Verlag.

39.

Fingleton

2004. “Regional Economic Growth and Convergence: Insights from a Spatial Econometric Perspective.” In Advances in Spatial Econometrics, edited by Anselin

Florax

Rey

, 397–432. Berlin, Germany: Springer-Verlag.

40.

Fingleton

Le Gallo

. 2008. “Estimating Spatial Models with Endogenous Variables, A Spatial Lag and Spatially Dependent Disturbances: Finite Sample Properties.” Papers in Regional Science 87:319–39.

41.

Fingleton

López-Bazo

. 2003. “Explaining the Distribution of Manufacturing Productivity in the EU Regions.” In European Regional Growth, edited by Fingleton

, 375–410. Berlin, Germany: Springer-Verlag.

42.

Fingleton

2006. “Empirical Growth Models with Spatial Effects.” Papers in Regional Science 85:177–98.

43.

Fingleton

McCombie

J. S. L.

. 1998. “Increasing Returns and Economic Growth: Some Evidence for Manufacturing from the European Union Regions.” Oxford Economic Papers 80:89–105.

44.

Fujita

. 2001. “Regional Disparity in China 1985-1994: The Effects of Globalization and Economic Liberalization.” Annals of Regional Science 35: 35–37.

45.

Fujita

Thisse

J.-F.

. 2002. Economics of Agglomeration; Cities, Industrial Location, and Regional Growth. Cambridge: Cambridge University Press.

46.

Gallup

J. L.

Sachs

J. D.

Mellinger

. 1999. “Geography and Economic Development.” International Regional Science Review 22:179–232.

47.

Glaeser

E. L.

Kallal

H. D.

Scheinkman

J. A.

Sheifer

. 1992. “Growth in Cities.” Journal of Political Economy 100:1126–52.

48.

Gomulka

1983. “Industrialization and the Rate of Growth: Eastern Europe 1955-75.” Journal of Post Keynesian Economics 5:388–96.

49.

Görg

Greenaway

. 2002. “Much Ado About Nothing? Do Domestic Firms Really Benefit from Foreign Direct Investment?” Research Paper 2001/37, Globalization and Labour Markets Programme, at Leverhulme Centre for Research on Globalization and Economic Policy, Nottingham.

50.

Graham

E. M.

Wada

2001. “Foreign Direct Investment in China: Effects on Growth and Economic Performance.” Working Paper No. 01-03. Washington, DC: Institute for International Economics.

51.

Guo

2005. “The Structure and Structure Changes in China’s Regional Economy.” PhD diss., University of Illinois at Urbana-Champaign.

52.

Hale

Long

. 2006

“What Determines Technological Spillovers of Foreign Direct Investment: Evidence from China.”

Yale University Economic Growth Center Discussion Paper No. 934, April 2006.

53.

Hansen

J. D.

Zhang

. 1996. “A Kaldorian Approach to Regional Economic Growth in China.” Applied Economics 28:679–85.

54.

Harris

R. I. D.

Lau

. 1998. “Verdoorn’s Law and Increasing Return to Scale in the UK Regions, 1968-1991: Some New Estimates based on the Cointegration Approach.” Oxford Economic Papers 50:201–18.

55.

Huang

Rozelle

. 1994. “Technological Change: The Re-discovery of the Engine of Productivity Growth in China’s Rural Economy.” Journal of Development Economics 49:337–69.

56.

Jeon

2006. “Manufacturing, Increasing Returns and Economic Development in China, 1979-2004: A Kaldorian Approach.” Working Paper No. 2006-08, Department of Economics, University of Utah.

57.

Jian

Sachs

J. D.

Warner

A. M.

. 1996. “Trends in Regional Inequality in China.” China Economic Review 7:1–21.

58.

Kaldor

1966. Cause of the Slow Rate of Economic Growth of the United Kingdom: An Inaugural Lecture. Cambridge: Cambridge University Press.

59.

Kaldor

1967. Strategic Factors in Economic Development. Ithaca, NY: New York State School of Industrial and Labor Relations, Cornell University.

60.

Kaldor

1968. “Productivity and Growth in Manufacturing Industry: A Reply.” Economica 35:385–91.

61.

Kaldor

1970. “The Case for Regional Policies.” Scottish Journal of Political Economy 17:337–48.

62.

Kelejian

H. H.

Prucha

I. R.

. 1998. “A Generalized Spatial Two-stage Least Squares Procedure for Estimating a Spatial Autoregressive Model with Autoregressive Disturbances.” Journal of Real Estate Finance and Economics 17:99–121.

63.

Kelejian

H. H.

2007. “HAC Estimation in a Spatial Framework.” Journal of Econometrics 140:131–54.

64.

Kokko

Tansini

Zejan

M. C.

. 1996. “Local Technological Capability and Productivity Spillovers from FDI in the Uruguayan Manufacturing Sector.” Journal of Development Studies 32:602–11.

65.

Konings

2001. “The Effects of Foreign Direct Investment on Domestic Firms: Evidence from Firm Level Panel Data in Emerging Economies.” Economics of Transition 9:619–33.

66.

Krugman

1991. “Increasing Returns and Economic Geography.” Journal of Political Economy 99:483–99.

67.

León-Ledesma

1999. “Verdoorn’s Law and Increasing Returns: An Empirical Analysis of the Spanish Regions.” Applied Economics Letters 6:373–76.

68.

León-Ledesma

2000. “Economic Growth and Verdoorn’s Law in the Spanish Regions, 1962–91.” International Review of Applied Economics, 14:55–69.

69.

León-Ledesma

2002. “Accumulation, Innovation and Catching-up: An Extended Cumulative Growth Model.” Cambridge Journal of Economics 26:201–16.

