Foreign Firms and Chinese Employment

Abstract

This paper examines the effect of foreign direct investment (FDI) on employment in the Chinese manufacturing sector. As one of the world's largest recipients of FDI, China has arguably benefited from foreign multinational enterprises in various respects. However, one of the main challenges for China, and other developing countries, is job creation, and the effect of FDI on employment is uncertain. The effect depends on the amount of jobs created within foreign firms as well as the effect of FDI on employment in domestic firms. We analyse FDI and employment in China using a large sample of manufacturing firms for the period 1998–2004. Our results show that FDI has positive effects on employment growth. The relatively high employment growth in foreign firms is associated with their firm characteristics and their high survival rate. Employment growth is also relatively high in private domestic Chinese firms. There also seems to be a positive indirect effect of FDI on employment in private domestically-owned firms, presumably caused by spillovers.

1. INTRODUCTION

JOB creation is arguably one of the main challenges for developing countries. Improvements in human welfare that have a broad basis are difficult to achieve without a substantial increase in modern sector employment. Without such employment, people must continue to seek a meagre existence in agriculture or the informal sector.

As an example, Asia Development Bank (2005) suggests there is a need to create at least 750 million new jobs in Asia over the next decade if the positive development with high economic growth and rapidly decreasing poverty rates is to continue. The figure can be compared to Asia's present labour force of 1.7 billion. Such a massive creation of new jobs is obviously a huge challenge and requires a broad set of policies. In the context of modern sector employment, Felipe and Hasan (2006, p. 7) argue for the need for industrial policy to ‘promote diversification of production activities into new areas, facilitate restructuring of existing activities, and foster coordination between public and private entities to make all of this happen’. It seems quite obvious that foreign direct investment (FDI) and multinational enterprises (MNEs) could play an important role in such industrial change with their knowledge of markets, technologies and distribution channels (e.g. UNCTAD, 2007). It is also clear that FDI has greatly contributed to developing East Asia's growth and industrial development (e.g. Dobson and Chia, 1997). Despite its high empirical and policy relevance, the contribution of FDI to employment in developing countries has been little explored so far.

This paper aims at examining the effect of FDI on employment in China, based on firm-level information from the Chinese manufacturing sector during the period 1998–2004. We examine both a direct employment effect, i.e. jobs created in foreign MNEs, and an indirect employment effect, i.e. the effect of FDI on jobs created in domestically-owned firms. As discussed above, there are reasons to expect that foreign MNEs can be important in generating employment. However, the positive effect on jobs created within foreign MNEs is not necessarily accompanied by a similar development in domestic firms. Two opposing effects on employment in domestic firms can be considered; a positive effect on suppliers and from various types of spillovers, and a negative effect from increased competition.

For a preview of our results, we find that foreign and private domestic firms have a comparably high growth in employment. This high growth in foreign firms is caused by favourable firm characteristics, such as high capital intensities and productivity, and by the relatively high survival rate of foreign-owned firms. Regarding the indirect effect, the empirical analysis finds positive effects of FDI on private domestically-owned firms, presumably because spillovers and learning or demonstration effects are more important than the competition effect.

The remainder of this paper is organised as follows. In Section 2, we provide a theoretical overview of potential channels through which FDI may affect employment, and in Section 3 we present the Chinese context and discuss some previous studies. The dataset and descriptive statistics are presented in Section 4 and we perform the empirical analysis in Section 5. We then conclude in Section 6.

2. FDI AND EMPLOYMENT

Economic growth in developing countries rests on a shift away from agriculture and informal services, and the ability of the manufacturing sector to absorb labour thus becomes a critical factor (Lewis, 1954). The total amount of people employed outside agriculture and the informal sector can presumably be affected by inflows of FDI. FDI might, for instance, increase the country's competitiveness by combining firm- and country-specific assets (e.g. Blonigen, 1997). This typically involves combining access to foreign markets and modern technology with a large supply of cheap labour. Such a combination of firm- and country-specific assets has frequently improved and expanded existing host-country industries, introduced production in new industries, and changed the comparative advantage of the host country (Lipsey, 2004, 2006).

In addition to introducing new industries and establishing new firms in the host country, inflows of FDI can increase employment through establishing linkages with domestic firms through purchases of locally produced goods and services. Moreover, firm-specific knowledge might diffuse from foreign to local firms through so-called spillovers. One channel of such spillovers is through turnover of employees (Fosfuri et al., 2001; Glass and Saggi, 2002; Görg and Strobl, 2005) and managers (Gershenberg, 1987; Pack, 2001). This channel might be particularly important also because foreign firms tend to provide more training than domestic firms (ILO, 1981; Lindsey, 1986; Djankov and Hoekman, 1999). It is also possible that FDI introduces new and better quality inputs to be used in the production of upstream domestic firms, thus making them more competitive and enabling them to expand production and employment.

There is another effect, however, which suggests that inflows of FDI might decrease employment in domestic firms. This will happen if foreign firms increase the competition for domestic firms and force them to exit the market or downsize their workforce. Laid-off workers might eventually be absorbed in other firms and industries but the adjustment costs can be substantial (Davidson and Matusz, 2001). It could be imagined that such a crowding-out effect is important when foreign MNEs not only focus on export markets, but also target the domestic market. There are at least two different channels through which such crowding out can take place. First, MNEs have firm-specific advantages, which give them a competitive edge over their domestic competitors despite a comparatively poor knowledge of local conditions. Second, MNEs might also raise the wage levels and press up the wages of their domestic competitors (Lipsey and Sjöholm, 2004a). Such wage increases will deter job growth in domestic firms when their cost advantages are diminishing. Finally, the replacement of domestic firms with foreign firms will have negative employment effects if the latter are more capital intensive, which is typically the case in developing countries.

