The present study usesfirm survey data of 1033 manufacturingfirms operating in Ethiopia in 2011 to examine the impact of Chinese outbound direct investment on the productivity of domesticfirms. Particularly, we attempt to answer two questions. Firstly, are Chineseowned (henceforth foreign)firms more productive than local ones? Secondly, does the presence of foreignfirms generate technology spillovers on domesticfirms operating in the same industry? Our empirical results show that foreignfirms are more productive and that their presence has different spillover effects on the productivity of domesticfirms. In particular, wefind that domesticfirms with higher absorptive capacity experience positive spillovers, while those with low absorptive capacity witness negative spillover. We alsofind that smallfirms and nonexporting firms benefit more from spillovers than do other types of domesticfirms. In this study, instrumental variables are used to address the potential endogeneity between foreignfirm presence and domesticfirm productivity
Trang 1Technology spillovers from Chinese outward direct investment:
The case of Ethiopia
a
Department of International Economics and Business, School of Economics, Xiamen University, Xiamen 361005, Fujian, China
b The Wang Yanan Institute for Studies in Economics, Xiamen University, Xiamen 361005, Fujian, China
a r t i c l e i n f o a b s t r a c t
Article history:
Received 28 January 2014
Received in revised form 10 November 2014
Accepted 3 January 2015
Available online 9 January 2015
in 2011 to examine the impact of Chinese outbound direct investment on the productivity of do-mesticfirms Particularly, we attempt to answer two questions Firstly, are Chinese-owned (henceforth foreign)firms more productive than local ones? Secondly, does the presence of foreignfirms generate technology spillovers on domestic firms operating in the same industry? Our empirical results show that foreignfirms are more productive and that their presence has different spillover effects on the productivity of domesticfirms In particular, we find that domes-ticfirms with higher absorptive capacity experience positive spillovers, while those with low ab-sorptive capacity witness negative spillover We alsofind that small firms and non-exporting firms benefit more from spillovers than do other types of domestic firms In this study, instrumen-tal variables are used to address the potential endogeneity between foreignfirm presence and domesticfirm productivity
© 2015 Elsevier Inc All rights reserved
JEL classification:
F21
D24
O33
L60
Keywords:
Absorptive capacity
China
Ethiopia
Outward direct investment
Spillovers
1 Introduction
Over the past decade, though insignificant in global terms,3China's outward direct investment (ODI)flows to Africa have increased rapidly The increase has generated interest and concern over the effects of China's ODI on developing Sub-Saharan host economies Some argue that Chinese ODI provides an alternative source of capital, technology and skills and that it has been instrumental in ful-filling financial and technological gaps for Africa (Brautigam, 2009; Foster, Butterfield, Chuan, & Pushak, 2008) On the negative side, some contend that the primary objective of Chinese ODI in Africa is tofind resources, and markets for their products where it drives African countries to resource-based economies and crowds-out local industries (Kaplinsky, 2008; Kaplinsky & Morris, 2009) China's ODI in Africa has generated considerable attention for several reasons One reason is the rapid pace at which China's ODI has risen and expanded in Africa.4Second, small and medium private Chinesefirms have recently become prominent investors in African manufacturing sector that there are uncertainties about the impact of their activities on the African economies where the investment
China Economic Review 33 (2015) 35–49
⁎ Corresponding author Tel.: +86 18959215421.
E-mail addresses: msmebratu@gmail.com (M Seyoum), renshui.wu@gmail.com (R Wu), Yangli1997@hotmail.com (L Yang).
1 Tel.: +86 13599515882.
2 Tel.: +1 703 473 7308.
3
The largest part of Chinese ODI goes to Asia, Latin America and Europe, respectively.