70.

Lin

J. Y.

1992. “Rural Reforms and Agricultural Growth in China.” American Economic Review 82:34–51.

71.

Lin

J. Y.

Cai

Zhou

. 2003. The China Miracle: Development Strategies and Economic Reform, revised ed. Hong Kong: Chinese University Press.

72.

Liu

2009. Transition of Regional Policy in China During the Thirty Year Reform. Sichuan, China: Regional Study Department, Sichuan Social Sciences Academy, mimeo.

73.

Lucas

1988. “On the Mechanics of Economic Development.” Journal of Monetary Economics 22:3–42.

74.

Markusen

Venables

. 1999. “Foreign Direct Investment as a Catalyst for Industrial Development.” European Economic Review 43:335–56.

75.

McCombie

1983. “Kaldor’s Law in Restrospect.” Journal of Post-Keynesian Economics 5:414–29.

76.

McCombie

de Ridder

. 1984. “The Verdoorn Law Controversy: Some New Empirical Evidence Using US State Data.” Oxford Economic Papers 36:268–84.

77.

McCombie

Roberts

. 2007. “Returns to Scale and Regional Growth: The Static-dynamic Verdoorn Law Paradox Revisited.” Journal of Regional Science 47:179–208.

78.

McCombie

Thirwall

A. P.

. 1994. Economic Growth and the Balance-of-Payments Constraint, London, UK: St. Martin’s Press.

79.

McCombie

J. S. L.

Pugno

Bruno

. 2002. Productivity Growth and Economic Performance: Essays on Verdoorn’s Law. Basingstoke, UK: Palgrave.

80.

McMillan

Whalley

Zhu

. 1989. “The Impact of China’s Economic Reforms on Agricultural Productivity Growth.” Journal of Political Economy 97:781–807.

81.

McMillan

2005. China Compendium of Statistics 1949-2005, Compiled by Department of Comprehensive Statistics of National Bureau of Statistics. Beijing, China: China Statistics Press.

82.

McMillan

2008. China Statistical Yearbook. Beijing, China: China Statistics Press.

83.

Nazara

Hewings

J. G. D.

. 2004. “Spatial Structure and Taxonomy of Decomposition in Shift-share Analysis.” Growth and Change 35:476–90.

84.

Necmi

1999. “Kaldor’s Growth Analysis Revisited.” Applied Economics 31:653–60.

85.

T. H.

2006. “Foreign direct investment and productivity: Evidence from the East Asian economies.” UNIDO Research and Statistics Branch Staff Working Paper, 3/2006.

86.

Parikii

1978. “Difference in Growth Rates and Kaldor’s Law.” Economica 45:83–91.

87.

Pieper

2003. “Sectoral Regularities of Productivity Growth in Developing Countries—A Kaldorian Interpretation.” Cambridge Journal of Economics 6:831–50.

88.

Pons-Novell

Viladecans-Marsal

. 1999. “Kaldor’s Laws and Spatial Dependence: Evidence for the European Regions.” Regional Studies 33:47–65.

89.

Qian

Smyth

. 2006. “Growth Accounting for the Chinese Provinces 1990-2000: Incorporating Human Capital Accumulation.” Journal of Chinese Economics and Business Studies 4:21–37.

90.

Romer

1986. “Increasing Returns and Long-run Growth.” Journal of Political Economy 94:1002–37.

91.

Romer

1993. “Ideas Gaps and Object Gaps in Economic Development.” Journal of Monetary Economics 32:543–73.

92.

Sandberg

2002. “The Dynamics of Economic Growth 1985–2000 in Chinese Provinces.” Interim Report IR-02-080, International Institute for Applied Systems Analysis, Austria.

93.

Shiu

Heshmati

. 2006. “Technical Change and Total Factor Productivity Growth for Chinese Provinces: A Panel Data Analysis.” Institute for the Study of Labor, Discussion Paper 2133.

94.

Solow

1956. “A Contribution to the Theory of Economic Growth.” Quarterly Journal of Economics 70:65–94.

95.

Thirlwall

A. P.

1983. “A Plain Man’s Guide to Kaldor’s Growth Laws.” Journal of Post-Keynesian Economics 5:345–58.

96.

Wang

. 2004. “Convergence and Transition Auspice of Chinese Regional Growth.” The Annuals of Regional Science 38:727–39.

97.

Wells

Thirlwall

A. P.

. 2003. “Testing Kaldor’s Growth Laws across the Countries of Africa.” African Development Review 15:89–105.

98.

Wen

1993. “Total Factor Productivity Change in China’s Farming Sector: 1952-1989.” Economic Development and Cultural Change 42:1–41.

99.

Yeung

Mok

. 2002. “Government Policy and the Competitive Advantages of Foreign-Financed Firms in Guangdong Province of Southern China.” Asian Business and Management 1:227–47.

100.

Ying

L. G.

2004. “Understanding China’s Recent Growth Experience: A Spatial Econometric Perspective.” Annals of Regional Science 37:613–28.

101.

Young

1928. “Increasing Returns and Economic Progress?” The Economic Journal 38:527–42.

102.

Zha

1992. “Change of Regional Economic Development Policy in China.” Chinese Geographical Science, 2:1–9.

103.

Zhang

2008. “Estimation of China's Provincial Capital Stock (1952-2004) with Applications.” Journal of Chinese Economic and Business Studies 6:177–96.

104.

Zhang

2002. “Productivity and Economic Growth: An Empirical Assessment of the Contribution of FDI to the Chinese Economy.” Journal of Economic Development 27:81–94.