![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
We construct new interregional input–output tables for China, which can be used to analyze changes in the interindustry linkages within and between eight Chinese regions, and their consequences. We claim that analyses based on these tables yield more accurate results than analyses using existing interregional input–output tables for China, because our tables explicitly account for a typical feature of the Chinse economy: the importance of processing exports activities. These activities rely heavily on imported inputs and much less on inputs sourced from domestic regions. Accounting for such differences between processing exports and other production activities reduces aggregation biases. We illustrate the usefulness of the tables by computing supply chain fragmentation indices for China and quantifying the biases that are avoided by using our input–output tables instead of conventional ones. We make our tables (for 2002, 2007 and 2012) publicly available.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 For example, Pei et al. (Citation2017) demonstrate that interior regions tend to specialize in upstream activities (such as providing natural resources and raw materials) while coastal regions predominantly carry out downstream activities and export final products. Adopting an environmental perspective.
2 Ma and Van Assche (Citation2016) analyze the factors that affect the location choice of China's export processing plants.
3 Northeast includes Heilongjiang, Jilin, and Liaoning; Northern Municipalities includes Beijing and Tianjin; North Coast includes Hebei and Shandong; East Coast includes Shanghai, Jiangsu, and Zhejiang; South Coast includes Guangdong, Fujian, and Hainan; Central Regions includes Shanxi, Henan, Hubei, Hunan, Anhui, and Jiangxi; Northwest includes Inner Mongolia, Shannxi, Ningxia, Gansu, and Xinjiang; Southwest includes Sichuan, Chongqing, Yunnan, Guizhou, Guangxi, Qinghai, and Tibet. See Appendix B of the Online Supplementary Information for a map. Note that Liaoning, Tianjin and Guangxi, which are coastal provinces, are respectively included in Northeast, Northern Municipalities and Southwest. Therefore, the share of processing exports in inland provinces is smaller than the sum of the shares in Northeast, Northern Municipalities, Central Regions, and Southwest.
4 Pei et al. (Citation2012) analyzed the contribution of changes in exports to China's value-added change between 2002 and 2007. They found to be 32% higher when the ordinary IO tables were used than when the tables capturing processing trade were used. Amiti and Freund (Citation2010) found a significant skill upgrade in China's total exports between 1992 and 2005, but they found no evidence of skill upgrading when processing exports were excluded from total exports. See also Dean et al. (Citation2011), Johnson and Noguera (Citation2012), Koopman et al. (Citation2012), and Upward et al. (Citation2013).
5 Pei et al. (Citation2017) only split the processing exports from the total exports in the IRIO tables (without estimating the intermediate input coefficients and value added coefficients of processing exports industries). They admit that this is only a half-way treatment, and formal IRIO tables that fully distinguishing processing trade are required to conduct more accurate analyses.
6 The IRIOP tables can be downloaded (as excel-files) from the Online Supplementary Information on the journal's website, or the corresponding author's personal homepage.
7 P&A trade and PIM trade differ in terms of ownership and payment for the imported materials. Under P&A trade, materials and components are supplied by a foreign company and processed by a Chinese enterprise on a consignment basis. Ownership of raw materials and components remains with the foreign company. The Chinese enterprise (i) does not pay for the imported materials, and (ii) receives a processing fee. After processing and assembly, the finished products are owned by the foreign company, which distributes them further. In contrast, under PIM trade, a Chinese enterprise purchases the raw materials and components. It becomes the owner of the imported commodities. After processing and assembly, the Chinese enterprise exports the finished products to foreign customers.
8 Bold-faced lowercase letters are used to indicate vectors, boldfaced capital letters indicate matrices, italicized lowercase letters indicate scalars (including elements of a vector or matrix). Vectors are columns by definition, row vectors are obtained by transposition, denoted by a prime (e.g. ). Diagonal matrices are denoted by a circumflex (e.g.
).
9 For brevity, we give the meanings of parts of the table related to specific activities (e.g. O) or pairs thereof (e.g. OP). Similar meanings apply for similar parts of the table.
10 The RAS method is commonly used to bi-proportionally scale a matrix of unbalanced preliminary estimates of an unknown real matrix to prescribed row and column sums (Stone & Brown, Citation1962; Bacharach, Citation1970; Lenzen et al., Citation2009, Citation2014). It is a mainstream semi-survey technique and is widely used in estimating interregional trade flows (Mi et al., Citation2018; Zheng et al., Citation2019). While other complex forms of the algorithm exist, we stick to the conventional algorithm to balance each block of the tables.
11 Koopman et al. (Citation2012) also constructed a 2007 NIOP table using quadratic optimization. We used the tables from the CAS and the NBS, since they are semi-official, publicly accessible and are available for more years.
12 The original CAS and NBS NIOP tables are not ‘bipartite’ but rather are ‘tripartite’. This is because ‘other production’ was split into two parts: production of domestic enterprises to satisfy domestic demand, and a combination of the production of non-processing exports and the production of foreign invested enterprises to satisfy domestic demand. Due to data limitations, however, this split of other production cannot be made at the regional level.
13 Also, other institutes (e.g. the Development Research Centre of the State Council of China) have constructed Chinese IRIO tables, using different compilation methods and different classifications of provinces into regions. We have used the SIC-NBS tables not only because they are semi-official and publicly accessible but also because they adhere to the division of regions that is most common in China.
14 In China's Customs statistics, two types of provincial trade data are available. One type is based on the source or destination of a delivery, the other type is according to the location of trading companies. Usually, the two types of trade data show significant differences. Products that are produced in one province are often exported (or imported) by a trading company that is located in a different province (Fu, Citation2004). For example, the exports of Guangdong by origin of goods are about 9.3% larger than the exports by location of trading companies in 2016 (NBS, Citation2017). In this paper, we have used data based on provincial trade statistics by origin or destination of a delivery. This is because in an interregional IO table we are particularly interested in knowing the region in which an imported product is used and in knowing the region in which the exported goods were produced.
15 The correspondence between this broad five-category classification and the 17-sector classification in the interregional IO table is given in Appendix C of the Online Supplementary Information.
16 For example, the total exports in 2007 are 24.6% larger in the NIOP table than in the IRIO table.
17 China's national and regional IO tables treat processing trade following the SNA 2008 system. Chinese customs provide the value of total exports and P&A exports, while BOP (Balance of payment) provides the information for processing fee. The export value in official IO tables is arrived by minus P&A exports from total exports plus processing fee. It worth mentioning that though processing fee is accounted as part of service trade in BOP, it is treated as trade in goods in Chinese IO tables.
18 This inconsistency issue does not exist for 2002. For that year, P&A imports were included both in the national IO table and the IRIO table. This was in line with the SNA 1993 guidelines.
19 According to Zhang and Qi (Citation2012), the value added in the IRIO table is obtained by adjusting the value added in provincial IO tables to the official national IO tables.
20 A major advantage of this revised BEC method is that it allows a good to go into more than one category.
21 The hierarchical estimation here follows a similar logic to TRAS (three-stage RAS). However, there are too many additional constraints other than the row and column sums, so we use the hierarchical estimation to simplify the algorithm. The estimation can make sure the domestic intermediate deliveries and final demand are exactly consistent with the variables in the NIOP tables, and therefore will yield more accurate estimations with minimum information loss (Gilchrist & St Louis, Citation1999).
22 The matrix T in Equation 1 has twice as many rows and columns if SCFs are computed on the basis of the IRIOP tables rather than on the basis of the IRIO tables. The dimensions of the summation vectors u vary accordingly.