- Email: [email protected]
The relationship between inward direct foreign investment and China's provincial export trade
China Economic Review 12 (2001) 82 ± 99
The relationship between inward direct foreign investment and China's provincial export trade Qing ZHANG1, Bruce FELMINGHAM* School of Economics, University of Tasmania, GPO Box 252-85, Hobart, Tasmania 7001, Australia Received 3 March 2000; accepted 15 January 2001
Abstract This study evaluates the causal links between inward foreign direct investment (FDI) and exports from the PRC as a whole and also from its provinces. The national study is based on a monthly time series for the years 1986 to 1999 and cointegration/error correction modelling (ECM) techniques are used to conclude that for the PRC as a whole the relationship between FDI and exports is bidirectional. The relationship between inward FDI/exports for the PRC's provinces, cities, and autonomous regions is analysed in three panel data sets: the high FDI recipients (HFDI, n = 144) concentrated along the Chinese coast, medium FDI recipients (MFDI, n = 192) in Central China, and the low FDI group (LFDI, n = 128) in Western China. The panel data studies reveal that in the HFDI and LFDI, bidirectional causality applies, while exports Granger cause FDI in the MFDI. All regional results are confirmed by Sims tests. The policy implications are explained. D 2001 Elsevier Science Inc. All rights reserved. JEL classification: 01 Keywords: FDI; Exports; Cointegration; ECM; Causality
1. Introduction The objective of this analysis is to explain the causal relationship between China's inward foreign direct investment (FDI) and exports for China as a whole and for its provinces and * Corresponding author. Tel.: +61-3-6226-2308; fax: +61-3-6226-7587. E-mail address: [email protected] (B. Felmingham). 1 Tel.: + 61-3-6226-2821; fax: + 61-3-6226-7587. E-mail address: [email protected] (Q. Zhang). 1043-951X/01/$ ± see front matter D 2001 Elsevier Science Inc. All rights reserved. PII: S 1 0 4 3 - 9 5 1 X ( 0 1 ) 0 0 0 4 4 - X
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
83
cities. This analysis focuses on the precedence and timing of the relationship between inward FDI and Chinese exports. Stern (1997) summarises the arguments about the complementary nature of inward FDI and the trade of the recipient (host) country. FDI and exports from the host country may be complementary, particularly if the foreign interest is secured through the establishment of foreign invested enterprises (FIEs). Inward FDI brings with it the expertise of the foreign partner in selecting and promoting exports on international market in this way FDI enhances the recipient country's export performance. The nature of the relationship between inward FDI and exports from the host country is inextricably associated with questions about precedence and timing. The literature about this aspect of the inward FDI/exports relationship is incomplete in the case of the PRC. This study is designed to fill this gap. The following question is addressed here. Does inward FDI precede or lead exports, or vice versa? Although this questions evaluated empirically for the PRC and its provinces, an economic rationale is presented for the possibility that the relationship can occur in either of two single directions; alternatively, that there exists a bidirectional relationship between inward FDI and exports or there is no causal link at all. The importance of the causal relationship between inward FDI and exports is central to development planning and strategies. If there is a definite unidirectional causality from export expansion to FDI (EXP ! FDI), then some credence is given to an export led growth strategy. Exports not only stimulate economic development and structural change, they also attract FDI into China. If this is so, then development planners can expect those regions of their domestic economies with a pre existing export orientation to act as conduits for FDI. An appropriate policy in these circumstances is trade liberalisation designed to reduce exporters' cost and to encourage exports. The FDI that is encouraged by export expansion will stimulate the overall development of the domestic economy. If the causative process is in the opposite direction (FDI ! EXP), then the implication is that the inflow of FDI is a prerequisite for the expansion of China's exports. Then, the appropriate development strategy is to provide incentives for FDI, which in time leads to export growth. This seems to have been the motivation for the establishment of the first special economic zones (SEZs) by the PRC in 1979. Alternatively, export growth and inward FDI may have a reciprocal causative relationship (EXP$FDI). In other words, FDI and exports interact in the process of development. Finally, if there is no evident direction of causation between export growth and FDI, then alternative strategies to the encouragement of FDI or export promotion are required for structural transformation and growth of the Chinese economy. In summary, a rigorous test of causality between PRC's exports and its inward FDI is necessary as a way of informing the Chinese government and development agencies about the success of past policy and future development strategies for China. To this point in time, the policy emphasis has been focused on the eastern or seaboard provinces of China or provinces and cities close to the major sources of FDI, such as Hong Kong. The Chinese government has encouraged the growth of exports in the east by establishing SEZs and economic and technological development zones (ETDZs) and providing incentives for investment in them. This experience suggests that inward FDI should lead, precede, or cause export growth in the coastal provinces of China and the
84
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
incentives provided to foreign investors in these constitutes an appropriate policy mix for the eastern region. However, the central and western provinces of China have not received the same incentives for FDI. Some export growth has occurred in the central region in particular without the same level of incentives being offered for FDI while in western provinces generally, limited amounts of FDI have been received. This uneven distribution of FDI suggests that regional differences in the exports/FDI relationship should be taken into account. The remainder of this paper is structured in the following way, an economic rationale is developed in Section 2, which is followed by a description of the methodology in Section 3. The data set and empirical analyses are discussed in Section 4. Finally, conclusions and policy implications appear in Section 5. 2. Theory, policy, and empirics The debate about causality in the relationship between inward FDI and exports from a host country is one aspect of a more general argument in the development literature. This is summed up in the following question: do outward-oriented development policies produce more rapid growth in comparison with inward looking policies? This question is of particular relevance to the PRC, which has encouraged openness and outwardness since 1978. Hein (1992) develops both inward and outward policy models based on the experiences of the Latin American and East Asian regions. The implications of Hein's analysis is that those economies that promote exports have attracted large inward FDI suggesting that exports lead or precede FDI. The potential precedence of exports over FDI is one implication of Lucas' (1993) analysis of the development of the South East Asian region. Lucas finds that inward FDI is more sensitive (elastic) to the demand for exports than it is to domestic aggregate demand in his sample of South East Asian countries. The Hein and Lucas results suggest that exports from developing countries could lead or precede FDI if outward-oriented export promotion is successful. The case for the EXP ! FDI link is supported also by an argument about the efficiency of private markets in contrast to the argument by Hein, which relates only to the development policies of governments. Privately owned, export-oriented domestic firms are subject to the disciplines imposed by highly competitive international markets. Competition induces innovativeness and managerial efficiency that make export-oriented firms more attractive to foreign investors. This reinforces our previous point that an export orientation attracts FDI particularly in regions noted for their export performance, their openness, and their more developed infrastructure. This combination of factors is likely to attract a higher concentration of foreign investment. A causal link from host country exports to inward FDI can also be rationalised from the supply side. The competitive edge created by the pressures of international competition increase the productivity of the host country's export sector. According to Porter (1990, p. 10), productivity improvements will be manifest either in lower costs of capital utilisation or in differentiated products that may command higher international
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
85
prices. Lower costs mean a higher rate of return to foreign investors. So, the productivity gains generated by host country exports will attract inward FDI and exports will lead inward FDI. Causality from inward FDI to exports is explained in a more detailed literature, generally in the context of the complementarity or substitutability of exports and FDI. Some of this literature is relevant to this study. The essence of the argument presented by Muchielli and Chedor (1999) is summarised briefly. Foreign investors in developing countries may have a better sense about the mix of host country export products that can succeed in world markets. Foreign investment brings with it international market knowledge and global distribution networks not necessarily available to local firms. FDI can create greater export potential and may alter a country's export product mix over time. In these circumstances, inward FDI will lead exports. There are mutual advantages to the partners in a joint venture formed to tap the export capacity of host country firms. The domestic partner benefits from new technology, physical capital, and managerial expertise, while for the foreign partner, an existing local producer provides an opportunity to enter the domestic market, which in the case of the PRC is a very large one. The FIEs are often attracted by cheaper labour costs, which give them a competitive advantage in their export markets. The significance of the FIEs in the development of China's export capacity since 1978 is described in Section 4.2 of this paper. In addition to these arguments, for a single direction of causation either from exports to FDI or from FDI to exports, there is always the prospect of a bidirectional link. This arises for example, if an outward-oriented export led development policy attracts FDI on a large enough scale for this foreign investment to create trade subsequently. Then, exports may lead FDI initially before FDI creates further trade. The evidence about the causal links between FDI and exports is inconclusive. An important study of this issue is conducted by Jun and Singh (1996) who examine the determinants of FDI in 30 developing countries. Part of this study is focused directly on the causality issue: One caveat in the empirical literature is that it is not clear whether FDI flows are attracted by economies that are already export-oriented (i.e., exports precede FDI flows) or whether FDI leads to export increases (i.e., FDI precedes exports). (Jun & Singh, 1996, p. 73)
Jun and Singh (1996) explore this causality issue for 11 high FDI recipient countries over the period 1969±1993. In four cases, Thailand, Ecuador, Portugal, and Greece exports Granger cause FDI and in only one case, namely Singapore, there is evidence of FDI Granger-causing exports. In the six remaining countries (Colombia, Costa Rica, Egypt, Malaysia, Mexico, and Nigeria), FDI and exports are unrelated. The PRC was not included in this analysis. The issue concerning the causal links between inward FDI and exports from the PRC is significant for the PRC's development policies and business development strategies and our review of the literature suggests that four relationships are conceivable: FDI ! EXP; EXP ! FDI; the relationship is bidirectional or there is no relationship between the PRC's exports and inward FDI.
86
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
3. Methodology There are four methodologies applied to the data set described in Section 4.1. These are discussed in sequence. The first step involves an analysis of the descriptive statistics outlining the distribution of average annual FDI (US$ million) across the PRC's provinces listed in Table 1. Here too, we identify the groupings of provinces according to the criterion developed in Section 4.1. The role of the FIEs in exports is analysed from the results in Table 2. There are two interesting aspects: first, there is the increased contribution of FIEs to provincial exports since 1992 and second, there is the relative importance of FIEs on provincial exports across the three groups of provinces.