3. THE CHINESE CONTEXT

Job creation is becoming one of the main economic challenges in China. One of the key reasons behind the employment pressure is the large dismantling of state-owned enterprises. The Chinese labour force consists of a staggering 779 million people (National Bureau of Statistics, 2007, Table 5-1), and it is predicted to grow at an annual rate of 1.3 per cent over the next few decades, putting a great deal of emphasis on the ability to generate enough employment opportunities (Chow et al., 1999, p. 483). The situation is further complicated by the large number of Chinese workers in the informal sector. At least 85 million Chinese are estimated to make a living in the informal sector (Cai et al., 2005). Bringing these people into modern sector employment would be tremendously beneficial for overall welfare in China.

Manufacturing seems to be the best possibility for modern sector employment expansion. The Chinese manufacturing sector is large, although the exact size is unknown and presumably underestimated in official statistics. For instance, Banister (2005) claims that the official figures underestimate the number of workers in township and village enterprises and the number of unregistered workers, and estimate manufacturing employment to be about 100 million workers, or about twice the size of total G7 manufacturing employment. Unfortunately, there are signs of stagnating and even declining Chinese manufacturing employment. Official labour statistics put manufacturing employment at about 98 million in 1996 and about 83 million in 2002 (National Bureau of Statistics, 2007, Table 5-4). Banister's estimates show a similar declining trend. The lack of employment in manufacturing is problematic in view of labour force growth and the large informal sector. One consequence is that China is experiencing rapidly increasing inequality, which to some extent is caused by stagnating incomes in the informal sector and increasing incomes in the formal modern sector (Lindbeck, 2008).

a. Previous Studies

Few studies examine employment growth in foreign- and domestically-owned firms. One notable exception is Alvarez and Görg (2007) who examine growth in employment at the plant level in Chilean manufacturing between 1990 and 2000. Their results suggest no major differences between employment growth in multinational and non-multinational firms. The authors note that the results could be biased by a selection of only surviving plants. Adjusting this potential bias by a Heckman procedure does not change their results. Based on a sample of Chinese state-owned enterprises for the period 1999 to 2003, Gong et al. (2006) examine the effect of privatisation and foreign acquisition on employment. Their results suggest that domestic privatisation leads to lower employment growth while foreign acquisition increases employment, as compared to firms that remained state owned.

There is also a literature that examines the employment effect of foreign acquisitions in terms of employment composition. Most of these papers examine the employment composition in developed countries (e.g. Almeida, 2003; Huttunen, 2007). One exception is Lipsey and Sjöholm (2002, pp. 10–11) which examines changes in employment in Indonesian plants after foreign acquisitions during the period 1975–99. Foreign acquisitions were found to target relatively large domestic plants and the acquisitions were followed by an increase in blue-collar and a decrease in white-collar workers. That different types of workers are affected differently by FDI has some implications for our study. We are only able to examine total employment but it is of course possible that, for instance, a positive effect on total employment hides both positive and negative effects for different types of workers.

While there are at least a few studies comparing employment growth in foreign- and domestically-owned firms, there are, to the best of our knowledge, no previous studies on how FDI affects employment in domestically-owned firms. There is, however, a very large literature on how FDI affects domestic firms in other respects. It has, for instance, been shown that FDI can have both positive and negative effects on domestic firms’ productivity (Görg and Greenaway, 2004; Lipsey and Sjöholm, 2005) and that it tends to increase exports and wages in domestic firms (Lipsey and Sjöholm, 2004b; Swenson, 2007).

4. DATA AND DESCRIPTIVE STATISTICS

The data used in this study have been compiled by the National Bureau of Statistics of China (NBS). The dataset is based on a census of large and medium-sized enterprises and a representative sample of small enterprises with more than 10 employees and an annual turnover of more than 5 million RMB for the period 1998–2004. The dataset covers a substantial share of Chinese industry. For instance, the data include about 50 per cent of officially reported employment in the Chinese manufacturing sector (National Bureau of Statistics, 2007).

The available firm-level economic variables include employment, wages, sales, value-added, export and fixed assets. The industry code at the four-digit level and a region code make it possible to aggregate the firm-level information up to the industry and regional levels. Using the ownership indicator, we follow Jefferson et al. (2003) and Hu et al. (2005) and create four different ownership categories: non-private domestic firms, private domestic firms, foreign firms and other firms. Non-private firms consist of state-owned enterprises and collective firms. Foreign firms include entirely foreign-owned firms and joint ventures with foreign co-owners. Other firms, finally, consist primarily of shareholding enterprises. A more detailed classification is given in Table A1 in the Appendix. It should be stressed that the ownership classification might be flawed. For instance, Dougherty et al. (2007, pp. 312–13) argue that the controlling shareholder might differ from the official registration status. We will therefore use sub-samples of firms where the controlling shareholder is more obvious, as a sensitivity check in the subsequent analysis.

Table 1 shows some descriptive statistics of the Chinese manufacturing sector by ownership. The number of private firms is by far the largest: about 10,000 private firms in 1998 increased more than tenfold to over 112,000 in 2004. This is partly a reflection of the dynamic private sector growth in China. According to officials at the National Bureau of Statistics, some of the increase is, however, caused by an improved coverage of the sample survey on small firms. Foreign firms and firms with other ownership categories have also increased by about 30 and 300 per cent, respectively. The number of foreign firms is more than 55,000 in 2004, only second to private domestic firms. In contrast to the dynamic development in the private and foreign sectors, the number of non-private domestic firms has declined by more than 50 per cent and amounts to 36,000 firms in 2004.

