FOREIGN DIRECT INVESTMENT AND SPILLOVER EFFECT: EVIDENCE FROM CHINA

FOREIGN DIRECT INVESTMENT AND

SPILLOVER EFFECT: EVIDENCE FROM

CHINA

Master Thesis of International Economics and Business

Faculty of Economics

Rijksuniversiteit Groningen

Landleven 5, Groningen

9747 AD, the Netherlands

ACKNOWLEDGEMENT

This thesis is the result of two years of work whereby I have been accompanied and supported by many people. It is a pleasant aspect that I have now the opportunity to express my gratitude for all of them.

The first persons I would like to thank are my parents, who provide constant support to me from the very beginning when I planned to do master studies in Rijksuniversiteit Groningen. I can never express too much gratitude to them. Secondly, I want to thank my thesis supervisor, Dr. J. Zhang, who has always been ready for my question and given me a lot of suggestions accompanying the whole process of my thesis. My thesis coassessor, Dr. D. Dikova, also offered me due responsibility and many very useful comments to polish my thesis. Besides, there are still many professors from Faculty of Economics and Management and Organization, who, during the two years of my study, have helped me to gain all knowledge to contribute to this final work. Finally, I would like to express thanks to my classmates, S. Agyemang, G. Batzios, H. D. Cruz, J. J. Kamote, P. Levchuk,

FOREIGN DIRECT INVESTMENT AND

SPILLOVER EFFECT: EVIDENCE FROM CHINA

Abstract: In this paper we investigate empirically whether foreign direct investment

benefits local components in the sector, in which it resides, and in surrounding sectors through productivity spillover. Using data covering 35 industrial sectors over the period from 1995 to 2003 in China, we do not find significant support for intra-industry spillover. But there is evidence for positive FDI spillover from 34 sectors to the remaining one and somewhat negative FDI spillover from one sector to the others. And if many sectors are taken as a whole, there is little evidence of positive association between FDI presence and productivity of local industries.

CONTENTS

1. Introduction………...5

2. Literature Review………..7

2.1. Theory on FDI spillover……….7

2.2. Empirical evidence……….8

3. Methodology………12

3.1 Empirical model………12

3.2 Data and measurement…...………...14

3.3 Econometric specification……….15

4. Results and Discussion………20

5. Conclusion………...24

INTRODUCTION

Host country effect of FDI has always been a hot issue for governments of all countries, who are fascinated with positive benefits brought by FDI inflow to their countries while fearing that its side effects, predictable or not, can grow out of their control. Navaretti and Venables (2004) aggregate empirical findings concerning three main aspects of FDI host country effects, factor market, employment volatility and behaviour of domestic firms. Firstly, they find most empirical studies (Griffith and Simpson 2001, Aitken, Harrison and Lipsey 1996, etc.) conclude that MNEs pay higher wages than local firms. The reason might be that multinational enterprises (MNE) employ higher-skilled labour than local firms. And there is no support for the view that wage can also decline because MNEs are large employers, especially in developing countries, who might enjoy monopolistic position and have strong bargaining power. Secondly, they find no significant support for the view that MNEs are generally more volatile employers than local firms. Following a labour demand shock they do not react by reducing employment to a much lower level but turn out to be more likely to preserve their employees though adjusting more quickly (Navaretti, Checchi and Turrini 2003, etc). Thirdly, MNEs can generate technological and pecuniary externality to local firms in the form of better performance when they are operational in the market and become part of the network of value adding chain. However, empirical studies produce mixed results: some suggest there is significant positive effect of FDI on local firms in host countries (Blomström and Persson 1983, Kokko 1994, etc.) while some others find insignificant or even negative effect (Aitken and Harrison 1999, Djankov and Hoekman 2000, etc.).

of national export and import compared to 39.1% in 1995. And FIEs’ share of value added in all industrial sectors rose from 14.8% in 1995 to 27.6% in 2003. Lemoine (2000) Daya-Gulati and Husain (2000) find that higher per capita income in coastal region of China can be attributed to high rate of foreign investments there, during the 1990s when FDI poured into the coastal areas (as also happened in the northern and northeastern regions when FDI began to flow there in the late 1990s). Despite the remarkable growth of FDI in every aspect, the role of FDI becomes more and more controversial due to concern about security and independence of national economy. It is feared that with various policy incentives towards FIEs, local firms are in a disadvantaged position of market competition and consequently will be crowded out. The share of sales revenue by FIE grew from 17.0% in 1995 to 30.5% in 2003 in 39 industrial sectors. In particular, in 2004 the top five FDI-intensive sectors, in terms of output share and value added, are Electronic and Telecommunications (73.36%)1, Instruments, Meters, Cultural and Official Machinery (61.84%), Stationery, Educational and Sports Goods (59.08%), Leather, Furs, Down and Related Products (53.23%), and Furniture Manufacturing (47.33%) (Southcn.com, 2004). Additionally, Young and Lan (1997) find in a case study that considerable local partners in joint ventures with foreign investors have distorted motives. They do not aim to gain anything from foreigners but to access favourable polices offered by the Chinese government.