Table 1 Average FDI by provinces or cities (in US$ million, 1983 ± 1998) Number
Provinces/ cities
Average FDI (1983 ± 1998)
Group
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
687.68 648.50 336.10 56.48 29.25 719.27 146.01 191.73 1460.93 1959.41 522.57 151.51 1638.42 134.35 988.59 210.22 285.93 243.38 4617.47 326.49 388.29 214.52 21.24 59.52 156.43 23.98 4.26 8.51 21.25
HFDI HFDI MFDI LFDI LFDI HFDI MFDI MFDI HFDI HFDI HFDI MFDI HFDI MFDI HFDI MFDI MFDI MFDI HFDI MFDI MFDI MFDI LFDI LFDI MFDI LFDI LFDI LFDI LFDI
A city or province is classified as HFDI, MFDI, or LFDI as follows: HFDI: average FDI US$500 million; MFDI: US$100 million < average FDI US$499 million; LFDI: average FDI < US$100 million.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
87
Table 2 Exports by FIEs (PRC and provincial groups, in US$ million, 1992 ± 1998) Year Exports PRC Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports HFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports MFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports LFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%)
Percent change total exports (1992 ± 1998) Percent change FIEs' exports contribution (1992 ± 1998)
1992
1993
1994
1995
1996
1997
1998
84,876.73
91,694.38
121,019.60
148,766.85
151,013.33
182,673.88
183,243.67
17,356.19
25,238.56
34,711.74
46,875.57
61,504.01
74,896.86
80,891.25
20.45
27.52
28.68
31.51
40.73
41.00
44.14
67,251.76
74,178.25
101,278.28
125,147.80
129,202.86
157,752.25
160,993.49
16,780.80
24,117.26
33,189.87
44,974.52
58,700.84
71,609.89
77,670.28
24.95
32.51
32.77
35.94
45.43
45.39
48.24
14,477.53
14,468.94
16,028.21
18,583.08
17,286.66
19,791.73
17,184.99
502.87
965.72
1297.61
1603.07
2418.86
2859.40
2821.37
3.47
6.67
8.09
8.62
13.99
14.45
16.42
3147.44
3047.19
3713.11
5035.97
4523.81
5129.90
5065.19
72.52
155.58
224.26
297.98
384.31
427.57
399.60
2.30
5.11
6.04
5.92
8.50
8.33
7.89
PRC
HFDI
115.89
139.39
MFDI 18.70
LFDI 60.93
366.07
362.85
461.05
451.02
Source: The data of total exports and exports by FIEs for 22 provinces, 4 autonomous regions, and 3 municipal cities were drawn from the Chinese Statistical Yearbook from 1995 to 1999.
This analysis of the relevant descriptive data is followed by a time series study of the relationship between inward FDI and exports from the PRC in the aggregate. This time series analysis will be based first on augmented Dickey±Fuller (ADF) and Phillips±Perron
88
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
(PP) tests for the presence of unit roots in the data on export and FDI levels and their first differences. Then, we proceed to test for a bivariate cointegration of aggregate monthly exports and FDI series. This bivariate cointegration model will establish if there is a longrun equilibrium relationship between exports and FDI. Finally, a standard error correction model (ECM) is formulated in both directions to see if there is a short-run adjustment towards long-run equilibrium in one or both directions. The ADF, PP tests for stationarity and Engle and Granger (1987) test for bivariate cointegration are well known and will not be repeated here; however, the causality test associated with this process will involve two ECMs as follows: FDIt a1
k X i1
EXPt a2
b1i FDIt
k X i1
b2i EXPt
k
k X i1
k
g1i EXPt
k X i1
g2i FDIt
k
k
r1 ECTt
r2 ECTt
k
k
e1t
1
e2t :
2
In Eqs. (1) and (2), ECT stands for the error correction terms. These are residuals generated from the cointegration of exports to FDI in Eq. (1) and from FDI to exports in Eq. (2). There are two bases for the presence of causality in these ECMs. The first concerns the short-run response of one variable, for example, exports to FDI and the second relates to the adjustment of the variables back towards long-run equilibrium. Standard Granger causality tests are misspecified if this second channel is not recognised. If in the estimation of Eqs. (1) and (2) both ^r1 and ^r2 are significant then causality is deemed to be bidirectional. However, if only one of ^r1 and ^r2 is significant, then a single direction of causation applies, and if neither is significant, then this second channel of causation is closed. The first channel may still apply even if the second is closed. In that case, it is appropriate to apply standard Granger causality modes to determine if a short-run interaction between inward FDI and exports exists. Ibrahim (2000) provides an interesting example of this situation. His bivariate cointegration tests fail to indicate a long-run relationship between Malaysian stock prices and the exchange rate (RM/US); however, he does find evidence of a short-run reaction between these variables when standard Granger causality tests are applied. A third aspect of our methodology is to determine if the findings of the national study are uniform across the provinces of the PRC. However, we acknowledge in Section 4.1 that there are too few time series observations to support an individual time series study for each province. In its place, three panel data sets are formed, one panel for each of the three provincial groups: high FDI (HFDI), medium FDI (MFDI), and low FDI (LFDI). From Table 1, it is evident that HFDI contains 9 provinces, while 12 and 8 provinces constitute the MFDI and LFDI panels, respectively. The 16 annual observations on exports and FDI when combined with the number of provinces in each group constitute the following data sets: HFDI 144 (9 16), MFDI 192(12 16), and LFDI 128 (8 16) observations.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
89
Standard Granger causality tests2 are applied to each of these panels. However, Granger tests are limited by their one sided nature. Granger tests relate to lags alone between two variables3. To overcome this limitation, we apply Sims (1972) test, which takes into account both lags and leads. The following interpretation (Eqs. 3±6) of the Sims' framework applies in this case: FDI f1 EXP; n past and m future lags of EXP;
3
FDI g1 EXP; n past lags of EXP;
4
EXP f2 FDI; n past and m future lags of FDI;
5
EXP g2 FDI; n past lags of FDI:
6
To test for unidirectional causality from FDI to exports, exports are regressed on past, current, and leading values of FDI. Causality from FDI to exports does hold if the coefficients of all leading (future) values of FDI are insignificant as a group. Likewise, exports lead FDI when all future values of exports are found to be insignificant in regression equations (3) and (4). 4. Data and results This section begins with a discussion of the data set that underpins the study and proceeds to an analysis of the results. The outcomes are discussed in the following sequence: first, there is a descriptive analysis of the data; second, the outcomes of the aggregate study for the PRC as a whole are discussed. Finally, the results of the provincial study based on grouped panel data are presented.
2
a FDIt
m X
b EXPt
li FDIt
i
i1 n X
dj EXPt
j1
ai FDIt
i
i1
FDI ! EXP : if
EXP ! FDI : if
n X j1
n X i1
bj EXPt
j
m2t :
j
m1t :
a^ i 6 0 in Eq: b and
n X i1
3
m X
a^ i 0 in Eq: b and
m X ^ dj 0 in Eq: a j1 m X ^ dj 6 0 in Eq: a: j1
We have made some common assumptions in estimating the Granger equations on panel data, in particular, that the disturbance terms in Eqs. (a) and (b) from footnote 2 have constant variances with finite moments and have independent increments.