Table 1.  Firm Characteristics by Ownership

Firm Characteristics	Domestic Non-private		Foreign		Domestic Private		Other
	1998	2004	1998	2004	1998	2004	1998	2004
Note: The nominal values of fixed assets and value added are deflated by producer price index (PPI) at the three-digit industry level, and wage is deflated by an annual consumer price index (CPI).
No. of firms	85,543	36,268	23,817	55,248	9,974	112,856	10,341	43,379
Average employment per firm, headcount	352	281	304	309	150	127	497	320
Export as a share of sales (%)	7.9	7.0	45.9	48.3	13.1	13.4	9.6	9.4
Average annual wage per employee, 1,000 yuan	6.5	11.3	12.0	16.6	6.8	10.6	7.5	12.1
Value added per employee, 1,000 yuan	114.5	249.5	259.6	369.1	209.1	283.8	168.6	306.5
Fixed assets (capital) per employee, 1,000 yuan	42.3	67.2	106.5	104.6	41.3	51.3	55.2	78.8

Comparing characteristics of firms by different kinds of ownership, we see that foreign firms are relatively capital intensive with high levels of productivity and wages. They are of about the same size as non-private domestic firms and other firms and substantially larger than private domestic firms. The main difference between foreign and domestic firms is the export orientation: about half of the production in foreign firms is exported.

The figures suggest that foreign firms are important as employers. However, from the descriptive statistics, we cannot draw the conclusion that they are important creators of new jobs. The reason is that the above figures might be caused by foreign firms acquiring domestic firms with little change in total employment. One possibility would be to examine the effect of takeovers on employment, but the data do not allow for such an analysis since the identification code of a firm changes after a takeover. Instead, in Table 2, we look at employment growth in firms of different ownership over the periods 1998–2001 and 2001–04. Only firms present in both periods are included.

Table 2.  Employment and Employment Growth by Ownership and Sector

	Ownership	Firms Existing both 1998 and 2001				Firms Existing both 2001 and 2004
		No. of Firms	Employment 1998	Employment 2001	Growth 1998–2001	No. of Firms	Employment 2001	Employment 2004	Growth 2001–2004
Total	Domestic non-private	31,919	14,762,545	12,658,873	−14.3%	15,987	7,319,868	6,062,059	−17.2%
	Foreign	13,939	4,739,187	5,237,500	10.5%	18,903	6,529,004	8,118,380	24.3%
	Domestic private	3,963	631,799	751,690	19.0%	15,064	2,384,847	2,870,745	20.4%
	Other	4,130	2,606,045	2,428,156	−6.8%	8,094	4,392,152	4,463,280	1.6%
Textile	Domestic non-private	1,737	1,277,583	1,119,767	−12.4%	737	480,768	424,221	−11.8%
	Foreign	1,028	313,019	336,074	7.4%	1,298	413,289	463,078	12.0%
	Domestic private	380	74,581	85,246	14.3%	1,574	304,234	356,902	17.3%
	Other	300	342,707	310,397	−9.4%	543	550,788	522,171	−5.2%
Non-metallic metal	Domestic non-private	3,737	1,346,829	1,161,815	−13.7%	1,819	543,249	459,812	−15.4%
	Foreign	616	163,115	179,139	9.8%	762	204,599	228,771	11.8%
	Domestic private	328	70,821	69,975	−1.2%	1,298	251,423	270,135	7.4%
	Other	457	269,502	242,837	−9.9%	806	391,785	373,013	−4.8%
Ferrous metal	Domestic non-private	754	1,490,880	1,306,723	−12.4%	350	991,916	825,911	−16.7%
	Foreign	100	44,066	45,582	3.4%	98	38,455	45,807	19.1%
	Domestic private	103	19,371	23,179	19.7%	297	46,884	71,722	53.0%
	Other	94	125,866	115,351	−8.4%	146	276,716	305,914	10.6%
Transport equipment	Domestic non-private	1,817	1,641,486	1,411,438	−14.0%	969	853,789	633,780	−25.8%
	Foreign	393	214,113	209,314	−2.2%	509	223,419	276,305	23.7%
	Domestic private	136	25,348	32,937	29.9%	523	100,284	136,222	35.8%
	Other	197	146,664	119,849	−18.3%	372	206,784	232,341	12.4%
Computer, telecom. equipment	Domestic non-private	404	321,963	259,482	−19.4%	195	151,497	130,371	−13.9%
	Foreign	900	493,094	599,365	21.6%	1,171	786,394	1,145,731	45.7%
	Domestic private	57	14,461	16,091	11.3%	188	33,237	43,366	30.5%
	Other	107	80,712	78,275	−3.0%	229	187,571	200,413	6.8%
Five industries with lowest capital intensity	Domestic non-private	2,194	717,290	670,741	−6.5%	957	339,690	319,559	−5.9%
	Foreign	3,587	1,442,631	1,642,486	13.9%	4,397	1,926,596	2,364,519	22.7%
	Domestic private	971	214,810	249,444	16.1%	2,493	552,850	646,591	17.0%
	Other	574	244,369	265,919	8.8%	911	376,179	404,507	7.5%
Half with lowest capital intensity	Domestic non-private	15,118	7,213,641	6,130,298	−15.0%	8,041	3,524,487	2,937,217	−16.7%
	Foreign	8,664	3,181,868	3,469,197	9.0%	11,793	4,231,020	5,257,685	24.3%
	Domestic private	4,376	847,083	928,804	9.6%	12,359	2,305,465	2,685,171	16.5%
	Other	3,804	2,341,609	2,155,580	−7.9%	6,708	3,216,926	3,317,835	3.1%
Half with highest capital intensity	Domestic non-private	16,671	7,820,476	6,789,958	−13.2%	8,384	4,176,768	3,435,055	−17.8%
	Foreign	5,618	1,720,839	1,930,965	12.2%	7,520	2,526,663	3,139,195	24.2%
	Domestic private	3,598	607,261	624,209	2.8%	10,025	1,627,668	1,780,204	9.4%
	Other	4,361	2,460,556	2,287,097	−7.0%	7,190	3,542,673	3,612,064	2.0%
Five industries with highest capital intensity	Domestic non-private	4,124	2,388,971	1,955,253	−18.2%	2,353	1,312,382	1,025,351	−21.9%
	Foreign	1,065	287,230	287,392	0.1%	1,456	348,074	358,842	3.1%
	Domestic private	855	123,631	124,516	0.7%	2,331	333,978	341,634	2.3%
	Other	1,204	728,213	686,705	−5.7%	1,910	1,045,451	991,516	−5.2%