Since the main purpose behind the FDI promotion policy is the strategy of China central government, ‘exchanging market for technology’. To evaluate this strategy we need to know whether there has been any technology advancement of industries in host country as a result of FDI presence when certain market share has already taken by FIEs. Hence, this paper aims to investigate if there is productivity spillover from FDI to local industries in China, which is defined as unintentional and indirect impact of FDI operation on efficiency and growth of economic units in host country (He, 2000). In another word, do Chinese local industries benefit from FDI presence through exhibiting a higher productivity? Next section introduces theories and empirical studies on technology spillover a. In the third section, we discuss the methodology of empirical analysis. In

Section 4 results and their implications will be presented and discussed. Finally, we reach our conclusion of this paper.

LITERATURE REVIEW

There are plenty of theories explaining why there is FDI spillover to local economic units and the most persuasive explanations are those that emphasis the co-existence of proprietary knowledge of some form and market failures in protecting that knowledge (Görg and Greenaway, 2004). To compete with local firms multinational enterprises (MNEs) have to have some unique assets that cannot be easily imitated. Otherwise MNEs are not able to survive in local market since they are already in a disadvantaged position of market information and local institutions compared with their local counterparts (Love, 1997). In another word, MNEs must possess more or less ownership-specific (O) advantage over their domestic competitors. This O advantage can be in the form of tangible assets, such as natural endowments, manpower and capital, and intangible assets such as technology and information, managerial, marketing and entrepreneurial skills (Dunning, 2001). And because it is almost impossible for owners to fully appropriate their O assets, which turns out to be the source of spillover and can be diffused to the industry where FDI has establishment or surrounding industries that are components of the same value added chain.

The fourth source of gain might be via export spillovers as suggested by the two scholars. Indigenous firms can learn to export from FIEs and thus are trained in international competition. Formation of joint venture and various cooperation projects between foreign investors and local firms can also facilitate intra-industry spillover due to physical or technological proximity.

Another source that local firms in host country can benefit from FDI presence is inter-industry spillover. The new technology brought by foreign investors may stimulate local suppliers of intermediate products to improve product quality and lower cost in order to tailor to the demand of FIEs and thus compete for FIEs market. New products introduced by FIEs may also stimulate productivity improvement of local firms purchasing their products (Blomström, 1991). He also claims that spillovers to the multinationals’ down-stream firms will be more important in the future. The reason is that newly emerging technology, such as new generation of computer based automation and information technology, are generally knowledge and research intensive, and expensive to develop, that only a few, large firms (MNEs) can afford such efforts. Small countries or enterprises facing technology revolution, thus, have to accept certain degree of dependence on MNEs’ technology. For them it is more realistic to use than to develop innovation, as shown in some small European countries (Blomström and Meller, 1991).

Empirical Evidence

TABLE 1

Papers on intra-industry Spillover (Source: Görg and Strobl, 2001)

Note: ‘CS’ denotes cross-sectional data, while ‘Panel’ denotes use of combined cross-sectional time-series data in the respective analysis.

‘+’ and ‘-’ refers to positive and negative results of spillover effect respectively.

We find five studies on intra-industry spillover with Chinese sample, which is not included in the discussion of Görg and Strobl (2001). Among them, two utilized panel data, two cross-sectional data, and one case study mentioned in previous section. With sector-level data of 29 manufacturing industries in Shenzhen Special Economic Zone over the period from 1994 to 1998, Liu (2002) investigates intra-industry effect by regressing value added of a sector on FDI share in the same sector, without distinguishing value added by national enterprises or FIEs, and the coefficient for FDI presence remains insignificant across various model specifications. He then includes an interaction term, the product of FDI share in one sector and average FDI share in all sectors, to test if sectors dominated by domestic firms benefit from FDI presence. But the result is still inconclusive. The other panel analysis done by Hu and Jefferson (2002) utilized firm-level data but from only two industrial sectors (textile and electronics). They find that in the short run FDI tend to reduce both productivity and market share of domestic firms. But in the long run, the negative effect on productivity of domestic firms disappears in both sectors. And both cross-sectional studies produce positive results of spillover effect (Chuang and Hsu 2004, and Wei and Liu 2003). In sum, no panel studies have been conducted with Chinese data covering all regions and sectors and this is what we are going to investigate.