90
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
4.1. The data set It was essential ab initio to determine if this study should focus on a national data set for the PRC as a whole, or if it should be pitched at a regional level and consider the link between FDI and exports in the PRC's provinces, cities, and autonomous regions (hereafter ``provinces''). Some authors, most recently, Sun and Chai (1998) and Tang (1998) argue that the provinces of the PRC are the appropriate basis because the culture, history, and structure of the individual provinces are so diverse. This argument turns out to be supported by the descriptive data included on Table 1 showing the distribution of inward FDI. Clearly, the range of variation in the average values of FDI received by individual provinces is large and so it is important to distinguish those provinces that receive large amounts of inward FDI from those that do not. However, it can be argued that the link between FDI and exports should be analysed for the PRC as a whole. An aggregate study of causality will inform policy makers as readily as separate provincial studies. For example, if a study based on China as a whole reveals a particular causal direction while an individual province or group of provinces displays a different outcome, then an explanation of this difference may be informative for policy. Our preference for separate national and provincial studies is also dictated by the availability of data. A monthly national time series for exports from the PRC and inward FDI was obtained from the China Monthly Economic Indicators for the years 1986 to 1999. This provides a time series of 168 observations, sufficient for the application of a more recent approach to causality tests based on ECM. However, reliable provincial data on exports and FDI is only available on an annual basis from various publications of the Chinese Statistical Yearbook (State Statistical Bureau of China [SSB], 1980±1998, various years) for the years 1983 to 1998 providing a data set of 16 observation that is not sufficient for individual provincial time series analysis. We address this problem by grouping the data in Table 1. Provinces with average annual inward FDI valued at US$500 million or more for the period 1983 to 1998 are referred to as HFDI recipients; those provinces receiving an average annual FDI in the range of US$100 million to US$499 million are classified as MFDI recipients, while provinces receiving less than US$100 million per annum on average over the reference period are labelled LFDI provinces. This approach to grouping the Chinese provinces is preferred to an alternative geographical method because FDI, as the basis of the classification, is also a key variable in this study. The alternative is a geographical basis that identifies separate coastal or eastern, middle, and western provinces of the PRC. However, there is a close correspondence between these geographical and functional modes of classification. The HFDI provinces are largely concentrated along the PRC's coast, while MFDI provinces are generally located in Central China and the LFDI provinces are to be found in Western China. 4.2. Analysis of descriptive data Table 1 contains the dollar value of inward FDI averaged over the years 1983±1998 for each of 29 provinces. Note that Chongqing is excluded from this list because it was
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
91
established only in 1997. Further, Tibet's receipt of FDI is negligible and it is excluded from further study. The following provinces receive FDI on average in excess of US$500 million per annum: Guangdong, Jiangsu, Fujian, Shanghai, Shandong, Liaoning, Beijing, Tianjin, and Zhejiang. These provinces comprise the HFDI group and are located along the Chinese coast or in close proximity to it. The role of FIEs in the expansion of the PRC's exports is an issue warranting some preliminary analysis. Indigenous Chinese firms and the FIEs generally export different types of goods. The FIEs' product range benefits from advanced technology, managerial expertise, and superior international networks. The foreign links established in the FIEs are largely concentrated in the HFDI group. This lends some plausibility to the argument that FDI leads exports or maybe that causality is bidirectional in the HFDI. However, among the LFDI and MFDI provinces concentrated in Western and Central China, respectively, it may be the case that exports are dominated by indigenous firms and that export growth is independent of FDI. Table 2 contains the proportion of the FIEs exports from the overall PRC and each of the three provincial groups since 1992. The concentration of the PRC's exports in the HFDI group is clear on Table 2. Of the export values measured in US$ million, 80% and 87% are accounted for by the HFDI provinces in 1992 and 1998, respectively. Further, the value of exports from the HFDI group have grown at a faster rate (139.39%) than export growth in either the MFDI (18.7%) and LFDI (60.93%). This result is not surprising given the concentration of FDI in the coastal provinces of the PRC and the effect of this on export development. This outcome constitutes prima facie evidence for causality flowing from FDI to exports among the HFDI group. A formal analysis will firm up this casual observation. A second important observation from Table 2 concerns the fall in total export values in 1998. Export growth slowed down in this year across the PRC and its provinces but remained positive for the nation and the HFDI group, but fell in the MFDI and LFDI groups. The behaviour of export growth in the MFDI group is of particular interest. Total exports in this group fell by 13%, but exports from the FIEs fell by less than 2%. This outcome may be explained by a greater degree of vulnerability of the indigenous exporters in the MFDI to a crisis such as that engendered by the Asian currency collapse in 1997. The remaining noteworthy feature of Table 2 is the increase in the proportion of total exports accounted for by the FIEs. This proportion has risen by 451% and 461%, respectively in the LFDI and MFDI, faster than the growth in the proportional contribution of FIEs exports to the total value of exports from the HFDI group. This outcome is suggestive of a particular pattern to the link between China's exports and its inward FDI. Initially, it is the export potential of a region that attracts FDI, in this case invested in FIEs. However, the effects of FDI on exports are subject to a gestation period in which time export potential is turned into export performance. 4.3. National (PRC) study of causality The test for causality in the relationship between inward FDI and PRC exports is based on a monthly time series dating from 1986 to 1999. We begin by testing each series for
92
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
stationarity applying standard ADF and PP procedures: the F and t statistics produced by these tests appear in columns 1 to 8 of Table 3. The results of these tests are inconclusive. The ADF F and t statistics suggests that FDI is nonstationary in levels while the corresponding PP statistics indicate that the FDI series is stationary. Similarly, the ADF t statistic for exports suggests nonstationarity, while the ADF F and PP F and t statistics indicate stationarity. In these circumstances, we are unable to conclude that the monthly series for PRC exports and inward FDI are stationary in levels, however the first difference of each series is clearly stationary: ADF and PP statistics exceed the relevant cut off scores in absolute value in each series. Taking the results of these stationarity tests collectively, it is appropriate to assume that both monthly time series are integrated of order one (I(1)) and to proceed to tests for cointegration. The results of Engle and Granger (1987) bivariate cointegration tests are disclosed in columns 9 and 10 of Table 3. The ADF test results indicate that a long-run equilibrium between inward FDI and exports does not exist. However, the PP test supports the view that there is cointegration of these time series. Greater reliance is placed on the PP results, because PP tests for cointegration are nonparametric and do not rely upon any distributional assumptions and are not characterised by the moments of a probability distribution function. This makes PP tests more robust to misspecification and data inconsistencies. The presence or otherwise of a long-run equilibrium relationship between inward FDI and exports for the PRC will be confirmed in the following ECM. The test results of ECM for the national study based on Eqs. (1) and (2) are reported in Table 4. The generic statistical characteristics of these estimates are sound. The RÅ2 for each equation is of acceptable magnitude. The Breusch±Godfrey (BG) test statistics for Eqs. (1) and (2) are 3.95 and 8.82, respectively, which indicate serial correlation is not present. It is clear that the coefficient of ECT in both equations is significant. These results confirm the presence of cointegration of the two variables: PRC's exports and its inward FDI. Statistically, the presence of cointegration rules out noncausality between the variables. As has been noted by Granger (1988), if a pair of I(1) series are cointegrated, there must be causation in at least one direction. The significance of ^r1 and ^r2 at 5% level in the estimates of Eqs. (1) and (2), respectively, suggests also that causality is bidirectional for the PRC as a whole. 4.4. Causality among PRC groups and pooled data The provinces of the PRC stand at different levels of development. This is recognised in our functional groupings of PRC provinces. As stated previously, these groups correspond broadly to the geographical regions of China: the eastern (coastal) region, which contains the HFDI group; the MFDI located in middle China; and the western provinces, which include the LFDI provinces. It is conceivable for the direction of causation between exports and FDI to differ from the national bidirectional pattern. The methodology applied for this purpose is detailed in Section 3 of the paper and involves causality tests based on Granger and Sims models. Each group of data is treated as a separate panel and Eqs. (a) and (b) contained in footnote 2 are estimated for the three panels.
0.93 1.72 3.13
* Significant at the 10% level.
2.11 9.21* 4.03
FDI EXP Critical value (10%)
30.73* 39.54* 4.03
F
4 t 9.60* 10.86* 3.13
4.50* 10.17* 4.03
F
5 t
6 3.41* 5.50* 3.13
3
t
2
1
F
ADF
PP
ADF
Variables
First difference
Levels
Stationarity
Table 3 Unit root test and cointegration test results: FDI and exports
155.11* 130.38* 4.03
F
7
PP 8 t 21.58* 19.79* 3.13
FDI EXP EXP FDI Critical value (10%)
Variables
9 0.97 1.48 3.50
ADF
10
PP
Cointegration
118.42* 143.80* 23.40
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99 93
94
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
Table 4 Causality tests based on bidirectional ECM Eq. (1) (dependent variable: FDIt) Coefficient b11 b12 b13 b14 b15 b16 g11 g12 g13 g14 g15 g16 r1 a1 R2 adj. BG df = 148
Estimate 0.908 0.808 0.838 0.787 0.604 0.10 0.023 0.012 0.042 0.046 0.029 0.024 0.434 1.042 0.617 3.951
Eq. (2) (dependent variable: EXPt) t ratio 9.814* 7.227* 6.843* 6.308* 5.137* 1.039 0.745 0.309 0.921 1.025 0.729 0.766 3.087* 1.882**
Coefficient g21 g22 g23 g24 g25 g26 b21 b22 b23 b24 b25 b26 r2 a2 R2 adj. BG df = 148
Estimate 0.829 0.858 0.571 0.271 0.285 0.094 0.053 0.111 0.473 0.975 0.467 0.159 0.335 4.21 0.458 8.82
t ratio 8.709* 6.579* 3.791* 1.802** 1.893** 0.931 0.191 0.326 1.286 2.592* 1.314 0.562 2.007* 2.541*
The number of lags applicable in this estimation was determined by applying the Akaike's FPE test to estimates of Eqs. (1) and (2). Six lags were chosen because this lag length minimised the FPE with highest adjusted R2. BG test results above indicate six lag model is free of serial correlation since the critical value of the c2 distribution is 9.48 at the 5% level for BG statistics. * Significant at the 5% level: critical value 1.96 (df = 148). ** Significant at the 10% level: critical value 1.65 (df = 148).
Provincial data are also pooled to get the overall picture for the PRC as a whole. The pooled series contains 464 observations. The application of Akaike's Information Criterion and Akaike's (1969, 1970) Final Prediction Error (FPE) criterion to both Sims and Granger causality models for each group panel and for pooled data suggests that the lag structures shown on Table 5 are appropriate.4 Four, two, and three lags were preferred for Granger tests of causality in the HFDI, MFDI, and LFDI groups, while three lags are required for pooled data. One backward and three forward lags are required for Sims test based on the three grouped panels. The preferred lag structure for the Sims test on pooled data is three backward and three forward lags. The grouped Granger and Sims causality test results for each of these groups in both directions (FDI ! EXP and EXP ! FDI) are listed on Table 6. The causal direction inferred for each group is summarised in the last column. The results in Table 6 indicate that the direction of causality between inward FDI and exports is bidirectional for both HFDI and LFDI groups based on the Granger and Sims tests. The HFDI group is composed of some
4
The appropriate lag length are chosen with minimised Akaike FPE and higher adjusted R2 as criteria.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
95
Table 5 Preferred lag structures HFDI (n = 144)
MFDI (n = 192)
LFDI (n = 128)
Pooled (N = 464)
Granger causality FDI ! EXP EXP ! FDI
4 4
2 2
3 3
3 3
Sims' model FDI ! EXP EXP ! FDI
(1,3) (1,3)
(1,3) (1,3)
(1,3) (1,3)
(3,3) (3,3)
The above lags were chosen after evaluating: two, three, or four lags for Granger causality and (2 back, 1 forward), (3,1), (1,2), (2,2), (3,2), (3,2), (1,3), (2,3), (3,3) in the case of Sims model. Four, two, and three lags were preferred for Granger tests of causality in the HFDI.