The figures show that employment growth in non-private domestic firms has been negative in both periods: firms present in both 1998 and 2001 saw their labour force decline by 14 per cent and the corresponding figure for firms present in both 2001 and 2004 is 17 per cent. The category ‘other firms’ also shows negative employment growth in the first period and a positive but small growth in the second period. Private firms, domestic as well as foreign ones, show positive growth in both periods. In the first period, private domestic firms increased their labour force by 19 per cent, almost twice as much as the increase in foreign firms. The situation changed in the second period when foreign firms increased their labour force by more than 24 per cent, i.e. slightly more than private domestic firms. To sum up, both private domestic and foreign firms have increased their number of employees by two-digit figures in both periods.

Some of the observed differences in employment growth between ownership groups could be caused by differences in the sector distribution of firms. Therefore, we show the development in the five largest sectors in Table 2. The previous results seem to hold at a more disaggregated level: employment has declined in non-private domestic firms and increased, with some exceptions, in private domestic and foreign-owned firms. Finally, we have divided the sample of industries (two-digit level) by the degree of capital intensity. The results are shown in Table 2 and are again consistent with the pattern described above.

5. EMPIRICAL ANALYSIS

a. Direct Effects of FDI

There is clearly a substantial difference in the ability of different types of firms to create jobs. The above figures show that employment growth has been high in foreign-owned firms, even higher in private domestic firms and negative in non-private firms. This pattern is relatively stable when disaggregating to the industry level, and the difference in performance cannot simply be explained by firms with different ownership being active in different sectors of the economy. To shed some light on the underlying causes of employment growth, we model employment growth as a function of firm characteristics, industry characteristics, regional characteristics and conditions at the macro level. This set of variables captures much of the diversity of the Chinese economy, both across industries and regions. Controlling for the differences in the environment faced by firms allows us to estimate the effect of firm-specific characteristics on job growth. The data on small firms is unfortunately quite limited and, in order to retain the small firms in the empirical analysis, we are forced to limit the number of variables in the estimated model. Regional and industry dummy proxies for (near) time-invariant region- and industry-specific characteristics and time dummies capture time-varying economy-wide factors. Since some of the firm characteristics are time invariant (notably ownership), we do not include firm-specific effects in the model. Many of the firm-specific variables are likely to be endogenous in a model for employment and we use the first lag of these variables to protect against endogeneity bias. More specifically, the basic model we estimate is given by

• (1)

where i is index for firms, j is index for industries and t is index for year. The variables included in the model are:

• X_it 
  = Employment.
• Firm_it 
  = A vector of lagged firm characteristics, i.e. firm size measured by employment, the export share of total sales, labour productivity, average wage and capital intensity (see Table A2 in the Appendix).
• Ownership_i 
  = Ownership dummy variables indicating the four ownership categories defined in the previous section.
• Year_t 
  = Year dummy variables.
• Industry_j 
  = Industry dummy variables at the four-digit level (477 industries).
• Province_dummy 
  = 31 province-level dummy variables.

The firm-specific variables control for the most important factors influencing firm performance. Lagged firm size will capture the effect of employment in the previous year on employment growth and export share controls for the importance of access to international markets. Labour productivity and capital intensity control for the efficiency-related factors and the average wage can indirectly capture skill differences between employees. A correlation matrix of the variables is included in the Appendix.

The OLS estimates are shown in columns (1)–(4) in Table 3. The coefficients are relatively stable over the different specifications.¹ Our preferred specification is displayed in column (4) of Table 3, and columns (1)–(3) show more parsimonious specifications. The dummy variable for ‘domestic non-private’ firms is omitted as the reference category. The export share is a particularly interesting explanatory variable since it measures a firm's ability to overcome the constraints of the domestic market and is closely related to whether the firm is foreign owned. To further investigate the effect of the export share, we include interaction terms with the ownership dummies to measure the differential effect of the export share.