• H1: there is intra-industry FDI spillover within a single sector.

added is not distinguished between national and foreign invested enterprises as mentioned above, there is another point that might draw misleading conclusion. A positive coefficient for MFDI may not be a reflection of inter-industry spillover from surrounding sectors to sector i but industry spillover if there is significant intra-industry spillover within sector i but no or weak inter-intra-industry spillover from all other sectors to sector i. In this case, one would conclude that there is empirical support for both intra- and inter-industry spillover even if the evidence of latter is dubious. To avoid this situation, due to the difficulty to identify all upper- and down-stream industries for one sector and to specify the magnitude of impact of a single surrounding sector, we investigate spillover between one sector and all others. Specifically, in this paper we regress value added in all but sector i on FDI presence in sector i to capture inter-industry effect and the other way around. Since sources and recipients involves different sectors in the process of inter-industrial spillover, there is need to specify the direction of this effect. We would also like to investigate the total effect of FDI presence on many sectors as a whole. Since there might be both negative and positive spillover effects, it is interesting to know what the net effect is.

• H2: There is FDI inter-industry spillover from one sector to all others. • H3: There is FDI inter-industry spillover form all other sectors to one sector. • H4: Average FDI presence in many sectors has generally a positive effect in

productivity of those sectors.

panel data at firm level, Aitken and Harrison (1999) identify two effects of foreign direct investment on Venezuelan plants. First, increases in foreign equity participation are correlated with increases in productivity for small enterprises. Second, increases in foreign ownership negatively affect the productivity of wholly domestically owned firms in the same industry, the magnitude of which appears large and robust to alternative model specification. The net impact of foreign investment, taking into account these two offsetting effects, is quite small. And the gains from foreign investment appear to be entirely captured by joint ventures. The unique contribution of this paper is, by utilizing national-wide Chinese industry-level data, an investigation of intra-industry spillover and, what is more important, a first attempt to measure inter-industry spillover from one sector to the others and the other way around.

METHODOLOGY

Previous studies investigate spillover by checking the association of FDI presence and labour productivity or total factors productivity (TFP). Estimation of FDI effect on labour productivity is based on a partial measurement of productivity, as output per worker can also be affected by the use of other factors of production that are not taken into account by measure of labour productivity. Thus, this paper attempts to investigate FDI effect on TFP. We follow Griffith (1999) and Liu (2002) to use Cobb-Douglas production function,

β α_K AL Y =

(1) where Y denotes output, L and K labour and capital input, A technical factors, α elasticity of labour and β elasticity of capital. By following Liu (2002), we decompose A, which includes technical factors, to include FDI presence as follow.

A = B FDIγ INDδ (2)

lnY = lnB + γlnFDI + δlnIND + αlnL + βlnK (3) Since we control for labour and capital inputs, a higher level of output is entirely due to productivity improvement. This means a positive and significant γ indicates a positive association between FDI presence and productivity. And since we aim to test both intra- and inter-industry spillover effects from FDI to local industries as stated in four hypotheses in previous section, we expand equation 3 to following two econometric models,

lnVAit = lnBBit+ γ1lnFDIit+ γ2lnAFDIjt+ δ2lnSOEit+ α1lnLit+ β1lnKit + εit (4)

lnVAjt = lnBBit+ γ3lnFDIit+ γ 4lnAFDIjt+ δ4lnASOEjt+ α2lnLjt+ β2lnKjt + εit (5)

where VAit denotes total value added by local firms in sector i and VAjt total value added

by local firms in all sectors except i. FDIit denotes FDI input share in sector i and AFDIjt

average FDI input share in all other sectors. SOEit and ASOEjt denote state-owned

enterprises’ input share in sector i and state-owned enterprises’ input share in all other sectors. Lit and Ljt denotes total labour input of local firms in sector i and in all other

sectors respectively while Kit and Kjt denotes to total capital input of local firms in sector i

and all other sectors. A positive and significant γ1 is supportive of H1, since FDI presence

in one sector is positively associated with total value added by local firms when labour and capital input is under control. For the same reason, a positive γ2 provides support for

H3 since AFDIjt quantifies FDI presence in other sectors. And a positive γ3 supports H2

when we regress total value added by local firms in other sectors on FDI presence in one sector. Finally, a significant and positive γ4 provides evidence for H4.

The only variables we introduce into equation 4 and 5 to represent industrial characteristics in equation 3 (IND) are SOEit and ASOEjt State-owned enterprises are an

VAjt). In addition, we do not add any variable to control technology difference between

FIEs and their local counterparts, such as R&D intensity and labour-capital input ratio. An FIE in local market is different from an indigenous competitor not just because it is of a different type of ownership but also of a different buddle of characteristics, such as types and combinations of inputs, size, etc., which accompany ownership (Navaretti and Venables, 2004). And this also applies to foreign and local parts of Chinese industry if we extend the focus from firm to sector level. If we control all technology factors, there might be no difference between foreign and local components that make the two distinguishable from each other without checking their origin, not to mention that there are many unobservable differences that are difficult to quantify, such as method of management. Measurements of these variables are presented after appropriate sample is chosen.