4
According to the Ministry of Commerce of China (MOC), China's ODI stock in Africa rose from 900 million dollars in 2003 to nearly 16 billion dollar in 2011.
http://dx.doi.org/10.1016/j.chieco.2015.01.005
Contents lists available atScienceDirect
China Economic Review
Trang 2is made.5The third reason relates to the way in which the Chinese are perceived to invest; there is a belief that Chinesefirms behave dif-ferently from otherfirms, either because the Chinese state is often behind ODI or because they have a different culture and institutional structure (Buckley et al., 2007)
The purpose of this study is to examine the impact of Chinese manufacturing ODI on the productivity of the Ethiopian manufactur-ing sector More specifically, we analyse two issues The first issue is to examine whether foreign-owned firms exhibit higher levels of productivity than domestic ones Then, we investigate whether the productivity of domesticfirms is correlated with the presence of foreignfirms in the same industry.6Identifying such effects would be consistent with the existence of intra-industry or horizontal technology transfer spillovers We further explore whether productivity gains stemming from horizontal spillovers vary with domes-ticfirms' characteristics
The study focuses on Ethiopia for two reasons First, Ethiopia has received a substantial amount of Chinese ODI in manufacturing, and is ranked among the top four recipient countries in Africa.7And over the past few years, China has emerged as the largest source of FDI in Ethiopia in terms of number of investment projects, with over 970 projects as of end-2012 (Ethiopian Investment Agency [EIA] data)
Second, to the best of our knowledge, no attempt has so far been made to systematically investigate the impact of Chinese manufacturing FDI on developing Sub-Saharan host economies.8Using a more disaggregated dataset for Chinese FDI in manufactur-ing, this empirical study is, to our knowledge, thefirst to present a detailed analysis of the impact of Chinese FDI on a host country in Sub-Saharan economies
The analysis is based on 1033 manufacturingfirms operating in Ethiopia in 2011 The data come from the survey of large and medium scale manufacturing industries conducted by theEthiopian Central Statistical Agency (CSA), Ministry of Finance and Economic Development (2012) The survey data coversfirms in the formal manufacturing sector, which employ 10 persons and more and use power driven machinery The dataset contains detailed information on the basic information of the establishment, own-ership structure, foreign equity participation, output, assets, employment, wages, input costs, location and other information Thefindings can be summarized as follows We find that foreign-owned firms are significantly more productive than their local counterparts, suggesting that there are direct benefits from Chinese FDI With regard to spillover effects, we find that Chinese FDI has different spillover effects on domesticfirms dependent upon their characteristics More specifically, our empirical results reveal that: (i) domesticfirms with high absorptive capacity (smaller technology gap with foreign firms) experience positive spillovers, while those with low absorptive capacity witness negative spillovers; (ii) smallfirms and non-exporting firms benefit more from spillovers than do other types of domesticfirms; and (iii) skilled labour of domestic firms does not enhance their capacity to attract FDI spillovers
The remainder of the paper is organized as follows.Section 2presents a brief overview of Chinese ODIflows to Ethiopia.Section 3
explains the theoretical framework for the role of foreign ownership on the productivity and technology spillovers.Section 4 intro-duces the econometric model, data and variable definitions InSection 5we present our regression results, whileSection 6concludes
2 Overview of Chinese ODIflows to Ethiopia
China's ODIflows to the African continent have grown rapidly over the past decade and Ethiopia is a good example of this trend
Fig 1shows the trends in China's ODIflows to Ethiopia from 2004 to 2010, using the Ministry of Commerce of China (MOC) 2010 Statistical Bulletin of China's Outward Foreign Direct Investment China's ODIflows to Ethiopia rose from virtually zero in 2004, reaching 24 million dollars in 2006 to a peak of 73.4 million dollars in 2009 before it had declined to 58.5 million dollars in 2010 According to MOC data, the stock of ODI from China to Ethiopia in 2010 was 368 million dollars