Table 3.  Determinants of Employment Growth

	OLS Estimates					Heckman Two-step Estimates
	(1)	(2)	(3)	(4)	(5) Restricted Sample	(6)	(7)	(8)
Notes: Robust standard errors within parentheses. * Significant at a 10 per cent level; ** significant at a 5 per cent level; *** significant at a 1 per cent level.
Dependent variable	Emp. growth	Emp. growth	Emp. growth	Emp. growth	Emp. growth	Survival	Emp. growth	Marginal effects
Foreign (dummy)	0.070*** (0.002)	−0.003** (0.002)	−0.018*** (0.002)	−0.030*** (0.002)	−0.069*** (0.010)	0.293*** (0.006)	−0.011*** (0.003)	0.046
Private (dummy)	0.065*** (0.001)	0.022*** (0.001)	0.021*** (0.001)	0.020 (0.002)***	−0.049*** (0.009)	0.142*** (0.005)	0.030*** (0.002)	0.054
Other (dummy)	0.033*** (0.002)	0.017*** (0.002)	0.016*** (0.002)	0.016*** (0.002)	−	0.178*** (0.005)	0.028*** (0.002)	0.058
Lagged firm size (employment)	−	−0.061*** (0.001)	−0.063*** (0.001)	−0.063*** (0.001)	−0.071*** (0.002)	−	−0.053*** (0.001)	−0.043
Lagged firm size (sales)	−	−	−	−	−	0.168*** (0.001)	−	0.035
Lagged export share	−	−	0.062*** (0.002)	0.040*** (0.003)	0.085** (0.036)	0.174*** (0.096)	0.052*** (0.004)	0.080
Lagged labour productivity	−	0.058*** (0.001)	0.059*** (0.001)	0.059*** (0.001)	0.052*** (0.002)	−	0.070*** (0.001)	0.057
Lagged average wage	−	0.095*** (0.001)	0.094*** (0.001)	0.094*** (0.001)	0.089*** (0.003)	0.108 (0.003)***	0.102*** (0.001)	0.106
Lagged capital intensity	−	0.024*** (0.000)	0.025*** (0.000)	0.025*** (0.000)	0.028*** (0.001)		0.024*** (0.000)	0.016
Lagged export share × foreign dummy	−	−	−	0.041*** (0.004)	−0.001 (0.037)	−	0.038*** (0.004)	0.031
Lagged export share × private dummy	−	−	−	0.012** (0.005)	−0.015 (0.037)	−	0.010** (0.005)	0.008
Lagged export share × other dummy	−	−	−	0.009 (0.006)	−	−	0.006 (0.006)	0.005
Year dummy	Yes	Yes	Yes	Yes	Yes	Yes	Yes	−
Industry dummy (four-digit)	Yes	Yes	Yes	Yes	Yes	−	Yes	−
Industry dummy (two-digit)	−	−	−	−	−	Yes	−	−
Province dummy	Yes	Yes	Yes	Yes	Yes	Yes	Yes	−
No. of observations	640,579	640,579	640,579	640,579	75,977	802,842	640,579	−
R^2	0.01	0.11	0.11	0.11	0.10	−	−	−
Mills ratio	−	−	−	−	−	−	0.166*** (0.016)	−

Column (1) shows that foreign-owned firms have a high growth in employment – about 7 per cent higher than in the reference group non-private domestic firms. It is also seen that both private domestic firms and firms belonging to the group ‘other’ have a higher growth than the reference group. The results are, hence, in accordance with the figures in Table 2 where both foreign and private firms display considerably higher job growth than non-private domestic firms.

Looking at the estimations where we control for various firm characteristics, it becomes clear that the high employment growth in foreign firms is caused by such characteristics: controlling for a number of factors believed to affect employment growth makes growth in foreign firms lower than in non-private firms. In other words, being foreign owned does not in itself cause higher job growth. Instead, these firms display higher job growth because they differ in other firm characteristics. Private firms show a different growth pattern: employment growth is comparably high in these firms even after controlling for firm characteristics. Growth in employment in private firms is about 6.5 per cent higher than in non-private domestic firms without controlling for firm characteristics and about 2 per cent higher after controlling for firm characteristics.

Turning to the other firm characteristics, we find positive and significant coefficient estimates except for the firm size variable. Large firms grow relatively slowly but firms with high labour productivity, high wages and high capital intensities grow relatively fast. The results are in accordance with most previous studies. It might be particularly interesting to note that high capital intensity leads to high employment growth which runs against the commonly held perception that labour-intensive technology generates more employment opportunities. One possible explanation is that capital-intensive technology leads to higher quality or lower prices of products and thereby a stronger growth in employment.²

The coefficient on export share is significant and positive. It implies a relatively strong effect on employment growth, i.e. on average, if export intensity increases by 1 per cent, it will generate an increase in employment growth by 4 per cent for the reference category non-private domestic firms. The interaction terms show a significantly higher effect of 8 per cent for foreign firms, while the effect in private domestic firms is higher than in non-private firms but lower than in foreign firms. It thus appears that foreign firms are considerably more adept at leveraging their access to foreign markets and turning this into job growth.

Ownership is quite heterogeneous within ownership groups with, for instance, different types of joint ventures (see Table A1 in the Appendix). One example is foreign firms, where wholly foreign-owned firms are often said to differ from overseas Chinese firms (Hong Kong, Macau, Taiwan) in important aspects. It also seems slightly arbitrary which group some of the ownership categories are included in. For instance, state-collective jointly operated enterprises might have characteristics similar to non-private firms, rather than to firms in the ‘Other’ category. We therefore use a restricted sample of firms in column (5) where private firms include only wholly-private firms (category 174 in Appendix A1); foreign firms are wholly foreign owned (330); and non-private firms are wholly state-owned enterprises (151). Looking only at this subset of wholly-owned firms, the results change in some important respects. First, whereas growth in employment in wholly foreign-owned firms remains lower (after controlling for firm characteristics) than in wholly non-private-owned firms, the results in column (5) suggest that the same is now true for wholly private firms. Moreover, there is no longer any statistically significant effect from the ownership–export interaction variables.

We continued to experiment with different ownership classifications. For instance, we included dummy variables for all the different ownership categories in Table A1 in the Appendix.³ Growth in employment was by and large lower in wholly state-owned enterprises than in any other ownership category when we did not control for firm characteristics, and higher than in any other category when we did control for firm characteristics.

One methodological problem is, as noted previously, that our panel is unbalanced. We only have information on surviving firms, and ownership might affect the likelihood of survival. The OLS estimates might thus suffer from selection bias. In addition, firm exit implies a job growth of −100 per cent for that firm and period. While it is possible to include this in the dataset, it would lead to huge outliers which can distort the results by themselves. We correct for these problems using the Heckman two-step procedure where firm survival and employment growth, conditional on survival, are modelled as two separate processes. In the first step, we estimate a probit model for firm survival, the selection equation, as:

• P̂(Survival) =Φ(Z_i,t)

• (2)

It is preferable to have different sets of control variables in the selection equation and the employment equation. The firm control variables included in the selection equation are firm size measured by sales, capital intensity, export share and average wage. We also control for the ownership-, year-, regional- and industry-specific effects by including dummy variables. Industry dummies are included at the two-digit level (29 industries). Column (6) of Table 3 reports the estimates of the selection equation. It is noteworthy that foreign firms have a significantly higher survival probability than all other firm categories. Private domestic firms and firms belonging to the ‘Other’ category have also higher survival probability than non-private domestic firms but, again, lower than foreign firms.