Data and Measurement

Our population under study is all industries in China with FDI presence. All sample information comes from China National Bureau of Statistics, the website of which (www.stats.gov.cn/tjsj/ndsj) publishes a yearbook recording statistics from all economic aspects from 1995 (yearbooks before 1995 are only available in library of the bureau). The industry chapter of this book, together with other datasets in this website, provides a rich source of industry level data. 35 out of 39 third-digit industrial sectors in the period from 1995 to 2003 are taken as the sample2_{. We use input share of foreign invested}

enterprise in the year book to proxy FDI presence since there are mainly three types of FIEs that involve FDI. Besides fully owned subsidiaries by MNEs, the other two forms are joint venture and cooperation operation enterprises with foreign participation. Yearbook 1998 does not include sector level indicators for FIEs as yearbooks in other years. As a result, there are missing values for six variables related to FIEs in 1998, FDIit,

AFDIit, SOEit, and ASOEjt. As there is much bigger gap between values of variables

between 1997 and 1999 than that between any two consecutive years and the definitions of various variables in 1998 have not changed, the missing values are estimated with

2 _{Observations from four sectors are excluded due to missing values. See a complete list of sectors in}

TABLE 2

Variables and Measurements

Dependent variables

VAit total real value added by local firms of sector i in year t3 (108 RMB yuan)*

VAjt total real value added by local firms of all sector except i in year t (108 RMB

yuan)

Independent variables

FDIit total registered assets of FIEs divided by total asset of the sector in year t

AFDIit total registered assets of FIEs divided by total asset of all sectors except i in

year t Control variables

SOEit total registered assets of state-owned enterprises divided by total asset of sector

i in year t

ASOEjt total registered assets by state-owned enterprises divided by total assets of all

sectors except i in year t

Lit number of employees owned by local firms of sector i in year t4 (104 persons)*

Ljt number of employees owned by local firms of all sectors except i in year t (104

persons)*

Kit total capital assets owned by local firms of sector i in year t (108 RMB yuan)*

Kjt total capital assets owned by local firms of all sectors except i in year t (108

RMB yuan)*

Note: * denotes the value of variable X of all local firms in one sector and it is calculated as Xall firms - XFIEs

average values of variables for 1997 and 1999. Hence, the balanced sample of this study comprises a total of 315 observations. With sample, we choose measurement for each variable and present it in Table 2.

Econometric Specifications

There are mainly three different models for panel analysis, pooled regression model, fixed effects (FE) model, and random effects (RE) model, among which there might be some ‘sub-models’ due to problems of heteroskedasticity and autocorrelation.

FE, RE, and pooled regression We assume FE model to be the best specification for our study for two reasons. Firstly, pooled regression model cannot be the suitable one for this study because 35 sectors, which have different characteristics not captured by

3 _{We follow Liu (2002) to use a price deflator for each year, the ratio of gross output at current and constant}

(1990) prices.

explanatory variables, such as speed of industrial cycle and X-efficiency5, can have different effects on value added. Secondly, RE model is not the best candidate either since it views the cross-sections, on which we have data, as a random sample from a larger population. However, the 35 industrial sectors are not random sample drawn from a population but nearly exhaustive and exclusive cross-sections. Hence, the dummy variables of different sectors are treated to be unknown but fixed parameters. We make inferences only about the sectors in our sample. In another word, there is no ‘superior sectors’ for the sector dummies in this research to extend to. The three statistical models differ mainly in their assumptions concerning the intercept and error terms. It is assumed that the error term in an equation can be decomposed into two independent elements:

(6) it i it =u +v ε

where ui is time-invariant and accounts for any unobservable industry specific effects not

captured by explanatory variables. The term, vit represents the remaining disturbance, and

varies over cross-sections and time. In a pooled regression model, the ui’s are assumed to

take the same value for all cross-sections. However, this assumption does not hold in any panel analysis in FDI spillover. Both FE and the RE models accommodate unobservable heterogeneity. In the former, the ui’s are fixed parameters to be estimated with dummies

for time or cross-sections, while in the RE model, the ui’s are assumed to be random,

independently and identically distributed, i.e. ui~N(0, σ u 2). The RE model is better than

the FE model if the unobservable effects are uncorrelated with regressors, because dummies in a FE model can exhaust degrees of freedom. To decide which model is more appropriate for this study, FE or RE, we follow Liu et al. (2000) to conduct two diagnostic tests to check the validity of the assumptions accompanying the two models. They are likelihood ratio (LR) test for the pooled regression model against the FE model and Hausman specification (HS) test for the RE model against the FE model. Rejection of both hypotheses favours FE model6. Besides, some events happened during the period from 1995 to 2003 can have significant impact on FDI and thus our results, such as return

5 _{In economics, X-efficiency is the effectiveness with which a given set of inputs are used to produce}

outputs. If a firm is producing the maximum output it can, given the resources it employs, such as men and machinery, and the best technology available, it is said to be X-efficient.