Table 6 The grouped Granger and Sim's causality for PRC (1983 ± 1998) FDI on export
Export on FDI
F ratio
Result
BG
F ratio
Result
BG
1
2
3
4
5
6
Causal inference
reject H0
1.90 (df = 4)
13.56* (4,132)
reject H0
0.98 (df = 4)
FDI ! EXP EXP ! FDI
1.96 (df = 2) 0.818 (df = 3)
7.85* (2,186) 24.80* (3,119)
reject H0
3.44 (df = 2) 2.48 (df = 3)
EXP ! FDI
Granger's test HFDI 3.28* group (4,132) MFDI group LFDI group
0.79 (2,186) 6.85* (3,119)
not reject H0
Pooled
14.02* (3,455)
reject H0
2.79 (df = 3)
53.73 (3,455)
reject H0
2.79 (df = 3)
FDI ! EXP EXP ! FDI
Sim's test HFDI group
46.62* (3,134)
reject H0
85.34 (df = 3)
10.35* (4,108)
reject H0
82.64 (df = 3)
FDI ! EXP EXP ! FDI
MFDI group LFDI group
2.99 (3,182) 41.70* (3,118)
not reject H0
115.74 (df = 3) 115.19 (df = 3)
2.93* (4,183) 16.07* (4,108)
reject H0
85.41 (df = 3) 8.57 (df = 3)
EXP ! FDI
Pooled
27.82* (3,450)
reject H0
258.43* (df = 3)
4.63* (3,450)
reject H0
285.58* (df = 3)
FDI ! EXP EXP ! FDI
reject H0
reject H0
The degrees of freedom are in the parentheses after the estimates. F test: 2,42, 3.04, 2.68, 2.68, 2.68, 2.68, and 2.60. BG: 9.487 (df = 4), 7.815 (df = 3), and 5.99 (df = 2). * The critical value at 5% level.
reject H0
reject H0
FDI ! EXP EXP ! FDI
FDI ! EXP EXP ! FDI
96
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
eastern coastal regions that are relatively advanced and includes the economic growth centres, such as Guangdong, Shanghai, and Beijing. They are major seaports and airports with superior transportation, communications, better service facilities, and human resources. For this reason, this group of provinces offers the most attractive investment environment in China. The HFDI group became major export providers and principal foreign investment recipients following the opening of China's doors in 1978. The HFDI group of provinces are relatively open to international trade and have well-established trade channels. Their export capacities have been strengthened with the help of foreign funds, especially in the SEZs. The growing exports of the HFDI group, in turn, not only stimulate the development of nonexport sectors, but also encourage regional economic growth as a whole, attracting more foreign investors by providing a more promising investment environment. The presence of an export trade encourages investment on the one hand while further injections of FDI stimulate export growth, so the link between FDI and exports is bidirectional. Most provinces in the LFDI group are backward inland regions with a relatively low degree of openness. The inland provinces were not allowed to trade with the rest of the world prior to the open-door policy. Their business activities were domestically oriented. Once an internationally oriented development policy was applied to China, export and foreign investment inflow became the two major determinants of economic growth among the LFDI. Some foreign investors seem to have taken advantage of the western region's mineral and energy resource abundance regardless of its poor infrastructure conditions. The results on Table 6 confirm the reciprocal relationship among the LFDI group. FDI is a prerequisite for export and economic development as a whole. On the other hand, exports facilitate the inflow of foreign funds. But the scale of FDI and export growth is relatively small in comparison with the HFDI group. The test results for the MFDI group suggests that the direction of causation runs from exports to FDI. The economies of most MFDI members are not as well developed as the HFDI group although some of them are located close to the eastern coast and had a successful export record prior to 1978. Exports, no doubt, played an important role in the economic development of the MFDI since most of the MFDI members did not receive government approval for foreign investment until at least 6 years after the first SEZs were established. Unidirectional causality can be explained by the gradualism with which China's doors were opened. The SEZs were established well ahead of the MFDI. The time lag in implementing the open-door policy across China did make a difference to regional development. This is reflected in the results on Table 6. Also, regionally different economic influences are evident in the decentralisation of the Chinese economy as a whole during the transition from a planned to a market economy. Results are also presented on Table 6 for the pooled data set comprising 29 provinces over the 16-year period. This is another way of analysing causality from the perspective of the PRC as a whole because it contains the cross section of provinces with a smaller set of time series observations. It is clear that these pooled data outcomes provide the same results as the national time series and ECM discussed in Section 4.2. In both Granger and Sims tests based on pooled data, the value of the F statistic leads to the rejection of the null hypothesis suggesting the absence of a causal flow from FDI to exports and then again from exports to FDI. So the conventional Granger and Sims tests produce the same outcome as the ECM
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
97
study. The pooled data analysis confirms that there is a bivariate causal relationship between inward FDI and exports for the PRC as a whole. The values of the BG test statistic in columns 3 and 6 of Table 6 indicate that these OLS estimates of the Granger but not Sims equations are free of serial correlation in general.5 We keep this in mind when analysing and interpreting the results of the Sims test. We take comfort from the fact that both Sims and Granger tests lead to the same conclusions. 5. Conclusions The major finding of this study is that bidirectional causality applies to the link between inward FDI and exports from the PRC. This is the clear outcome of the national time series analysis of the issue and is reinforced in a pooled regression model across 29 PRC provinces. We also find bidirectional causality between inward FDI and exports from the HFDI recipient provinces concentrated along the PRC coast. More than 80% of the PRC's total export values originate from these coastal provinces, so it is not surprising that the outcome for the HFDI panel mirrors the national pattern. This major outcome of the study provides some testimony to the effectiveness of the policy of opening the PRC's doors leading to greater interaction between the PRC and the rest of the world. In particular, FDI was encouraged from 1978 and beyond while the emphasis of the PRC's development policy was refocused to allow a greater emphasis to be given to export promotion. The FIEs contribution to the PRC's development is also emphasised in our statistics where the proportion of export values contributed by the FIEs doubled between 1992 and 1998. If it was the intention of the PRC government to create trade through foreign involvement in 1978, then the policy intention has been realised at least in the coastal provinces. However, the linkage between inward FDI and exports in the central provinces of the PRC (MFDI) differs from the national and HFDI study. In the case of the MFDI, causality appears to flow from exports or export potential to FDI. This provides some support for those researchers such as Dollar (1992) and Jun and Singh (1996) who find that outward-oriented economies particularly those exploiting their export potential attract more foreign investment and grow faster. It is their outward orientation and policies such as export enhancement that attract FDI and create trade subsequently. It is interesting to note that our results for the MFDI correspond with Jun and Singh's analysis, which finds that export potential leads FDI frequently. This finding for the MFDI may be indicative of a process of staged development. In a first stage, the degree of openness attracts FDI, which in time creates trade in the form of increased exports. If causality is assessed at an early stage of this process, then exports or potential exports could appear to be leading FDI, however, the direction could be reversed at a later stage when the FDI has turned export potential into real export growth. Our 5
Values of the BG test are below the relevant critical c2 value at a=.05 in testing that in the error term regressions: et r1 et 1 r2 et 2 r3 et 3 r4 et 4 m t all coefficients are simultaneously zero.
98
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
conclusion about the MFDI provinces may simply reflect an early stage of the export development process. The role of the FIEs in insulating the PRC's exports against the effects of economic downturns is evident in the case of the MFDI. The US value of total exports from the MFDI provinces fell by 13.1% between 1997 and 1998 as a consequence of the Asian currency crisis while FIE exports from the MFDI provinces fell by only 1.33%. The results for the LFDI provinces show some evidence of FDI playing a leading role in the development of exports from Western China. However, the development of the full potential of this region will depend on foreign capital being invested in infrastructure projects before the full export and domestic growth potential of this region of China is realised. Our finding of causality flowing from FDI to exports in the LFDI must be seen in this context. Acknowledgments The authors wish to thank the editor and anonymous referees for their advice, which has improved the manuscript substantially. The authors are responsible for any remaining errors and omissions. References Akaike, H. (1969). Fitting autogression for prediction. Annals of the Institute of Statistical Mathematics, 21, 203 ± 217. Akaike, H. (1970). Statistical predictor identification. Annals of the Institute of Statistical Mathematics, 22, 243 ± 247. Dollar, D. (1992). Outward-oriented developing economies really do growth more rapidly: evidence from 95 LDCs 1976 ± 1985. Economic Development and Cultural Change, 40 (3), 523 ± 544. Engle, R.F., & Granger, C.W.J. (1987). Cointegration and error correction: representation estimation and testing. Econometrica, 2(55), 251 ± 276. (March). Granger, C.W.J. (1988). Some recent developments in a concept of causality. Journal of Econometrics, 39, 199 ± 211. Hein, S. (1992). Trade strategy and the dependency hypothesis: a comparison of policy, foreign investment and economic growth in Latin America and East Asia. Economic Development and Cultural Change, 40 (3), 495 ± 521. Ibrahim, M.H. (2000). Cointegration and Granger causality tests of stock price and exchange rate interactions in Malaysia. ASEAN Economic Bulletin, 17 (1), 36 ± 47. Jun, K.W., & Singh, H. (1996). The determinants of foreign direct investment in developing countries. Transnational Corporations, 2(5), 67 ± 105 (August). Lucas, R. (1993). On the determinants of direct foreign investment: evidence from East and Southeast Asia. World Development, 21 (3), 391 ± 406. Mucchielli, J.L., & Chedor, S. (1999). Foreign direct investment, export performance and the impact on home employment: an empirical analysis of French firms. In: S.G. Lee, & P.-B. Ruffini (Eds.), New horizons in international business. Cheltenham, UK: Edward Elgar. Porter, M.E. (1990). The competitive advantage of nations. London: Macmillan. Sims, C.A. (1972). Money, income and causality. American Economic Review, LXII (4), 540 ± 552.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
99
State Statistical Bureau of China. (1980 ± 1998). Statistical Yearbook of China (SYOC) 1980 ± 1998. Beijing: China Statistical Publishing House. State Statistical Bureau of China (various years). Statistical Yearbook of China. Beijing: China Statistical Publishing House. Stern, R.E. (1997). Foreign direct investment, exports and eastwest integration: theory and practice. (Chapter 10, pp. 329 ± 357) In: R.N. Cooper, & J. Cacs (Eds.), Trade growth in transition economies export impediments for Central and Eastern Europe. Cheltenham, UK: Edward Elgar. Sun, H., & Chai, J. (1998). Direct foreign investment and inter-regional economic disparities in China. International Journal of Social Economics, 25 (2 ± 4), 424 ± 427. Tang, K. K. (1998). Economic integration of the Chinas provinces: a business cycle approach National Centre for Development Studies, ANU, RSPS. (Working Paper No. 98/1, pp. 1 ± 16).