The official statistics reported by MOC (Fig 1) seem to understate the true investment volume of Chinese ODI in Ethiopia Accord-ing to the EIAfigures, the accumulated stock of Chinese ODI in Ethiopia stood at nearly two billion dollars at the end of 2010.9As shown inTable 1, the amount of annual ODIflows from China to Ethiopia has been increasing rapidly, albeit from a low base: from 181.71 million dollars between 2002 and 2004, to 414.29 million dollars during the period 2005–2007 and rose further to over one billion dollars between 2008 and 2010 Similarly, the total number Chinese investment projects in Ethiopia reached 944 projects in
2010, a whopping 782.24% increase from the period 2002–2004 According to EIA data, out of the 944 investment projects, 632 projects (66%) are engaged in the manufacturing sector
Fig 2looks at China's ODIflows as a share of total FDI inflows for the past decade There is a visible trend exhibiting that Chinese ODI has been growing very fast, and that it is taking over the principal position in new FDI attracted to the country In terms of the number of investment projects, China's contribution rose from 11% in the period of 2000–2005, reaching 29% in 2007 to a peak of 32% in 2008 During the period 2006–2011, China took the top place contributing 25% of the total FDI attracted to Ethiopia
5
According to Shen (2013) estimates, small and medium private Chinese enterprises, predominantly concentrated in manufacturing and service industries, accounted for 55% of all Chinese investment projects in Africa by the end of 2011.
6 In our case, sectors are defined at a more aggregate level, hence some intra-industry spillovers may, in reality, capture vertical spillovers (see Table 2 ) 7
According to Shen (2013) estimates, based on data from MOC and African host governments, Nigeria, South Africa, Zambia, Ethiopia and Ghana (in that order) are the top five Chinese ODI recipient countries in Sub-Saharan Africa.
8 We use “ODI” and “FDI” interchangeably.
9 It is fair to argue that the EIA data captures Chinese ODI in Ethiopia more comprehensively and accurately than the official data reported by MOC, simply because the
Trang 33 FDI and productivity spillovers: literature review
FDI is considered to influence the productivity and competitiveness of host-country economic activities for at least two quite different reasons First, multinationalfirms bring to their host-country superior productive assets such as technological know-how, managerial and entrepreneurial skills, and marketing techniques Second, there are spillovers of technology transfer from foreign firms, which affect the productivity of domestic firms Spillovers occur when foreign firms cannot fully internalize all quasi-rents as
a result of their productive activities
3.1 Are foreignfirms more productive than local ones?
The orthodox literature suggests that foreignfirms undertake FDI to exploit firm-specific ownership advantages that arise from the possession of intangible assets such as better access to advanced technology inputs and more efficient organization in production and distribution In addition, they tend to operate on a lower (production and distribution) cost curve than domesticfirms, which allows them to compute successfully with domesticfirms who have intimate knowledge of local market conditions including customs, consumer preferences, legal environment and business practices If so, other things being equal, we expect foreignfirms to be more productive than domestic ones
Consonant with the conventional literature, a number of studies conducted in Sub-Saharan African countries confirm that foreign firms are more productive than domestic firms A recent Africa investor survey conducted by the United Nations Industrial Develop-ment Organization (UNIDO) in 2011 reveals that, on average for all surveyed countries, foreignfirms are 11% more productive than domestic ones.10For country-specific studies in Africa,Melese and Waldkirch (2011)use a sample of 6574 manufacturingfirms for the period 2002 to 2009 to show that foreign-ownedfirms are more productive than domestically owned firms in Ethiopia In Kenya,Gachino (2013)employsfirm survey data for 180 manufacturing firms, and finds similar productivity advantages in favour foreignfirms
Thus, thefirst issue examined in this paper is whether Chinese-owned firms exhibit significantly higher levels of productivity than domesticfirms in Ethiopia If so, the presence of Chinese firms can be expected to impact positively on the host-country, because their higher productivity levels help productivity in their respective industries to shift upward, which is also reflected in the aggregate productivity of the host country (Caves, 1996)