TABLE 3 FE and RE Models

FE.Eq.(4) FE.Eq.(5) RE.Eq.(4) RE.Eq.(5) FDI 0.103** (0.036) 0.002* (0.001) 0.091** (0.018) 0.002** (0.000) AFDI -2.936** (0.588) 0.119** (0.033) 1.876** (0.335) 0.165** (0.016) SOE -0.276** (0.088) -0.174** (0.059) ASOE 0.0002 (0.0002) -0.003 (0.002) L 0.230* (0.109) 0.107 (0.093) AL 0.087** (0.032) -0.157 (0.094) K 0.611** (0.098) 0.888** (0.100) AK 0.292** (0.060) 0.527** (0.059) Adjusted R2 0.979 0.999 0.648 0.541 Test LR LR HS HS Cross-section effect 619.6** 663.6** 23.95** 173.3** Degrees of freedom 34 34 5 5 Period effect 64.3** 1674.5** 23.947** 169.5** Degrees of freedom 8 8 5 5 Note: FE model is applied in column 2 and 3 and RE model in column 4 and 5; Figures in parentheses are standard deviations;

* and ** denote significance at the 5% and 1% levels respectively;

White cross-section covariance estimators are utilized for FE model but not for RE model because HS test can only be carried out without White estimators;

When we carry out HS test for Equation 3, we find insignificant test statistics for period (chi-square 10.345, p-value .0660) or cross-section (chi-square 8.155, p-value .1479) alone. But when we combine two effects we find significant statistics as shown in column 4. Since we want to control both industry and year effects, we accept the result of joint effects. When we carry out HS test for Equation 4, we find significant statistics as shown in column 5 but Eview 5.1 refuses to produce results for joint effects. Hence, we accept results for two independent effects.

of Hong Kong, accession to WTO as well as Asian financial crisis. For this reason, we also need to decide a specification to control period effects besides testing a best model specification for cross-sections. The statistics of the two tests will be presented with estimation output.

As the results show, LR statistics for two equations are significant (619.6 and 663.6), rejecting pooled regression in favour of FE model. HS statistics are significant too (23.95 and 173.3), rejecting RE in favour of FE model. Hence, hereafter we apply FE model for all regressions for cross-sections. And the results are the same for year effect, favouring FE model, as the test statistics show.

Heteroskedasticity and Autocorrelation Since the dataset is organized in the form of cross-sections in each time point (year in this case), our regression can easily suffer from heteroskedasticity. Cross-sectional data invariably involves observations on economic units of different sizes, different industrial sector with different number of employees for instance. And frequently, the larger a cross-section is, the more difficult it is to explain the variation in some outcome variables by the variation in a set of explanatory variables because larger cross-sections, such as firms and households, are likely to be more diverse and flexible with respect to the way in which values for the dependent variable are determined (Hill, Griffiths, and Judge 2001). Hence, many studies with cross-sectional or panel data are found to try to curb heteroskedasticity and weighted least squares (WLS) is a model specification most frequently applied. Weights are the share of each plant in total annual industry output in the study done by Aitken and Harrison (1999), and square root of the number of firms in different groups of manufacturing enterprises by Liu (2002). Similarly, we need to assign weights to different industrial sectors to trim heterogeneity of sector size. However, we are not able to find an appropriate candidate for weight along cross-sectional dimension. The residual of OLS barely shows any correlation with change in any explanatory variable measuring size of a sector, number of employees, number of firms, or average number of employees per firm. Hence, another technique, White cross-sectional covariance estimators, is utilized, which correct standard deviation of no longer B.L.U.E. (best linear unbiased estimator) least squares caused by heteroskedasticity.

the assumption of FE model that errors are independent and distributed for all individuals and in all time periods (

) , 0 ( 2 ε σ N

Hill, Griffiths & Judge, 2001). To test the presence of autocorrelation we apply Lagrange Multiplier (LM) test. The null hypothesis of LM test is ρ=0 in Equation 7. (7) it t i i it = ρε ,−1+v ε