The relationship between inward direct foreign investment and China's provincial export trade Qing ZHANG1, Bruce FELMINGHAM* School of Economics, University of Tasmania, GPO Box 252-85, Hobart, Tasmania 7001, Australia Received 3 March 2000; accepted 15 January 2001
Abstract This study evaluates the causal links between inward foreign direct investment (FDI) and exports from the PRC as a whole and also from its provinces. The national study is based on a monthly time series for the years 1986 to 1999 and cointegration/error correction modelling (ECM) techniques are used to conclude that for the PRC as a whole the relationship between FDI and exports is bidirectional. The relationship between inward FDI/exports for the PRC's provinces, cities, and autonomous regions is analysed in three panel data sets: the high FDI recipients (HFDI, n = 144) concentrated along the Chinese coast, medium FDI recipients (MFDI, n = 192) in Central China, and the low FDI group (LFDI, n = 128) in Western China. The panel data studies reveal that in the HFDI and LFDI, bidirectional causality applies, while exports Granger cause FDI in the MFDI. All regional results are confirmed by Sims tests. The policy implications are explained. D 2001 Elsevier Science Inc. All rights reserved. JEL classification: 01 Keywords: FDI; Exports; Cointegration; ECM; Causality
1. Introduction The objective of this analysis is to explain the causal relationship between China's inward foreign direct investment (FDI) and exports for China as a whole and for its provinces and * Corresponding author. Tel.: +61-3-6226-2308; fax: +61-3-6226-7587. E-mail address: [email protected] (B. Felmingham). 1 Tel.: + 61-3-6226-2821; fax: + 61-3-6226-7587. E-mail address: [email protected] (Q. Zhang). 1043-951X/01/$ ± see front matter D 2001 Elsevier Science Inc. All rights reserved. PII: S 1 0 4 3 - 9 5 1 X ( 0 1 ) 0 0 0 4 4 - X
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
83
cities. This analysis focuses on the precedence and timing of the relationship between inward FDI and Chinese exports. Stern (1997) summarises the arguments about the complementary nature of inward FDI and the trade of the recipient (host) country. FDI and exports from the host country may be complementary, particularly if the foreign interest is secured through the establishment of foreign invested enterprises (FIEs). Inward FDI brings with it the expertise of the foreign partner in selecting and promoting exports on international market in this way FDI enhances the recipient country's export performance. The nature of the relationship between inward FDI and exports from the host country is inextricably associated with questions about precedence and timing. The literature about this aspect of the inward FDI/exports relationship is incomplete in the case of the PRC. This study is designed to fill this gap. The following question is addressed here. Does inward FDI precede or lead exports, or vice versa? Although this questions evaluated empirically for the PRC and its provinces, an economic rationale is presented for the possibility that the relationship can occur in either of two single directions; alternatively, that there exists a bidirectional relationship between inward FDI and exports or there is no causal link at all. The importance of the causal relationship between inward FDI and exports is central to development planning and strategies. If there is a definite unidirectional causality from export expansion to FDI (EXP ! FDI), then some credence is given to an export led growth strategy. Exports not only stimulate economic development and structural change, they also attract FDI into China. If this is so, then development planners can expect those regions of their domestic economies with a pre existing export orientation to act as conduits for FDI. An appropriate policy in these circumstances is trade liberalisation designed to reduce exporters' cost and to encourage exports. The FDI that is encouraged by export expansion will stimulate the overall development of the domestic economy. If the causative process is in the opposite direction (FDI ! EXP), then the implication is that the inflow of FDI is a prerequisite for the expansion of China's exports. Then, the appropriate development strategy is to provide incentives for FDI, which in time leads to export growth. This seems to have been the motivation for the establishment of the first special economic zones (SEZs) by the PRC in 1979. Alternatively, export growth and inward FDI may have a reciprocal causative relationship (EXP$FDI). In other words, FDI and exports interact in the process of development. Finally, if there is no evident direction of causation between export growth and FDI, then alternative strategies to the encouragement of FDI or export promotion are required for structural transformation and growth of the Chinese economy. In summary, a rigorous test of causality between PRC's exports and its inward FDI is necessary as a way of informing the Chinese government and development agencies about the success of past policy and future development strategies for China. To this point in time, the policy emphasis has been focused on the eastern or seaboard provinces of China or provinces and cities close to the major sources of FDI, such as Hong Kong. The Chinese government has encouraged the growth of exports in the east by establishing SEZs and economic and technological development zones (ETDZs) and providing incentives for investment in them. This experience suggests that inward FDI should lead, precede, or cause export growth in the coastal provinces of China and the
84
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
incentives provided to foreign investors in these constitutes an appropriate policy mix for the eastern region. However, the central and western provinces of China have not received the same incentives for FDI. Some export growth has occurred in the central region in particular without the same level of incentives being offered for FDI while in western provinces generally, limited amounts of FDI have been received. This uneven distribution of FDI suggests that regional differences in the exports/FDI relationship should be taken into account. The remainder of this paper is structured in the following way, an economic rationale is developed in Section 2, which is followed by a description of the methodology in Section 3. The data set and empirical analyses are discussed in Section 4. Finally, conclusions and policy implications appear in Section 5. 2. Theory, policy, and empirics The debate about causality in the relationship between inward FDI and exports from a host country is one aspect of a more general argument in the development literature. This is summed up in the following question: do outward-oriented development policies produce more rapid growth in comparison with inward looking policies? This question is of particular relevance to the PRC, which has encouraged openness and outwardness since 1978. Hein (1992) develops both inward and outward policy models based on the experiences of the Latin American and East Asian regions. The implications of Hein's analysis is that those economies that promote exports have attracted large inward FDI suggesting that exports lead or precede FDI. The potential precedence of exports over FDI is one implication of Lucas' (1993) analysis of the development of the South East Asian region. Lucas finds that inward FDI is more sensitive (elastic) to the demand for exports than it is to domestic aggregate demand in his sample of South East Asian countries. The Hein and Lucas results suggest that exports from developing countries could lead or precede FDI if outward-oriented export promotion is successful. The case for the EXP ! FDI link is supported also by an argument about the efficiency of private markets in contrast to the argument by Hein, which relates only to the development policies of governments. Privately owned, export-oriented domestic firms are subject to the disciplines imposed by highly competitive international markets. Competition induces innovativeness and managerial efficiency that make export-oriented firms more attractive to foreign investors. This reinforces our previous point that an export orientation attracts FDI particularly in regions noted for their export performance, their openness, and their more developed infrastructure. This combination of factors is likely to attract a higher concentration of foreign investment. A causal link from host country exports to inward FDI can also be rationalised from the supply side. The competitive edge created by the pressures of international competition increase the productivity of the host country's export sector. According to Porter (1990, p. 10), productivity improvements will be manifest either in lower costs of capital utilisation or in differentiated products that may command higher international
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
85
prices. Lower costs mean a higher rate of return to foreign investors. So, the productivity gains generated by host country exports will attract inward FDI and exports will lead inward FDI. Causality from inward FDI to exports is explained in a more detailed literature, generally in the context of the complementarity or substitutability of exports and FDI. Some of this literature is relevant to this study. The essence of the argument presented by Muchielli and Chedor (1999) is summarised briefly. Foreign investors in developing countries may have a better sense about the mix of host country export products that can succeed in world markets. Foreign investment brings with it international market knowledge and global distribution networks not necessarily available to local firms. FDI can create greater export potential and may alter a country's export product mix over time. In these circumstances, inward FDI will lead exports. There are mutual advantages to the partners in a joint venture formed to tap the export capacity of host country firms. The domestic partner benefits from new technology, physical capital, and managerial expertise, while for the foreign partner, an existing local producer provides an opportunity to enter the domestic market, which in the case of the PRC is a very large one. The FIEs are often attracted by cheaper labour costs, which give them a competitive advantage in their export markets. The significance of the FIEs in the development of China's export capacity since 1978 is described in Section 4.2 of this paper. In addition to these arguments, for a single direction of causation either from exports to FDI or from FDI to exports, there is always the prospect of a bidirectional link. This arises for example, if an outward-oriented export led development policy attracts FDI on a large enough scale for this foreign investment to create trade subsequently. Then, exports may lead FDI initially before FDI creates further trade. The evidence about the causal links between FDI and exports is inconclusive. An important study of this issue is conducted by Jun and Singh (1996) who examine the determinants of FDI in 30 developing countries. Part of this study is focused directly on the causality issue: One caveat in the empirical literature is that it is not clear whether FDI flows are attracted by economies that are already export-oriented (i.e., exports precede FDI flows) or whether FDI leads to export increases (i.e., FDI precedes exports). (Jun & Singh, 1996, p. 73)
Jun and Singh (1996) explore this causality issue for 11 high FDI recipient countries over the period 1969±1993. In four cases, Thailand, Ecuador, Portugal, and Greece exports Granger cause FDI and in only one case, namely Singapore, there is evidence of FDI Granger-causing exports. In the six remaining countries (Colombia, Costa Rica, Egypt, Malaysia, Mexico, and Nigeria), FDI and exports are unrelated. The PRC was not included in this analysis. The issue concerning the causal links between inward FDI and exports from the PRC is significant for the PRC's development policies and business development strategies and our review of the literature suggests that four relationships are conceivable: FDI ! EXP; EXP ! FDI; the relationship is bidirectional or there is no relationship between the PRC's exports and inward FDI.
86
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
3. Methodology There are four methodologies applied to the data set described in Section 4.1. These are discussed in sequence. The first step involves an analysis of the descriptive statistics outlining the distribution of average annual FDI (US$ million) across the PRC's provinces listed in Table 1. Here too, we identify the groupings of provinces according to the criterion developed in Section 4.1. The role of the FIEs in exports is analysed from the results in Table 2. There are two interesting aspects: first, there is the increased contribution of FIEs to provincial exports since 1992 and second, there is the relative importance of FIEs on provincial exports across the three groups of provinces.