3.2 Are there any spillover gains from FDI?
There is a large literature on the existence and direction of spillovers of technology transfer from FDI The existence of spillovers is based on two assumptions Thefirst assumption lies on the fact that multinational firms have better access to advanced technology and other advantages and have, therefore, higher levels of productivity (Caves, 1996) The second is related to the fact that technology-based assets transferred to the host economy have the characteristics of a public good (Romer, 1990) Dissemination and appropriation of their qualities may take place in the form of unintentional transmission or intentional transfer from
multination-al to locmultination-alfirms through demonstration effects, worker mobility or direct linkage with local agents
Furthermore, multinationalfirms may inject a higher level of competitive intensity in the host-country, which may produce addi-tional spillover benefits Foreign firms typically enter markets characterized by high entry barriers and consequently strong monop-olistic distortions Their entry thus may reduce monopmonop-olistic distortions and raise the productivity of local agents by improving resource allocations in the host country (Caves, 1996) On the other hand, negative effects may arise from competition if foreign firms, which happen to produce at a lower marginal cost, gain market shares from local firms.Aitken and Harrison (1999)argue that even if localfirms benefit from technology spillovers, productivity of local firms may still decline (rise in average cost) if the output demanded from them is reduced as foreignfirms take a large share of the market However, local firms may also respond to foreign competition by making better use of existing resources or investing in new technologies in order to maintain their market share and, therefore, drive a higher level of productivity on the process (Blomstrom & Kokko, 1998) Technical efficiency in the indus-try is thus improved
Spillovers of technology transfer from FDI also depend on the characteristics of domesticfirms, which shape absorptive capacity to internalize spillovers (Farole & Winkler, 2012) Domesticfirms' absorptive capacity is their ability to recognize, assimilate and apply
10 The survey was carried out in Burkina Faso, Burundi, Cameroon, Cape Verde, Côte d'Ivoire, Ethiopia, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mali, Mozambique,
Table 1
China's outward FDI flows to Ethiopia, 2002–2010.
Source: authors’ computations based on data from the Ethiopian Investment Agency (EIA).
37
M Seyoum et al / China Economic Review 33 (2015) 35–49
Trang 4outside knowledge to commercial ends (Cohen & Levinthal, 1989) It is argued that where localfirms are constrained by their limited absorptive capacity, the expected productivity spillover benefits of FDI either do not show up or are negative (the latter case may occur if the entry or presence of foreignfirms shrinks local firms' market share) Conversely, the greater the local firms' capacity to absorb new technology and processes, the more productivity spillover benefits The factors that influence local firms' absorptive capacity include (i) technology gap between foreign and localfirms; (ii) local firm's size; (iii) exporting behaviour of domestic firms; and (iv) share of skilled labour at the firm level
The technology gap between foreign and localfirms has been acknowledged as one of the most important moderating factors for the realization of FDI spillover potential (Findlay, 1978; Wang & Blomstrom, 1992; Kokko, Tanzani, & Zejan, 1996, among others) However, studies on the role of technology gap for spillover effects of FDI conflict For instance, some argue that a large technological gap between foreign and localfirms should enhance positive spillovers, because the potential for improve-ment is large (Sjoholm, 1999; Wang & Blomstrom, 1992) Others argue that domesticfirms need some minimum amount of technical capacity to be able to benefit from spillovers and thus smaller technology gap between foreign and domestic firm re-sults in larger spillovers (Blomstrom, Globerman, & Kokko, 1999; Blomstrom & Kokko, 1998) WhileBlalock and Gertler (2009)
suggest that if the technology gap between foreign and localfirms is too large or too small productivity spillover benefits may not be realized