The estimated coefficients for the remaining variables, with the exception of capital intensity, have the expected sign and are significant. Turning to the equation for employment growth in column (7), the significant coefficients of the Mills ratio and the implied estimate of the correlation between equations of 0.4 confirm that it is necessary to correct for the sample selection effect and that the previous OLS estimates might be biased. Nevertheless, the Heckman two-step estimates are very similar to the OLS estimates in column (4). Recall, however, that these estimates are conditional on firm survival and do not take account of the employment effect of failing firms. The marginal effects, ∂E(y)/∂x_i, reported in column (8), account for this by also considering the effect on the survival probability of a change in an explanatory variable. Hence, the marginal effect captures both the probability of surviving and growth in employment in surviving firms. The marginal effects are calculated at the sample means of the explanatory variables and reflect a step change from 0 to 1, rather than the derivative for the ownership dummies. The results differ from the previous estimates in one important respect: employment growth is higher in foreign firms than in the reference group of non-private domestic firms. The positive effect is primarily caused by a high survival rate and is found even though we control for firm characteristics in the Heckman estimation. The results for the other variables are broadly in line with previous estimations and it is particularly interesting to see that the other two domestic ownership categories are having a comparable high growth in employment.

b. Indirect Effects of FDI

The above discussion focuses on employment within foreign firms. There are, as previously mentioned, reasons to expect that the entry of foreign firms can have positive as well as negative effects on employment in domestic firms. Positive effects could be caused by the support of linkage industries or demonstration effects, and negative effects could be caused by increased competition.

We try to identify this indirect effect by relating the FDI intensity (measured by the share of sales by foreign firms) of a sector to employment growth in domestic firms. It should be noted that there are several potential problems with this approach. First, the definition of a sector is important. The more narrowly defined is the industry classification we choose, the more weight will be put on the competition effect and the less on the linkage effect. Therefore, we try with industry classifications at both two- and four-digit levels of the Chinese industry classification, which is similar to the industry classification in ISIC Rev. 3. As previously mentioned, the two-digit classification includes 29 industries and the four-digit classification includes 477 industries.

The second related issue is how the geographic distinction of a market should be defined. This is an important issue in such a large country as China. As an example, will a foreign firm in Shanghai use suppliers from the Guangdong province and increase competition for firms in the Guangdong province? There are no theoretical answers to this question and, once more, we adopt a pragmatic approach and use two different geographic classifications, at the national and the regional level, which divide the 31 provinces of China into three regions: east, mid and west.

Taking our previous model (1) as the starting point, we add the FDI intensity and the Herfindahl index as a measure of the competitive pressure as explanatory variables. The ownership dummies and the interactions with the export share are dropped since we estimate the model separately for the subsamples of domestic private and non-private firms. The model for employment growth is thus:

• (3)

where Industry_jt represents the FDI intensity and the Herfindahl index at the two- or four-digit industry level for the relevant region or at the national level. For the estimates of the Heckman sample selection model, we make similar modifications to the selection equation but do not add the FDI intensity or the Herfindahl index.

The results are shown in Table 4 for private firms and Table 5 for non-private firms. For clarity, we have only included the coefficients on the FDI intensity and the Herfindahl index. The coefficients on the other included control variables only changed marginally as compared to the results shown previously in Table 3.⁴

Table 4.  The Effect of FDI on Employment in Private Domestic Firms

		National Level Two-digit	Regional Level Two-digit	National Level Four-digit	Regional Level Four-digit
Note: * Significant at a 10 per cent level; ** ** significant at a 5 per cent level; *** *** significant at a 1 per cent level.
OLS	Herfindahl (lagged)	−1.217** (0.623)	−0.511*** (0.133)	0.048 (0.045)	−0.013 (0.022)
	FDI intensity (lagged)	0.064*** (0.021)	0.089*** (0.017)	0.078*** (0.013)	0.057*** (0.011)
Fixed effect	Herfindahl (lagged)	−0.560 (0.853)	−0.577** (0.280)	0.035 (0.058)	0.030 (0.035)
	FDI intensity (lagged)	0.049 (0.031)	0.016 (0.031)	0.045** (0.019)	0.008 (0.018)
IV	Herfindahl (lagged)	−0.922 (1.419)	−0.676*** (0.142)	0.025 (0.138)	−0.076* (0.041)
	FDI intensity (lagged)	0.004 (0.270)	0.170* (0.099)	0.094 (0.497)	−0.019 (0.181)
Heckman two-step	Herfindahl (lagged)	−1.110* (0.623)	−0.504*** (0.132)	0.045 (0.045)	−0.015 (0.022)
	Marginal effect	−0.896	−0.407	0.037	−0.012
	FDI intensity (lagged)	0.065*** (0.020)	0.091*** (0.017)	0.079*** (0.013)	0.057*** (0.011)
	Marginal effect	0.053	0.074	0.063	0.046
	Mills ratio	0.548*** (0.038)	0.549*** (0.038)	0.549*** (0.038)	0.548*** (0.038)