Table 1 Average FDI by provinces or cities (in US$ million, 1983 ± 1998) Number
Provinces/ cities
Average FDI (1983 ± 1998)
Group
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
687.68 648.50 336.10 56.48 29.25 719.27 146.01 191.73 1460.93 1959.41 522.57 151.51 1638.42 134.35 988.59 210.22 285.93 243.38 4617.47 326.49 388.29 214.52 21.24 59.52 156.43 23.98 4.26 8.51 21.25
HFDI HFDI MFDI LFDI LFDI HFDI MFDI MFDI HFDI HFDI HFDI MFDI HFDI MFDI HFDI MFDI MFDI MFDI HFDI MFDI MFDI MFDI LFDI LFDI MFDI LFDI LFDI LFDI LFDI
A city or province is classified as HFDI, MFDI, or LFDI as follows: HFDI: average FDI US$500 million; MFDI: US$100 million < average FDI US$499 million; LFDI: average FDI < US$100 million.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
87
Table 2 Exports by FIEs (PRC and provincial groups, in US$ million, 1992 ± 1998) Year Exports PRC Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports HFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports MFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%) Exports LFDI Total exports (US$ million) FIEs' exports (US$ million) Proportion FIEs' (%)
Percent change total exports (1992 ± 1998) Percent change FIEs' exports contribution (1992 ± 1998)
1992
1993
1994
1995
1996
1997
1998
84,876.73
91,694.38
121,019.60
148,766.85
151,013.33
182,673.88
183,243.67
17,356.19
25,238.56
34,711.74
46,875.57
61,504.01
74,896.86
80,891.25
20.45
27.52
28.68
31.51
40.73
41.00
44.14
67,251.76
74,178.25
101,278.28
125,147.80
129,202.86
157,752.25
160,993.49
16,780.80
24,117.26
33,189.87
44,974.52
58,700.84
71,609.89
77,670.28
24.95
32.51
32.77
35.94
45.43
45.39
48.24
14,477.53
14,468.94
16,028.21
18,583.08
17,286.66
19,791.73
17,184.99
502.87
965.72
1297.61
1603.07
2418.86
2859.40
2821.37
3.47
6.67
8.09
8.62
13.99
14.45
16.42
3147.44
3047.19
3713.11
5035.97
4523.81
5129.90
5065.19
72.52
155.58
224.26
297.98
384.31
427.57
399.60
2.30
5.11
6.04
5.92
8.50
8.33
7.89
PRC
HFDI
115.89
139.39
MFDI 18.70
LFDI 60.93
366.07
362.85
461.05
451.02
Source: The data of total exports and exports by FIEs for 22 provinces, 4 autonomous regions, and 3 municipal cities were drawn from the Chinese Statistical Yearbook from 1995 to 1999.
This analysis of the relevant descriptive data is followed by a time series study of the relationship between inward FDI and exports from the PRC in the aggregate. This time series analysis will be based first on augmented Dickey±Fuller (ADF) and Phillips±Perron
88
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
(PP) tests for the presence of unit roots in the data on export and FDI levels and their first differences. Then, we proceed to test for a bivariate cointegration of aggregate monthly exports and FDI series. This bivariate cointegration model will establish if there is a longrun equilibrium relationship between exports and FDI. Finally, a standard error correction model (ECM) is formulated in both directions to see if there is a short-run adjustment towards long-run equilibrium in one or both directions. The ADF, PP tests for stationarity and Engle and Granger (1987) test for bivariate cointegration are well known and will not be repeated here; however, the causality test associated with this process will involve two ECMs as follows: FDIt a1
k X i1
EXPt a2
b1i FDIt
k X i1
b2i EXPt
k
k X i1
k
g1i EXPt
k X i1
g2i FDIt
k
k
r1 ECTt
r2 ECTt
k
k
e1t
1
e2t :
2
In Eqs. (1) and (2), ECT stands for the error correction terms. These are residuals generated from the cointegration of exports to FDI in Eq. (1) and from FDI to exports in Eq. (2). There are two bases for the presence of causality in these ECMs. The first concerns the short-run response of one variable, for example, exports to FDI and the second relates to the adjustment of the variables back towards long-run equilibrium. Standard Granger causality tests are misspecified if this second channel is not recognised. If in the estimation of Eqs. (1) and (2) both ^r1 and ^r2 are significant then causality is deemed to be bidirectional. However, if only one of ^r1 and ^r2 is significant, then a single direction of causation applies, and if neither is significant, then this second channel of causation is closed. The first channel may still apply even if the second is closed. In that case, it is appropriate to apply standard Granger causality modes to determine if a short-run interaction between inward FDI and exports exists. Ibrahim (2000) provides an interesting example of this situation. His bivariate cointegration tests fail to indicate a long-run relationship between Malaysian stock prices and the exchange rate (RM/US); however, he does find evidence of a short-run reaction between these variables when standard Granger causality tests are applied. A third aspect of our methodology is to determine if the findings of the national study are uniform across the provinces of the PRC. However, we acknowledge in Section 4.1 that there are too few time series observations to support an individual time series study for each province. In its place, three panel data sets are formed, one panel for each of the three provincial groups: high FDI (HFDI), medium FDI (MFDI), and low FDI (LFDI). From Table 1, it is evident that HFDI contains 9 provinces, while 12 and 8 provinces constitute the MFDI and LFDI panels, respectively. The 16 annual observations on exports and FDI when combined with the number of provinces in each group constitute the following data sets: HFDI 144 (9 16), MFDI 192(12 16), and LFDI 128 (8 16) observations.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
89
Standard Granger causality tests2 are applied to each of these panels. However, Granger tests are limited by their one sided nature. Granger tests relate to lags alone between two variables3. To overcome this limitation, we apply Sims (1972) test, which takes into account both lags and leads. The following interpretation (Eqs. 3±6) of the Sims' framework applies in this case: FDI f1 EXP; n past and m future lags of EXP;
3
FDI g1 EXP; n past lags of EXP;
4
EXP f2 FDI; n past and m future lags of FDI;
5
EXP g2 FDI; n past lags of FDI:
6
To test for unidirectional causality from FDI to exports, exports are regressed on past, current, and leading values of FDI. Causality from FDI to exports does hold if the coefficients of all leading (future) values of FDI are insignificant as a group. Likewise, exports lead FDI when all future values of exports are found to be insignificant in regression equations (3) and (4). 4. Data and results This section begins with a discussion of the data set that underpins the study and proceeds to an analysis of the results. The outcomes are discussed in the following sequence: first, there is a descriptive analysis of the data; second, the outcomes of the aggregate study for the PRC as a whole are discussed. Finally, the results of the provincial study based on grouped panel data are presented.
2
a FDIt
m X
b EXPt
li FDIt
i
i1 n X
dj EXPt
j1
ai FDIt
i
i1
FDI ! EXP : if
EXP ! FDI : if
n X j1
n X i1
bj EXPt
j
m2t :
j
m1t :
a^ i 6 0 in Eq: b and
n X i1
3
m X
a^ i 0 in Eq: b and
m X ^ dj 0 in Eq: a j1 m X ^ dj 6 0 in Eq: a: j1
We have made some common assumptions in estimating the Granger equations on panel data, in particular, that the disturbance terms in Eqs. (a) and (b) from footnote 2 have constant variances with finite moments and have independent increments.