Where ρ_i is a parameter that determines the correlation properties of error term ε_it in sector i, and are uncorrelated random variables with a constant variance. If Hv_it o is

rejected, we will transform our model by implementing GLS along time dimension. First, we run regression of Equation 4 and 5 with fixed industry and fixed year dummies. Second, we regress dependent variables on explanatory variables and one year lagged error term obtained from the first step. Then, we find the one year error term is significant in both equations (t-value of the error term is 7.705 and p-value .000 for Equation 4, and t-value of the error term is 3.326 and p-value .001 for Equation 5). Hence, we reject the null hypothesis that ρ=0 in Equation 6 and conclude the presence of autocorrelation. To tackle it, we implement GLS along time dimension for both regression equations. Specifically, we estimate ρ for each industry and transform variables into

1 2 *

1 1 i i

i x

x = −ρ for observations in 1995 and for observations in the remaining eight years. In this way transformation does not have to sacrifice observation of the first year and the transformed errors have the same variance as the errors (

1 , * − − = _{it i it} it x x x ρ it v Hill,

Griffiths & Judge, 2001).

In consideration of the possibility that it can take a substantial period of time for FDI and SOE to take effect on output and thus productivity, we also regress dependent variables on up to three-year lagged value of explanatory variables, FDIit, AFDIit, SOEit, and

ASOEjt. But we do not regress dependent variable on current and lagged value of an

independent variable in one equation since coexistence of these terms on the right hand of an equation creates serious collinearity. We run an auxiliary regression as follow

3 , 2 , 1 , ln ln ln

TABLE 4

Results of Auxiliary Regression

Coefficient LOGFDI(-1) 0.946** (0.126) LOGFDI(-2) -0.453* (0.188) LOGFDI(-3) 0.424* (0.168) Note: Adjusted R-square is 0.999.

Fixed dummies are added for both periods and industries. White cross-sectional covariance estimators are utilized.

TABLE 5

Correlation Coefficients of Explanatory Variables in Equation 4

lnFDIit lnAFDIjt lnSOEit lnLit lnKit

lnFDIit 1 lnAFDIjt -0.022 1 lnSOEit -0.480 -0.032 1 lnLit 0.097 -0.322 0.090 1 lnKit 0.041 0.007 0.423 0.783 1 TABLE 6

Correlation Coefficients of Explanatory Variables in Equation 5

lnFDIit lnAFDIjt lnASOEit lnALit lnAKit

lnFDIit 1

lnAFDIjt -0.022 1

lnASOEit -0.006 0.040 1

lnALit -0.080 -0.081 -0.603 1

lnAKit 0.106 0.093 0.331 -0.810 1

and present the results in Table 4. Hence, we run three regressions for both Equation 4 and 5 with independent variable of one year in each. And there is no serious correlation among explanatory variables except for the relationship between labour and capital input as showed in Table 5 and 6 below. It is not difficult to understand that there must be some labour-capital ratio in production

and the four variables work only as control and do not affect the main findings of this paper. In addition, we do not lag capital and labour input because capital and labour are production inputs that directly decide output and productivity of the current year.

TABLE 7

Regression Results of Equation 4

No lag One-year lag Two-year lag Three-year lag FDI 0.199** (0.033) 0.079 (0.084) -0.047 (0.090) -0.024 (0.074) AFDI 1.341** (0.328) 2.639** (0.367) 2.866** (0.419) 2.965** (0.211) SOE -0.112 (0.177) 0.011 (0.148) 0.219* (0.091) 0.066 (0.133) L 0.001** (0.000) 0.119 (0.104) 0.085* (0.041) 0.053 (0.063) K 1.083** (0.068) 0.473 (0.315) 0.168 (0.296) 0.048 (0.295) Adjusted R2 0.989 0.996 0.998 0.998 Sample 378 294 252 210

Note: Fixed dummies are added for both industries and periods. GLS along time dimension is utilized.

We first present the regression results of Equation 4 in Table 7.

For regression without time lag, both Hypothesis 1 and 3 are supported (positive and significant coefficients for both γ1 and γ2, suggesting the presence of intra-industry

spillover within a single industrial sector from foreign to local components and also spillover from other sectors to that sector. However, lagged FDI presence in one sector leads to a different story: evidence of positive intra-industry spillover becomes insignificant if we regress value added in one sector on FDI presence in the same sector one year, two years and three years before. This result suggests that the desired spillover within one sector exists only for a short period, specifically, in the same year of concern. The reason might be that negative effect of FDI does not take place until a certain period after entry, which in turn offset positive spillover. FIEs can do so by attracting skilled and experienced labour from their local competitors, or as suggested by Aitken and Harrison (1999), competition following entry of FDI reduces market share of domestic firms and cause them to cut production, in which way productivity of domestic firms would fall as they spread their fixed cost over a smaller market, forcing them back up their average cost curves at least in the shorn run. Figure 1 in next page illustrates this demand effect.