Table 5.  The Effect of FDI on Employment in Non-private Domestic Firms

		National Level Two-digit	Regional Level Two-digit	National Level Four-digit	Regional Level Four-digit
Note: * Significant at a 10 per cent level; ** ** significant at a 5 per cent level; *** *** significant at a 1 per cent level.
OLS	Herfindahl (lagged)	0.666 (0.521)	−0.057 (0.074)	0.020 (0.038)	−0.030** (0.014)
	FDI intensity (lagged)	−0.001 (0.020)	0.013 (0.014)	−0.003 (0.12)	0.006 (0.008)
Fixed effect	Herfindahl (lagged)	0.895 (0.615)	−0.154 (0.168)	0.041 (0.043)	0.015 (0.022)
	FDI intensity (lagged)	−0.009 (0.026)	−0.032 (0.025)	0.008 (0.015)	0.006 (0.013)
IV	Herfindahl (lagged)	1.685** (0.767)	−0.124 (0.081)	−0.049 (0.063)	−0.062** (0.025)
	FDI intensity (lagged)	−0.258* (0.152)	−0.107* (0.059)	−0.290 (0.289)	−0.191 (0.121)
Heckman two-step	Herfindahl (lagged)	0.651 (0.519)	−0.058 (0.074)	0.019 (0.038)	−0.030** (0.014)
	Marginal effect	0.501	−0.045	0.014	−0.023
	FDI intensity (lagged)	0.002 (0.020)	0.016 (0.014)	−0.003 (0.012)	0.007 (0.008)
	Marginal effect	0.001	0.012	−0.002	0.005
	Mills ratio	0.229*** (0.028)	0.226*** (0.028)	0.225*** (0.028)	0.226* *** (0.028)

The OLS estimates show that FDI tends to increase employment growth in domestic private firms (Table 4) irrespective of at which level of aggregation the FDI intensity is measured. Concentration, as measured by the Herfindahl index, only yields a significant effect when we use the two-digit industry classification where increased concentration (lower competition) decreases employment growth. The estimated effects are, in most cases, smaller and insignificant when controlling for unobserved heterogeneity across firms by including fixed firm effects, and the FDI intensity is only significant at the national four-digit level.

There is a distinct possibility that the lagged FDI intensity is endogenous, i.e. that foreign firms invest in certain industry sectors or regions in anticipation of a favourable development. To control for the possible endogeneity, we also estimate equation (3) with instrumental variables using instruments calculated at the industry and regional levels. Specifically, we instrument the FDI intensity with the R&D and import intensities for the industry sector (four or two digit) at the national or regional level as well as the industry's and region's share of total patent applications, share of government S&T funding and the share of foreign S&T funding. (A more detailed variable definition can be found in Table A2 in the Appendix.⁵) The instrumental variable (IV) estimates for private firms are reported in the third block of Table 4. Correcting for the possible endogeneity of FDI, the effect of the FDI intensity is only significant at the regional two-digit level.⁶ There is, however, little evidence that the FDI intensity is endogenous for private firms. We test the null hypothesis that the lagged FDI intensity is exogenous with a difference-in-Sargan test and fail to reject this at the 10 per cent level for all specifications. The difference in results compared with OLS can thus be explained by the lower efficiency of the IV estimator.

Finally, in the Heckman two-step estimation, the qualitative results are identical to the OLS estimates and the estimates differ little. Still, the statistically significant coefficients on Mills ratio suggest that it is necessary to correct for the sample selection bias caused by attrition. Taking account of the possibility that firms might cease to exist, the marginal effects are close to the two-step estimates due to the high probability of survival.

To sum up, the estimated effect of FDI on employment in private domestic firms varies by specification. An overall assessment of the results suggests that FDI seems to increase employment in private domestic firms within the same sector. A conservative assessment would instead be that there are at least no signs of a negative effect. One could, of course, claim that any negative effect of FDI might be captured by the Herfindahl index if FDI affects industry concentration. We therefore tried alternative estimations where the Herfindahl index was excluded. The results remained robust and similar to those shown in Table 4, implying that inclusion of the Herfindahl index is of little importance for the estimated positive effect of FDI.⁷

Considering the different effects of national and regional presence of FDI and of FDI in the same two- or four-digit industry, it is difficult to find any strong pattern. We would expect the negative competition effect from FDI to be relatively important at the four-digit level and the positive linkage effect to be relatively important at the two-digit level, but there is no strong support for this hypothesis in Table 4. There are some indications that the positive effect is stronger at the two-digit level at a regional level and at the four-digit level at a national level. That FDI even within the same four-digit industry might have a positive effect on employment in private domestic firms leads us to the much studied and debated issue of spillovers from FDI. The positive effect within narrowly defined industries is consistent with the existence of such spillovers. Swenson (2007) finds evidence of spillovers in terms of export behaviour in China. She argues that this is caused by information on foreign markets and technologies flowing from foreign to domestic firms. Such flows could stem from demonstration effects or job turnover when employees in foreign firms join domestic competitors. It is also worthwhile to mention the literature on spillovers in China that looks at productivity in domestic firms. There are several such studies, and some of them find a positive effect of FDI on the productivity of domestic firms (see e.g. Cheung and Lin, 2004; Hale and Long, 2007). It is plausible that more productive firms will grow faster, as is seen in our econometric results.

Next we turn our attention to the effect of FDI on employment growth in non-private firms. The results are shown in Table 5 and differ substantially from those in Table 4. We find no signs of a positive effect of the lagged FDI intensity. The estimates are in general small and insignificant. The only significant (at the 10 per cent level) estimates for the FDI intensity are the IV estimates⁸ at the national and regional two-digit level, which are both negative.

It should, perhaps, not come as a surprise that foreign firms have a positive effect on private firms, but impose less of an effect on non-private firms, since the latter are still to a large extent operating outside normal market economic restrictions.

6. CONCLUDING REMARKS

FDI is considered to be one of the key driving forces behind the spectacular economic growth in China in the last two decades. However, academic research and public policy discussions tend to ignore the effect of FDI on employment. This is unfortunate, considering the large importance of employment growth in developing countries. This paper contributes to this issue by providing empirical results on the effect of FDI on employment in China, based on a large firm-level dataset for the period 1998–2004.