90
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
4.1. The data set It was essential ab initio to determine if this study should focus on a national data set for the PRC as a whole, or if it should be pitched at a regional level and consider the link between FDI and exports in the PRC's provinces, cities, and autonomous regions (hereafter ``provinces''). Some authors, most recently, Sun and Chai (1998) and Tang (1998) argue that the provinces of the PRC are the appropriate basis because the culture, history, and structure of the individual provinces are so diverse. This argument turns out to be supported by the descriptive data included on Table 1 showing the distribution of inward FDI. Clearly, the range of variation in the average values of FDI received by individual provinces is large and so it is important to distinguish those provinces that receive large amounts of inward FDI from those that do not. However, it can be argued that the link between FDI and exports should be analysed for the PRC as a whole. An aggregate study of causality will inform policy makers as readily as separate provincial studies. For example, if a study based on China as a whole reveals a particular causal direction while an individual province or group of provinces displays a different outcome, then an explanation of this difference may be informative for policy. Our preference for separate national and provincial studies is also dictated by the availability of data. A monthly national time series for exports from the PRC and inward FDI was obtained from the China Monthly Economic Indicators for the years 1986 to 1999. This provides a time series of 168 observations, sufficient for the application of a more recent approach to causality tests based on ECM. However, reliable provincial data on exports and FDI is only available on an annual basis from various publications of the Chinese Statistical Yearbook (State Statistical Bureau of China [SSB], 1980±1998, various years) for the years 1983 to 1998 providing a data set of 16 observation that is not sufficient for individual provincial time series analysis. We address this problem by grouping the data in Table 1. Provinces with average annual inward FDI valued at US$500 million or more for the period 1983 to 1998 are referred to as HFDI recipients; those provinces receiving an average annual FDI in the range of US$100 million to US$499 million are classified as MFDI recipients, while provinces receiving less than US$100 million per annum on average over the reference period are labelled LFDI provinces. This approach to grouping the Chinese provinces is preferred to an alternative geographical method because FDI, as the basis of the classification, is also a key variable in this study. The alternative is a geographical basis that identifies separate coastal or eastern, middle, and western provinces of the PRC. However, there is a close correspondence between these geographical and functional modes of classification. The HFDI provinces are largely concentrated along the PRC's coast, while MFDI provinces are generally located in Central China and the LFDI provinces are to be found in Western China. 4.2. Analysis of descriptive data Table 1 contains the dollar value of inward FDI averaged over the years 1983±1998 for each of 29 provinces. Note that Chongqing is excluded from this list because it was
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
91
established only in 1997. Further, Tibet's receipt of FDI is negligible and it is excluded from further study. The following provinces receive FDI on average in excess of US$500 million per annum: Guangdong, Jiangsu, Fujian, Shanghai, Shandong, Liaoning, Beijing, Tianjin, and Zhejiang. These provinces comprise the HFDI group and are located along the Chinese coast or in close proximity to it. The role of FIEs in the expansion of the PRC's exports is an issue warranting some preliminary analysis. Indigenous Chinese firms and the FIEs generally export different types of goods. The FIEs' product range benefits from advanced technology, managerial expertise, and superior international networks. The foreign links established in the FIEs are largely concentrated in the HFDI group. This lends some plausibility to the argument that FDI leads exports or maybe that causality is bidirectional in the HFDI. However, among the LFDI and MFDI provinces concentrated in Western and Central China, respectively, it may be the case that exports are dominated by indigenous firms and that export growth is independent of FDI. Table 2 contains the proportion of the FIEs exports from the overall PRC and each of the three provincial groups since 1992. The concentration of the PRC's exports in the HFDI group is clear on Table 2. Of the export values measured in US$ million, 80% and 87% are accounted for by the HFDI provinces in 1992 and 1998, respectively. Further, the value of exports from the HFDI group have grown at a faster rate (139.39%) than export growth in either the MFDI (18.7%) and LFDI (60.93%). This result is not surprising given the concentration of FDI in the coastal provinces of the PRC and the effect of this on export development. This outcome constitutes prima facie evidence for causality flowing from FDI to exports among the HFDI group. A formal analysis will firm up this casual observation. A second important observation from Table 2 concerns the fall in total export values in 1998. Export growth slowed down in this year across the PRC and its provinces but remained positive for the nation and the HFDI group, but fell in the MFDI and LFDI groups. The behaviour of export growth in the MFDI group is of particular interest. Total exports in this group fell by 13%, but exports from the FIEs fell by less than 2%. This outcome may be explained by a greater degree of vulnerability of the indigenous exporters in the MFDI to a crisis such as that engendered by the Asian currency collapse in 1997. The remaining noteworthy feature of Table 2 is the increase in the proportion of total exports accounted for by the FIEs. This proportion has risen by 451% and 461%, respectively in the LFDI and MFDI, faster than the growth in the proportional contribution of FIEs exports to the total value of exports from the HFDI group. This outcome is suggestive of a particular pattern to the link between China's exports and its inward FDI. Initially, it is the export potential of a region that attracts FDI, in this case invested in FIEs. However, the effects of FDI on exports are subject to a gestation period in which time export potential is turned into export performance. 4.3. National (PRC) study of causality The test for causality in the relationship between inward FDI and PRC exports is based on a monthly time series dating from 1986 to 1999. We begin by testing each series for
92
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
stationarity applying standard ADF and PP procedures: the F and t statistics produced by these tests appear in columns 1 to 8 of Table 3. The results of these tests are inconclusive. The ADF F and t statistics suggests that FDI is nonstationary in levels while the corresponding PP statistics indicate that the FDI series is stationary. Similarly, the ADF t statistic for exports suggests nonstationarity, while the ADF F and PP F and t statistics indicate stationarity. In these circumstances, we are unable to conclude that the monthly series for PRC exports and inward FDI are stationary in levels, however the first difference of each series is clearly stationary: ADF and PP statistics exceed the relevant cut off scores in absolute value in each series. Taking the results of these stationarity tests collectively, it is appropriate to assume that both monthly time series are integrated of order one (I(1)) and to proceed to tests for cointegration. The results of Engle and Granger (1987) bivariate cointegration tests are disclosed in columns 9 and 10 of Table 3. The ADF test results indicate that a long-run equilibrium between inward FDI and exports does not exist. However, the PP test supports the view that there is cointegration of these time series. Greater reliance is placed on the PP results, because PP tests for cointegration are nonparametric and do not rely upon any distributional assumptions and are not characterised by the moments of a probability distribution function. This makes PP tests more robust to misspecification and data inconsistencies. The presence or otherwise of a long-run equilibrium relationship between inward FDI and exports for the PRC will be confirmed in the following ECM. The test results of ECM for the national study based on Eqs. (1) and (2) are reported in Table 4. The generic statistical characteristics of these estimates are sound. The RÅ2 for each equation is of acceptable magnitude. The Breusch±Godfrey (BG) test statistics for Eqs. (1) and (2) are 3.95 and 8.82, respectively, which indicate serial correlation is not present. It is clear that the coefficient of ECT in both equations is significant. These results confirm the presence of cointegration of the two variables: PRC's exports and its inward FDI. Statistically, the presence of cointegration rules out noncausality between the variables. As has been noted by Granger (1988), if a pair of I(1) series are cointegrated, there must be causation in at least one direction. The significance of ^r1 and ^r2 at 5% level in the estimates of Eqs. (1) and (2), respectively, suggests also that causality is bidirectional for the PRC as a whole. 4.4. Causality among PRC groups and pooled data The provinces of the PRC stand at different levels of development. This is recognised in our functional groupings of PRC provinces. As stated previously, these groups correspond broadly to the geographical regions of China: the eastern (coastal) region, which contains the HFDI group; the MFDI located in middle China; and the western provinces, which include the LFDI provinces. It is conceivable for the direction of causation between exports and FDI to differ from the national bidirectional pattern. The methodology applied for this purpose is detailed in Section 3 of the paper and involves causality tests based on Granger and Sims models. Each group of data is treated as a separate panel and Eqs. (a) and (b) contained in footnote 2 are estimated for the three panels.
0.93 1.72 3.13
* Significant at the 10% level.
2.11 9.21* 4.03
FDI EXP Critical value (10%)
30.73* 39.54* 4.03
F
4 t 9.60* 10.86* 3.13
4.50* 10.17* 4.03
F
5 t
6 3.41* 5.50* 3.13
3
t
2
1
F
ADF
PP
ADF
Variables
First difference
Levels
Stationarity
Table 3 Unit root test and cointegration test results: FDI and exports
155.11* 130.38* 4.03
F
7
PP 8 t 21.58* 19.79* 3.13
FDI EXP EXP FDI Critical value (10%)
Variables
9 0.97 1.48 3.50
ADF
10
PP
Cointegration
118.42* 143.80* 23.40
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99 93
94
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
Table 4 Causality tests based on bidirectional ECM Eq. (1) (dependent variable: FDIt) Coefficient b11 b12 b13 b14 b15 b16 g11 g12 g13 g14 g15 g16 r1 a1 R2 adj. BG df = 148
Estimate 0.908 0.808 0.838 0.787 0.604 0.10 0.023 0.012 0.042 0.046 0.029 0.024 0.434 1.042 0.617 3.951
Eq. (2) (dependent variable: EXPt) t ratio 9.814* 7.227* 6.843* 6.308* 5.137* 1.039 0.745 0.309 0.921 1.025 0.729 0.766 3.087* 1.882**
Coefficient g21 g22 g23 g24 g25 g26 b21 b22 b23 b24 b25 b26 r2 a2 R2 adj. BG df = 148
Estimate 0.829 0.858 0.571 0.271 0.285 0.094 0.053 0.111 0.473 0.975 0.467 0.159 0.335 4.21 0.458 8.82
t ratio 8.709* 6.579* 3.791* 1.802** 1.893** 0.931 0.191 0.326 1.286 2.592* 1.314 0.562 2.007* 2.541*
The number of lags applicable in this estimation was determined by applying the Akaike's FPE test to estimates of Eqs. (1) and (2). Six lags were chosen because this lag length minimised the FPE with highest adjusted R2. BG test results above indicate six lag model is free of serial correlation since the critical value of the c2 distribution is 9.48 at the 5% level for BG statistics. * Significant at the 5% level: critical value 1.96 (df = 148). ** Significant at the 10% level: critical value 1.65 (df = 148).
Provincial data are also pooled to get the overall picture for the PRC as a whole. The pooled series contains 464 observations. The application of Akaike's Information Criterion and Akaike's (1969, 1970) Final Prediction Error (FPE) criterion to both Sims and Granger causality models for each group panel and for pooled data suggests that the lag structures shown on Table 5 are appropriate.4 Four, two, and three lags were preferred for Granger tests of causality in the HFDI, MFDI, and LFDI groups, while three lags are required for pooled data. One backward and three forward lags are required for Sims test based on the three grouped panels. The preferred lag structure for the Sims test on pooled data is three backward and three forward lags. The grouped Granger and Sims causality test results for each of these groups in both directions (FDI ! EXP and EXP ! FDI) are listed on Table 6. The causal direction inferred for each group is summarised in the last column. The results in Table 6 indicate that the direction of causality between inward FDI and exports is bidirectional for both HFDI and LFDI groups based on the Granger and Sims tests. The HFDI group is composed of some
4
The appropriate lag length are chosen with minimised Akaike FPE and higher adjusted R2 as criteria.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
95
Table 5 Preferred lag structures HFDI (n = 144)
MFDI (n = 192)
LFDI (n = 128)
Pooled (N = 464)
Granger causality FDI ! EXP EXP ! FDI
4 4
2 2
3 3
3 3
Sims' model FDI ! EXP EXP ! FDI
(1,3) (1,3)
(1,3) (1,3)
(1,3) (1,3)
(3,3) (3,3)
The above lags were chosen after evaluating: two, three, or four lags for Granger causality and (2 back, 1 forward), (3,1), (1,2), (2,2), (3,2), (3,2), (1,3), (2,3), (3,3) in the case of Sims model. Four, two, and three lags were preferred for Granger tests of causality in the HFDI.
