Do local manufacturing firms benefit from transactional linkages with multinational enterprises in China?

Abstract

This paper examines the linkage effects of foreign direct investment (FDI) on firm-level productivity in Chinese manufacturing. It is found that FDI generates positive vertical linkage effects in Chinese manufacturing at both the national and regional levels, and limited positive horizontal spillovers at the regional level. While OECD firms gain from both vertical and (probably) horizontal linkages, Hong Kong, Macao and Taiwanese firms benefit only from backward linkage effects. In the domestic sector, in which we are most interested, both state-owned enterprises (SOEs) and non-SOEs are hurt by competition from foreign firms in the same industries. While SOEs gain from vertical linkages with foreign firms, non-SOEs are unable to do so. The patterns of productivity spillovers from FDI in Chinese manufacturing seem to be determined by one key factor – the technological capabilities of the firms involved. Important data limitations and policy implications of this research are discussed.

Appendices

APPENDIX A: DATA SOURCES AND VARIABLE DEFINITIONS

The firm-level data set used in this study is from the Annual Report of Industrial Enterprise Statistics compiled by the State Statistical Bureau of China, covering firms during the period 1998–2001 in the following industries: food processing, food manufacturing, beverage production, garments and other fiber products, medical and pharmaceutical products, ordinary machinery manufacturing, transport equipment manufacturing, electrical machines and apparatuses, and electronic and telecommunications equipment. For each industry, the Bureau collects detailed data on each industrial firm in operation. The data include information on ownership classification, value added, output, capital stock, number of employees, sales, intangible assets, new product sales and exports. To remove the influence of inflation, variables have been adjusted by relevant deflators. Price indices for total manufacturing fixed assets and industrial illustrationoutput are used, which are obtained from the China Statistical Yearbook 2002.

Because of entry and exit and ownership restructuring, the number of firms in operation is changing over time. In this study, the same firms can be identified based on their identifiers. The data are cleaned via extensive checks for nonsense observations, outliers, coding mistakes and the like. This finally produces an unbalanced set of 32,008 firms. The detailed distribution by year and by ownership is provided in the table above. A firm is defined to be domestically owned if its foreign equity participation, if any, is below 25%. In this data set there are two types of foreign presence: overseas Chinese from Hong Kong, Macao and Taiwan (HMT), and other foreign investors, mainly from OECD countries (OECD).

In the paper, region is defined at the level of province, autonomous region and central municipality. Mainland China can be broadly divided into three macro areas and 31 provinces, autonomous regions and central municipalities. The coastal area includes Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Guangxi and Hainan; the central area includes Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei and Hunan; and the western area includes Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang and Tibet. There are no data available for Tibet. Chongqing was not separated from Sichuan until 1996, and has therefore been treated as one combined province. Hence, in total, there are 29 regions included in our sample.

The I–O table uses an industry classification different from the SIC classification of the database of Annual Report of Industrial Enterprise Statistics. Nonetheless, the I–O table provides the correspondence with the SIC classification at a three-digit level. We aggregate the 59 three-digit industries to 25 sectors in the I–O tables (see Table A1).

Table a1 Classification scheme

APPENDIX B: LEVINSOHN–PETRIN ESTIMATION PROCEDURE

Although a full description of Levinsohn–Petrin estimation is beyond the scope of this paper (interested readers are referred to Levinsohn and Petrin, 2003), a brief sketch of the procedure is provided below.

The estimation method starts with the production function13

where y _t is the logarithm of the value added, and k _t and l _t are the logarithms of capital and labor, respectively. Following Levinsohn and Petrin (2003), Eq. (B.1) can be rewritten as

where η _t is the i.i.d. component of the disturbance term, and ω _t is the state-dependent unobserved productivity. Labor is assumed to be a variable input, while capital is a state variable. Demand for the intermediate inputs is assumed to be a function of capital and the state-dependent productivity term: m _t =m _t (k _t ; ω _t ). When this demand function is monotonically increasing in ω _t , one can express ω _t by inverting the intermediate inputs demand function. In this case, the unobservable productivity is expressed in terms of observable variables. A final assumption required for the identification of the parameters of the production function is that ω _t follows a first-order Markov process: ω _t =E(ω _t ∣ω_t−1)+ξ _t , where ξ _t is an innovation to productivity that is uncorrelated with k _t (Levinsohn and Petrin, 2003). With this model in hand, one can consistently estimate the parameters of the production function.