On the other hand, Hypothesis 3 receives constant support at 1% significance level along three year lagged terms. Besides the main finding, with δ2 only significant for one out of

FIGURE 1

Competition Effect on Productivity (source: Aitken and Harrison, 1999)

SOEs can have negative effect in productivity because they suffer from weak corporate governance and heavy social welfare burdens. The reason might be that though SOEs in China suffers from social welfare burden and inefficiency, they have always been in the forefront among domestic sectors (Liu 2002). The two forces in opposite direction thus might have offset each other. The coefficients for labour and capital are not always significantly positive along four years though they directly contribute to output.

After discussion on the results of Equation 4, we present regression of Equation 5 in Table 8 next page.

Coefficient of γ3 is not significant when we regress the dependent variable, total value

added by local firms in 34 sectors, on current and one-year lagged value of FDI presence in the remaining sector, suggesting no support for Hypothesis 2. But when we regress it on two-year and three-year lagged value, Hypothesis 2 is rejected. At the beginning two years, there is no sign that FDI presence in one sector has observable correlation with productivity of local firms as a whole from other sectors. And in an even longer time, there is even a significantly negative correlation. The possible explanation may be that it is easy for local units in one sector to feel the benefit of FDI presence in surrounding sectors (a significant and positive γ2 in Equation 4) while it is difficult for local units in

TABLE 8

Regression Results of Equation 5

No lag One-year lag Two-year lag Three-year lag FDI -0.003 (0.005) 0.002 (0.002) -0.008** (0.002) -0.007** (0.002) AFDI 0.139 (0.269) 1.081** (0.136) 0.227 (0.150) 0.021 (0.148) ASOE -0.037 (0.038) -0.011 (0.021) -0.001 (0.009) 0.007 (0.006) AL -0.442** (0.072) -0.024 (0.139) -0.181** (0.055) -0.143 (0.215) AK 1.185** (0.034) 1.215** (0.266) 1.752** (0.198) 1.957** (0.260) Adjusted R2 0.999 0.999 0.999 0.999 Sample size 378 294 252 210

Note: Fixed dummies are added for both industries and periods

GLS along time dimension is utilized.

participants of limited number, not to mention that we have abandoned some sectors due to lack of data. When we regress output of local firms in all other sectors on FDI presence in one sector, the average positive effect might become too small to capture. What is more, the entry of FDI brings products of new standard to local buyers and distributors

and different requirements for suppliers. In this way, they can create at least temporary disturbance to partners in front of and behind FIEs on the same value adding chain, thus lower productivity of surrounding industries. Hypothesis 4 does not receive strong support as γ4 turns to be significant and positive only in one period, indicating that that

there is no substantial positive FDI spillover effect within 34 sectors if we take them as a whole. This is a surprising result given that there is significant intra-industry spillover and inter-industry spillover in one direction, and γ4 attempts to capture productivity effect

of FDI presence in 34 sectors as a whole, which could be joint effect of intra- and inter-industry spillover. This cannot be a dubious result of collinearity since we find in previous section no serious correlation between lnAFDIjt and any other explanatory

In sum, FDI intra-industry spillover exists only in the initial year and disappears in the long run (H1). Coefficients for inter-industry effect in from 34 sectors to the remaining sector turns out to be positive and significant throughout four regressions with or without lag (H2), while spillover from FDI in one sector to local units as a whole in the others seems insignificant at the beginning and even becomes negative in the long run (H3), suggesting that positive spillover only flows from all sectors to one sector and some negative spillover from one sector to the others. However, since there are many channels for FDI spillover to take place it is possible that different channel direct spillover of different signs or magnitude, which produces above-mentioned contradicting results. And the two effects may offset each other (an insignificant γ4). Thus finally, there is no strong

support for general spillover effect of FDI within 34 industrial sectors taken as a whole (H4) probably as a result of the above two offsetting effects. We are not able to investigate in a longer time dimension, which requires dataset in an even longer time dimension and consumes degrees of freedom quickly with period and industry dummies.

CONCLUSION

Using data on 35 industrial sectors in China over the period from 1995 to 2003, we estimate the relationship between FDI presence and productivity of local industries. Besides intra-industry effect investigated by previous literatures on similar topic, we also investigate inter-industry spillover effects of two opposite direction, and a general effect of FDI presence in productivity of all local units in many sectors. Similar to many previous studies, there is no strong support for intra-industry spillover effect (an insignificant γ1). The unique findings of this paper are a constantly positive spillover

effect from FDI in all other sectors to local components in one sector (a significant and positive γ2) and, in contrast, a somewhat negative spillover effect in the opposite direction

though only significant in the last two periods (γ3). Perhaps as a result of these two

conflicting forces, there is also no clear evidence of spillover effect within many sectors (γ4). Hence, Chinese local government should encourage a certain basket of FDI tailored

sectors of main local industry are preferred to sectors that have little or no connection with local industry.