Table 6 The grouped Granger and Sim's causality for PRC (1983 ± 1998) FDI on export
Export on FDI
F ratio
Result
BG
F ratio
Result
BG
1
2
3
4
5
6
Causal inference
reject H0
1.90 (df = 4)
13.56* (4,132)
reject H0
0.98 (df = 4)
FDI ! EXP EXP ! FDI
1.96 (df = 2) 0.818 (df = 3)
7.85* (2,186) 24.80* (3,119)
reject H0
3.44 (df = 2) 2.48 (df = 3)
EXP ! FDI
Granger's test HFDI 3.28* group (4,132) MFDI group LFDI group
0.79 (2,186) 6.85* (3,119)
not reject H0
Pooled
14.02* (3,455)
reject H0
2.79 (df = 3)
53.73 (3,455)
reject H0
2.79 (df = 3)
FDI ! EXP EXP ! FDI
Sim's test HFDI group
46.62* (3,134)
reject H0
85.34 (df = 3)
10.35* (4,108)
reject H0
82.64 (df = 3)
FDI ! EXP EXP ! FDI
MFDI group LFDI group
2.99 (3,182) 41.70* (3,118)
not reject H0
115.74 (df = 3) 115.19 (df = 3)
2.93* (4,183) 16.07* (4,108)
reject H0
85.41 (df = 3) 8.57 (df = 3)
EXP ! FDI
Pooled
27.82* (3,450)
reject H0
258.43* (df = 3)
4.63* (3,450)
reject H0
285.58* (df = 3)
FDI ! EXP EXP ! FDI
reject H0
reject H0
The degrees of freedom are in the parentheses after the estimates. F test: 2,42, 3.04, 2.68, 2.68, 2.68, 2.68, and 2.60. BG: 9.487 (df = 4), 7.815 (df = 3), and 5.99 (df = 2). * The critical value at 5% level.
reject H0
reject H0
FDI ! EXP EXP ! FDI
FDI ! EXP EXP ! FDI
96
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
eastern coastal regions that are relatively advanced and includes the economic growth centres, such as Guangdong, Shanghai, and Beijing. They are major seaports and airports with superior transportation, communications, better service facilities, and human resources. For this reason, this group of provinces offers the most attractive investment environment in China. The HFDI group became major export providers and principal foreign investment recipients following the opening of China's doors in 1978. The HFDI group of provinces are relatively open to international trade and have well-established trade channels. Their export capacities have been strengthened with the help of foreign funds, especially in the SEZs. The growing exports of the HFDI group, in turn, not only stimulate the development of nonexport sectors, but also encourage regional economic growth as a whole, attracting more foreign investors by providing a more promising investment environment. The presence of an export trade encourages investment on the one hand while further injections of FDI stimulate export growth, so the link between FDI and exports is bidirectional. Most provinces in the LFDI group are backward inland regions with a relatively low degree of openness. The inland provinces were not allowed to trade with the rest of the world prior to the open-door policy. Their business activities were domestically oriented. Once an internationally oriented development policy was applied to China, export and foreign investment inflow became the two major determinants of economic growth among the LFDI. Some foreign investors seem to have taken advantage of the western region's mineral and energy resource abundance regardless of its poor infrastructure conditions. The results on Table 6 confirm the reciprocal relationship among the LFDI group. FDI is a prerequisite for export and economic development as a whole. On the other hand, exports facilitate the inflow of foreign funds. But the scale of FDI and export growth is relatively small in comparison with the HFDI group. The test results for the MFDI group suggests that the direction of causation runs from exports to FDI. The economies of most MFDI members are not as well developed as the HFDI group although some of them are located close to the eastern coast and had a successful export record prior to 1978. Exports, no doubt, played an important role in the economic development of the MFDI since most of the MFDI members did not receive government approval for foreign investment until at least 6 years after the first SEZs were established. Unidirectional causality can be explained by the gradualism with which China's doors were opened. The SEZs were established well ahead of the MFDI. The time lag in implementing the open-door policy across China did make a difference to regional development. This is reflected in the results on Table 6. Also, regionally different economic influences are evident in the decentralisation of the Chinese economy as a whole during the transition from a planned to a market economy. Results are also presented on Table 6 for the pooled data set comprising 29 provinces over the 16-year period. This is another way of analysing causality from the perspective of the PRC as a whole because it contains the cross section of provinces with a smaller set of time series observations. It is clear that these pooled data outcomes provide the same results as the national time series and ECM discussed in Section 4.2. In both Granger and Sims tests based on pooled data, the value of the F statistic leads to the rejection of the null hypothesis suggesting the absence of a causal flow from FDI to exports and then again from exports to FDI. So the conventional Granger and Sims tests produce the same outcome as the ECM
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
97
study. The pooled data analysis confirms that there is a bivariate causal relationship between inward FDI and exports for the PRC as a whole. The values of the BG test statistic in columns 3 and 6 of Table 6 indicate that these OLS estimates of the Granger but not Sims equations are free of serial correlation in general.5 We keep this in mind when analysing and interpreting the results of the Sims test. We take comfort from the fact that both Sims and Granger tests lead to the same conclusions. 5. Conclusions The major finding of this study is that bidirectional causality applies to the link between inward FDI and exports from the PRC. This is the clear outcome of the national time series analysis of the issue and is reinforced in a pooled regression model across 29 PRC provinces. We also find bidirectional causality between inward FDI and exports from the HFDI recipient provinces concentrated along the PRC coast. More than 80% of the PRC's total export values originate from these coastal provinces, so it is not surprising that the outcome for the HFDI panel mirrors the national pattern. This major outcome of the study provides some testimony to the effectiveness of the policy of opening the PRC's doors leading to greater interaction between the PRC and the rest of the world. In particular, FDI was encouraged from 1978 and beyond while the emphasis of the PRC's development policy was refocused to allow a greater emphasis to be given to export promotion. The FIEs contribution to the PRC's development is also emphasised in our statistics where the proportion of export values contributed by the FIEs doubled between 1992 and 1998. If it was the intention of the PRC government to create trade through foreign involvement in 1978, then the policy intention has been realised at least in the coastal provinces. However, the linkage between inward FDI and exports in the central provinces of the PRC (MFDI) differs from the national and HFDI study. In the case of the MFDI, causality appears to flow from exports or export potential to FDI. This provides some support for those researchers such as Dollar (1992) and Jun and Singh (1996) who find that outward-oriented economies particularly those exploiting their export potential attract more foreign investment and grow faster. It is their outward orientation and policies such as export enhancement that attract FDI and create trade subsequently. It is interesting to note that our results for the MFDI correspond with Jun and Singh's analysis, which finds that export potential leads FDI frequently. This finding for the MFDI may be indicative of a process of staged development. In a first stage, the degree of openness attracts FDI, which in time creates trade in the form of increased exports. If causality is assessed at an early stage of this process, then exports or potential exports could appear to be leading FDI, however, the direction could be reversed at a later stage when the FDI has turned export potential into real export growth. Our 5
Values of the BG test are below the relevant critical c2 value at a=.05 in testing that in the error term regressions: et r1 et 1 r2 et 2 r3 et 3 r4 et 4 m t all coefficients are simultaneously zero.
98
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
conclusion about the MFDI provinces may simply reflect an early stage of the export development process. The role of the FIEs in insulating the PRC's exports against the effects of economic downturns is evident in the case of the MFDI. The US value of total exports from the MFDI provinces fell by 13.1% between 1997 and 1998 as a consequence of the Asian currency crisis while FIE exports from the MFDI provinces fell by only 1.33%. The results for the LFDI provinces show some evidence of FDI playing a leading role in the development of exports from Western China. However, the development of the full potential of this region will depend on foreign capital being invested in infrastructure projects before the full export and domestic growth potential of this region of China is realised. Our finding of causality flowing from FDI to exports in the LFDI must be seen in this context. Acknowledgments The authors wish to thank the editor and anonymous referees for their advice, which has improved the manuscript substantially. The authors are responsible for any remaining errors and omissions. References Akaike, H. (1969). Fitting autogression for prediction. Annals of the Institute of Statistical Mathematics, 21, 203 ± 217. Akaike, H. (1970). Statistical predictor identification. Annals of the Institute of Statistical Mathematics, 22, 243 ± 247. Dollar, D. (1992). Outward-oriented developing economies really do growth more rapidly: evidence from 95 LDCs 1976 ± 1985. Economic Development and Cultural Change, 40 (3), 523 ± 544. Engle, R.F., & Granger, C.W.J. (1987). Cointegration and error correction: representation estimation and testing. Econometrica, 2(55), 251 ± 276. (March). Granger, C.W.J. (1988). Some recent developments in a concept of causality. Journal of Econometrics, 39, 199 ± 211. Hein, S. (1992). Trade strategy and the dependency hypothesis: a comparison of policy, foreign investment and economic growth in Latin America and East Asia. Economic Development and Cultural Change, 40 (3), 495 ± 521. Ibrahim, M.H. (2000). Cointegration and Granger causality tests of stock price and exchange rate interactions in Malaysia. ASEAN Economic Bulletin, 17 (1), 36 ± 47. Jun, K.W., & Singh, H. (1996). The determinants of foreign direct investment in developing countries. Transnational Corporations, 2(5), 67 ± 105 (August). Lucas, R. (1993). On the determinants of direct foreign investment: evidence from East and Southeast Asia. World Development, 21 (3), 391 ± 406. Mucchielli, J.L., & Chedor, S. (1999). Foreign direct investment, export performance and the impact on home employment: an empirical analysis of French firms. In: S.G. Lee, & P.-B. Ruffini (Eds.), New horizons in international business. Cheltenham, UK: Edward Elgar. Porter, M.E. (1990). The competitive advantage of nations. London: Macmillan. Sims, C.A. (1972). Money, income and causality. American Economic Review, LXII (4), 540 ± 552.
Q. Zhang, B. Felmingham / China Economic Review 12 (2001) 82±99
99
State Statistical Bureau of China. (1980 ± 1998). Statistical Yearbook of China (SYOC) 1980 ± 1998. Beijing: China Statistical Publishing House. State Statistical Bureau of China (various years). Statistical Yearbook of China. Beijing: China Statistical Publishing House. Stern, R.E. (1997). Foreign direct investment, exports and eastwest integration: theory and practice. (Chapter 10, pp. 329 ± 357) In: R.N. Cooper, & J. Cacs (Eds.), Trade growth in transition economies export impediments for Central and Eastern Europe. Cheltenham, UK: Edward Elgar. Sun, H., & Chai, J. (1998). Direct foreign investment and inter-regional economic disparities in China. International Journal of Social Economics, 25 (2 ± 4), 424 ± 427. Tang, K. K. (1998). Economic integration of the Chinas provinces: a business cycle approach National Centre for Development Studies, ANU, RSPS. (Working Paper No. 98/1, pp. 1 ± 16).