The descriptive statistics suggest that both FDI and private domestic firms have relatively high employment growth, as compared to non-private domestic firms. The cross-ownership comparison also shows that foreign firms, in general, have more advantageous firm characteristics as compared to firms with other kinds of ownership.

It is important to distinguish between the sources of favourable employment growth. Is it a pure ownership effect and/or an outcome of other firm characteristics that may yield a positive effect of employment growth? In the first step of our econometric analysis, we investigate the direct ownership effect and find that employment growth is strongly correlated with firm characteristics such as high productivity, capital intensity and wage. Furthermore, the higher export share, as a proxy for access to international markets, seems to give foreign firms additional competitive advantages as compared to domestic firms. Finally, foreign firms are considerably less likely to exit the market than any domestic firms. Controlling for survival ratios shows that employment growth in foreign firms is relatively high even after taking firm characteristics into account.

In the second part of the analysis, we look into the indirect effect of FDI in terms of spillovers and competition. From a theoretical point of view, the effect of FDI on employment can be both positive and negative, depending on the strength of the spillover effect and competition, which are simultaneously at work. Interestingly, we find that the spillover effect of FDI seems more important than the competition effect, in particular on private domestic firms and even at a highly disaggregated industry level. In contrast, such a positive indirect effect of FDI is not observed among non-private domestic firms.

Based on the empirical analysis, we conclude that FDI has contributed to employment in the Chinese manufacturing sector through its access to international markets and other firm characteristics which favour growth in employment, and through positive effects on employment in private domestic firms.

Footnotes

• 1

  We did also experiment with including industry dummy variables at a two-digit level or by including regional (east, mid and west) instead of provincial dummy variables. These changes had no major impact on the results.

• 2

  Note that it could still be that the level of employment is higher in labour-intensive sectors and firms.

• 3

  The results are not shown but are available on request.

• 4

  The complete results are available from the authors on request.

• 5

  Naturally, it can also be argued that the FDI variables are endogenous in the model (1) used to assess the direct effects of FDI on employment growth. The paucity of data on small firms prevents us from constructing appropriate firm-level instruments that can be used with equation (1). When the FDI intensity is measured at the industry/regional level, our data allow for the construction of instruments at the industry/regional level and we take advantage of this to assess the endogeneity of FDI.

• 6

  The Sargan test of the overidentifying restrictions rejects the validity of some of the instruments for some combinations of industry and regional classifications, and the R&D intensity is dropped for the national two-digit level.

• 7

  The results are not shown but are available on request.

• 8

  In contrast with the results for private firms, we reject the exogeneity of the FDI share at the 10 per cent level for the regressions at the national and regional two-digit level. The exogeneity of some of the potential instruments is also rejected and we drop the patent share and foreign S&T funding for the national two-digit level and the patent share for the regional two-digit level.

APPENDIX

Table A1.  Ownership Classification

Ownership	Code	Definition
Note: Overseas Chinese refer to firms from Hong Kong, Macau and Taiwan.
Non-private	110	State-owned enterprises
	120	Collective-owned enterprises
	130	Shareholding cooperatives
	141	Stated-owned, jointly operated enterprises
	142	Collective-owned, jointly operated enterprises
	151	Wholly stated-owned enterprises
Private	171	Wholly private-owned enterprises
	172	Private-cooperative enterprises
	173	Private limited liability enterprises
	174	Private shareholding enterprises
Foreign	210	Overseas Chinese joint venture
	220	Overseas Chinese cooperative
	230	Wholly overseas Chinese-owned enterprises
	310	Foreign joint venture
	320	Foreign cooperative
	330	Wholly foreign-owned enterprises
Other	143	State-collective jointly operated enterprises
	149	Other jointly operated enterprises
	159	Other limited liability enterprises
	160	Shareholding limited enterprises
	190	Other enterprises

Table A2.  Construction of Variables

Variable Name	Definition
Firm-level variables
Employment growth	log (number of employees)t− log (number of employees)t−1
Firm size	log (number of employees)
Labour productivity	log (real value-added/number of employees)
Average wage	log (real annual wage bill/number of employees)
Capital intensity	log (real capital stock/number of employees)
Export share	Export/total sales
Industry/regional-level variables
FDI intensity	The share of sales by FDI firms in total sales at the four-digit or two-digit industry levels
Herfindahl index	, where Si is the market share, in terms of sales of the ith firm in industry j at the four-digit or two-digit level
R&D intensity	R&D expenditure to sales ratio at the industry/regional level
Technology import intensity	Technology import expenditure to sales ratio at the industry/regional level
Patent share	The industry/region's share of total patent applications
Government S&T funding	The industry/region's share of total government S&T funding
Foreign S&T funding	The industry/region's share of total foreign S&T funding

Table A3.  Correlations between Main Variables

	Employment Growth	Firm Size	Export Share	Produc- tivity	Wage	Capital- intensity	FDI 2-digit	FDI 4-digit	Herf 2-digit	Herf 4-digit
Employment growth	1
Firm size	−0.165	1
Export share	0.040	0.168	1
Productivity	0.229	−0.24	−0.008	1
Wage	0.203	−0.02	0.133	0.409	1
Capital- intensity	0.130	−0.02	−0.131	0.339	0.228	1
FDI 2-digit	0.030	0.048	0.322	−0.013	0.114	−0.139	1
FDI 4-digit	0.046	0.028	0.356	0.046	0.147	−0.010	0.707	1
Herf 2-digit	0.010	0.071	−0.008	0.071	0.086	0.049	0.193	0.114	1
Herf 4-digit	0.000	0.024	0.022	0.042	0.056	0.0150	0.107	0.062	0.211	1