Despite many attempts by others (cf.Hale & Long, 2006; Kokko, 1994; Sjoholm, 1999) to examine the role of technological gap on FDI spillovers, results are inconclusive.Kokko (1994)employs cross-section industry level data for Mexico to show that productivity spillovers are smaller because of high technology gap between foreign and localfirms Similarly,Kokko et al (1996)extendKokko's (1994)analysis for Mexico to examine the role of technology gap on productivity spillovers, using cross-sectionfirm-level data for Uruguay Theyfind evidence supporting the notion that smaller technology gap between foreign and local firms enhances productiv-ity spillovers.Hale and Long (2006)observe the same phenomenon in China On the contrary,Sjoholm (1999), using cross-section data for 8086firms in Indonesia, shows that spillovers from FDI are found in sectors with a high degree of competition and less ad-vanced technology Putting it differently, hefinds that the larger the technology gaps between domestic and foreign firms, the larger the spillovers—it seems that larger technology gap leaves more ground for improvement
Another important characteristic that affects the extent and nature of FDI spillovers isfirm size (Dimelis & Louri, 2004; Sinani & Meyer, 2004) Larger domesticfirms tend to have stronger capacity to compete with foreign firms and imitate their technology and management practices (Crespo & Fontoura, 2007) Moreover, they are better positioned to spreadfixed costs of R&D over a larger sales base and hence are able to exploit economies of scale and scope in R&D activities (Cohen & Levinthal, 1989) They also pay better wages and thereforefind it easier to attract workers employed by foreign firms (Markusen & Trofimenko, 2007) On the other hand, larger domesticfirms may be competitive and hence operating at their maximum efficiency; therefore, the scope for technology transfer from foreignfirms could be limited (Dimelis & Louri, 2004) Others also suggest that smallerfirms may benefit more from multinationalfirms if, for example, they are endowed with higher proportion of skilled labour (Sinani & Meyer, 2004); moreover, they may operate at suboptimal efficiency level lacking the necessary technology and knowledge of productive assets and therefore can enjoy higher spillover benefits from FDI presence (Dimelis & Louri, 2004)
Aitken and Harrison (1999)employfirm level panel data to assess the effects of foreign equity participation on the productivity of domesticfirms in Venezuela They find significant negative spillovers from FDI for small enterprises only (less than 50 employees), which they attributed to market-stealing effect Likewise, the study byBoly, Coniglio, Prota, and Seric (2013)usesfirm level data for 19 Sub-Saharan economies to show that large and youngfirms enjoy more positive spillovers than do other type of firms In con-trast, based on a sample of 3742 manufacturingfirms operating in 1997 in Greece,Dimelis and Louri (2004)find that productivity
Fig 1 China's outward FDI flows to Ethiopia, 2004–2010 Source: MOFCOM, 2010 Statistical bulletin of China's outward foreign direct investment.
Trang 5spillovers accrue mostly for small localfirms.Sinani and Meyer (2004)observe the same phenomenon in Estonia, especially for small enterprise with a higher proportion of skilled labour
Exporting behaviour has been linked to a domesticfirm's capacity to absorb new technology and management practices Two opposing views exist in the literature on the role of exporting behaviour in capturing FDI spillovers On the one hand, some argue that domestic exportingfirms tend to have a stronger capacity that places them in a better position to mitigate negative spillovers
of FDI—because they are generally characterized by higher productivity, be it via self-selection process where more productive firms become exporters or learning by-exporting (Melitz, 2003; Crespo & Fontoura, 2007)
On the other hand, some suggest that local exportingfirms have exposure to additional channels through which they can learn about advanced knowledge, skill and technology from their international connections and hence the potential for FDI-induced exter-nalities is limited (Sinani & Meyer, 2004) In addition, exportingfirms may already enjoy higher productivity and hence there may be little knowledge spillover to be transferred to them from FDI Besides, local exportingfirms may not have the incentives to upgrade their technology if they face lower competitive pressure from FDI (assuming that foreignfirms do not export to the same market), which lowers the scope and magnitude of positive FDI spillovers
In the bulk of existing empirical literature, studiesfind no clear evidence whether exporting behaviour enhances or reduces the productivity spillovers from FDI For instance,Barrios and Strobl (2002)employfirm level panel data for Spanish manufacturers from the period 1990 to 1998 to show that the gains from FDI spillovers are larger for exportingfirms.Girma, Gorg, and Pisu (2008)observe the same phenomenon in the U.K for intra-industry spillover So doSchoors and Van der Tol (2002)in Hungary In contrast, several empirical studiesfind little or no productivity spillovers for exporting local firms (cf.Sinani & Meyer, 2004; Blomström and Sjöholm, 1999, among others)