BIBLIOGRAPHY

Aitken, B. J. & Harrison, A. E. 1999. Do Domestic Firms Benefit from Direct Foreign Investment? Evidence from Venezuela American Economic Review, Vol. 89, pp. 605-618

Aitken, B., Harrison, A. and Lipsey, R. E. 1996. Wages and Foreign Ownership. Journal of International Economics, Vol. 40 Issue 3/4, p345-371, 17p, 11 charts (AN

9702146391)

Arellano, M., Imbens, G., Mizon, G. E., Pagan, A, & Watson, M. (Ed.) 2003. Panel Data Econometrics. Oxford, England: Oxford University Press

Blomström, M. 1991. Host Country Effects of Foreign Investment. Working paper No. 3615, Massachusetts, the US, National Bureau of Economic Research

Blomström, M. & Meller, P. (Ed.) 1991. Diverging Path: Comparing a Century of Scandinavian and Latin American Economic Development. Baltimore, Johns Hopkins

Blomström, M. & Persson, H. 1983. Foreign Investment and Spillover Efficiency in an Underdeveloped Economy: Evidence from the Mexican Manufacturing Industry. World Development, Vol. 11, pp. 493-501

Brash, D.T. 1966. American Investment in Australian Industry. Massachusetts, Harvard University Press

Chuang, Y. & Hsu, P. 2004. FDI, Trade, and Spillover Efficiency: Evidence from China’s Manufacturing Sector Applied Economics, 2004, 36, 1103-1115

Dunning, J. H. 2001. The Eclectic (OLI) Paradigm of International Production: Past, Present and Future. International Journal of the Economics of Business. Jul 2001, Vol. 8 Issue 2, p173-190

Görg, H. & Greenaway, D. 2004. Much Ado about Nothing? Do Domestic Firms Really Benefit from Foreign Direct Investment? World Bank Research Observer, Fall 2004, v. 19, iss. 2, pp. 171-97

Griffith, R. 1999. Productivity and Foreign Ownership in the UK Car Industry, Institute for Fiscal Studies, Working paper 99/11

Griffith, R. & Simpson, H. 2001. Characteristics of Foreign-owned Firms in British Manufacturing . Institute for Fiscal Studies, IFS Working Papers: W01/10

He, J. 2000. Further Quantification of FDI Spillover to China Manufacturing Industry. World Economy. Vol 12

Hill, R. C., Griffiths, W. E. & Judge, G. G. 2001. Undergraduate Econometrics. 2 edition, John Wiley & Sons, Inc

nd

Hu, A.G. Z. & Jefferson, G. H. 2002. FDI Impact and Spillover: Evidence from China's Electronic and Textile Industries World Economy, Aug 2002, Vol. 25 Issue 8, p1063-1076

Kokko, A. 1994. Technology, Market Characteristics, and Spillovers. Journal of Development Economics, Vol. 43, p. 279-293.

Liu, Z. 2002. Foreign Direct Investment and Technology Spillover: Evidence from China. Journal of Comparative Economics 30, 579-602

Liu, X., Siler, P., Wang, C. & Wei, Y. 2000. Productivity Spillovers From Foreign Direct Investment: Evidence From UK Industry Level Panel Data. Journal of International Business Studies, 2000 3rd Quarter, Vol. 31 Issue 3, p407, 19p

Love, J. H. 1997. The Transaction Cost Theory of the (Multinational) Firm: Note . Journal of Institutional and Theoretical Economics, December 1997, v. 153, iss. 4, pp. 674-81

Navaretti, G. B., Checchi, D. & Turrini, A. 2003. Adjusting Labor Demand: Multinational Versus Local firms: A Cross-European Analysis. Journal of the European Economic Association. Vol. 1 Issue 2/3, p708-719, 12p, 4 charts; DOI: 10.1162/154247603322391332; (AN 12231115)

Navaretti, G. B. & Venables, A. J. 2004. Multilocal firms in the World Economy, Chapter 7, 151-185. New Jersey, Princeton University Press (ISBN: 0-691-11920-1)

Southcn.com. 2004. FDI, a Threat to Security of National Economy? http://www.southcn.com/finance/hot/200411240473.htm. Accessed July 9, 2006

UNCTAD. 2005. World Investment Report 2005. http://www.unctad.org/wir or http://www.unctad.org/fdistatistics. Accessed June 10, 2006

Wei, Y & Liu, X. 2003. Mutual Productivity Spillovers between Foreign & Local Firms in China. Birmingham, Aston Business School Research Institute. ISBN: 1854495038