A domesticfirm's capacity to absorb new technology and managerial practices is also linked to its share of skilled labour The economics literature posits that investment in skilled labour and R&D not only increases innovation, but also raises a firm's ability to recognize, assimilate and apply outside knowledge to commercial ends (Cohen & Levinthal, 1989; Glass & Saggi, 2002; Sinani & Meyer, 2004).Blalock and Gertler (2009)apply panel data of Indonesian manufacturers from 1988 to
1996 to argue that the share of skilled labour (the percentage of workers with collage degrees) considerably increases domes-ticfirms' productivity spillovers from FDI However, for highly skilled countries such as the U.K.,Girma and Wakelin (2007)
confirm such a finding for small enterprises only whereasSinani and Meyer (2004)find that a larger share of skilled labour enhances positive spillovers for large enterprises in Estonia On the other hand,Cuyvers, Soeng, Plasmans, and Van den Bulcke (2008)find that a firm's human capital intensity does not determine its ability to adopt foreign technology in Cambodian manufacturing sector
Therefore, as discussed above, the characteristics of the recipientfirms are important mediating factors for technology spillover potential to turn into actual technology spillovers The absence or presence of suchfirm-specific characteristics may crucially influ-ence observed spillovers from FDI and thus not taking them into account can bias empirical results It is along these lines that in this study, we focus on how spillovers differ according to the characteristics of domesticfirms in addition to estimating the produc-tivity increase (or decrease) of domesticfirms in the same industry
4 FDI and productivity spillovers: estimation strategy, data and variables
4.1 Estimation strategy
If the superior technology embodied in foreign-ownedfirms is diffused to local firms, productivity levels of local firms should increase To examine productivity spillovers from foreign-owned to locally ownedfirms, we follow an approach similar to that
Fig 2 The percentage of Chinese investment of total FDI Source: Shen (2013), based on EIA.
39
M Seyoum et al / China Economic Review 33 (2015) 35–49
Trang 6taken by earlier literature and estimate an augmented Cobb–Douglas production function FollowingDimelis and Louri (2004), we specify the following general form for the production function:
Yi¼ LαiKβiMγie
X
where Yidenotes the output offirm i, measured by gross sales; Li, Ki, Midenote labour,fixed capital and intermediate material inputs used in eachfirm i; α, β, γ are the elasticities of output with respect to capital, labour and intermediate material inputs, respectively
Xipdenotes the p-th control variable which is not explicitly included, for example, the one correlated with FDI;λois a constant param-eter Finally, the error term,εicaptures all other unobservable factors influencing output Log-linearizing Eq.(1)yields the following estimation equation:
lnYi¼ λ0þ αlnLiþ βlnKiþ γlnMiþX
p
p ¼1
Since our main objective relates to labour productivity, we transform Eq.(2)to obtain its labour-intensive form
ln Yi
Li
¼ λ0þ βln Ki
Li
þ γln Mi
Li
þ α þ β þ γ−1ð Þ lnLiþX
p p¼1
Eq.(3)can be rewritten by adding several control variables, as follows:
lnLPi¼ α0þ α1lnKIiþ α2lnMIiþ α3lnLiþ α4SKILLiþ α5lnAGEiþ α6SCALEiþ α7C FORiþ εi
LP stands for labour productivity, which is influenced by capital intensity (KI), material inputs intensity (MI), labour inputs (L), skilled labour (SKILL),firm age (AGE), firm scale (SCALE) and foreign presence (CFOR).Appendix Adescribes the explanatory variables adopted in the econometric investigation, which follow the theoretical and empirical literature discussed above We expect coef fi-cients for the independent variables to be positive and significant A statistically significant positive coefficient correlated with CFOR implies that foreignfirms have higher levels of productivity than their locally owned counterparts
To examine whether the presence of foreignfirms affects the productivity of local firms in the same industry, we build on the standard equation encountered in Eq.(4) The difference now is that the left-hand side concerns only localfirms, rather than all, firms:
lnLPi j¼ α0þ α1lnKIi jþ α2lnMIi jþ α3lnLi jþ α4SKILLi jþ α5lnAGEi jþ α6SCALEi j
where subscript j stands for sector j; LP, KI, MI, L, SKILL AGE and SCALE are same as Eq.(4) FDI is the measure of foreignfirms' presence in sector j
Table 2
Distribution of firms with foreign capital by industrial sectors in 2011.
Source: authors’ computations based on data from the Central Statistical Agency (CSA) of Ethiopia, Large and Medium Manufacturing and Electricity Industries Survey 2012.
Trang 7Following the literature, we employ the share of foreignfirms' output as percentage total output at the 4-digit sectorial level as our measure of intra-industry FDI presence
FDIYj ¼
X
i∈ jYif or
X
where i∈ j indicates a firm in a given sector, Yiisfirm-level output in a given sector, and Yiforis the output if thefirm is foreign.11A statistically significant positive coefficient correlated with FDI would be consistent with the existence of intra-industry or horizontal technology spillovers
To examine the effects of domesticfirm's characteristics on technology spillovers, we will add four measures of absorptive capacity namely technology gap (TG),firm size (SIZE), skilled labour force (SKILL) and exporting behaviour (EXP) and the interaction term between these measures and our FDI presence as additional explanatory variables in Eq.(5)
TG is measured by the difference of the average labour productivity (the ratio of total sales to total employment, weighted byfirm asset size) of foreignfirms at the 4-digit industry level and the labour productivity of a local firm in the same industry, following
Kokko et al (1996).12A negative value for the individual domesticfirm indicates that the local firm is more productive than the average foreignfirms, while a positive value indicates that the firm is less productive We define a positive gap dummy that takes the value one when TG is positive and zero otherwise This dummy variable allows the isolation of localfirms with low absorptive capacity We interact the positive gap dummy with FDI presence and expect it to have a negative effect
Following the literature,firm size is defined by the number of workers employed in a firm where firms with 50 or more employees are considered as large andfirms with less than 50 employees are categorized as small firms.13Accordingly, we define a SIZE dummy that takes the value 1 if thefirm has more than 50 employees and 0 otherwise We interact the SIZE dummy with our FDI presence to examine the effect of SIZE on technology transfer.14In a similar vein, we include an interaction term between the individual domestic firm's share of high skilled-labour to its total workforce (SKILL) and our FDI presence to determine if SKILL has an effect on technology spillovers
We further explore if FDI spillovers differ between domestic exporters and non-exporters Afirm is considered to be an exporter if its export sales are equal to or greater than 5% of its total sales We consider this simple (and also most widely used) definition adequate for the sake of identifying differences infirm characteristics between domestic exporters and non-exporters Accordingly,
we spilt the sample into exporting and non-exportingfirms and estimate Eq.(5)separately for each group
11
A firm is considered foreign if the share of foreign ownership is equal to or greater than 10%.
12
As an alternative measurement, we measured TG as the difference of the (un-weighted) average productivity of foreign firms at the 4-digit industry level and that of
a local firm in the same industry The results obtained following this approach were consistent to the ones reported.
13 The definition of large and small firms varies by country The Central Statistical Agency of Ethiopia follows similar classification where firms are defined as: small (between 10 and 19 employees), medium (between 19 to 49 employees) and large (greater than 50 employees).
14 An alternative approach is to split the entire sample into large and small firms and estimate Eq (5) separately for each group The results obtained following this
Table 3
Number and size, exporting behaviour and productivity of domestic and foreign firms.
Source: authors’ computations based on data from the Central Statistical Agency (CSA) of Ethiopia, Large and Medium Manufacturing and Electricity Industries Survey 2012.
Exporting behaviour of firms
Productivity of firms
Y
L (mean value a
Small firm (b50 employees).
a In thousand birr.
41
M Seyoum et al / China Economic Review 33 (2015) 35–49
Trang 8Given the cross-sectional nature of our dataset, solving the problem of estimating technology spillovers from FDI when FDI is en-dogenous is a major challenge The major challenge arises from identifying the direction of causality between foreign output share and domesticfirm productivity Foreign firms may enhance the productivity of domestic firms through technology and knowledge trans-fers and spillover, but it may be also the case that they are attracted to certain industries that already exhibit higher productivity In the latter case, the estimated coefficient on foreign output share would be biased upward On the contrary, if foreign firms are attracted to
or are mainly concentrated in industries that exhibit lower productivity, the estimated coefficient on foreign output share would be biased downward This endogeneity concern is addressed adequately by using instrumental variables two-stage least squares (IV 2SLS) approach, as explained in greater detail below
4.2 Data and instruments
The data employed in this study is the annualfirm survey data from the large and medium scale manufacturing industries conducted in 2012 by the CSA, Ministry of Finance and Economic Development This survey data was obtained directly from CSA The survey coversfirms in the formal manufacturing sector, which employ 10 persons and more and use power driven machinery The survey data is well suited to examine spillover effects from FDI, because it contains information on variables that are commonly applied in econometric estimation offirm level production functions Particularly, the data includes financial information as well as a wide range of indicators onfirm characteristics such as foreign ownership, employment and skills, exporting behaviour and region and sector offirms Sectors are defined according to the International Standard Industrial Classification (ISIC Revision-3.1), but in some cases are further aggregated
The number offirms used in our econometric estimation is reduced to 1033 (out of 1937) The number of firms in the estimation is reduced for the following reasons: i) we omitfirms for which we cannot calculate key variables due to missing information; and ii) we include only sectors that have foreignfirms (seven sectors at the 4-digit level in the survey had no foreign presence) After these considerations, thefinal data consists of 943 observations for domestic firms and 90 for foreign firms.15
The sectoral distribution offirms with foreign capital in 2011 is presented inTable 2 The relative presence of foreign capital and ownership is more visible, in terms of percentage shares, in bodies for motor vehicles, trailers and semi-trailers; spinning, weaving andfinishing of textiles; and chemical and chemical products.Table 2indicates that there is difference in the sector wise distribution among foreignfirms implying the need to control for industry specific factors that influence firm level productivity
Table 3below presents summary statistics on number and size distribution offirms, exporting behaviours of firms and productivity byfirm ownership In terms of number and size distribution of firms, the majority of foreign firms (60%) are considered to be largefirms (N50 employees), while the majority of domestic firms (71%) are categorized as small firms (b50 employees) Furthermore, about 93% of the domestic firms report that they do not export some or all of their production Interestingly, and consistent with other surveyfindings (cf.Shen, 2013; UNIDO, 2011; World Bank, 2012),Table 3suggests that the majority of Chinese investors in our survey data (about 84%) are essentially local market seekers, i.e., they do not export some or all of their production In terms of labour productivity, Chinesefirms are on average 2 times more productive than domesticfirms
Moreover, there is a definitive preference for independent market entry among Chinese firms As reflected inTable 4
below, 89% of the Chinesefirms in the survey prefer wholly or majority ownership (N50%) as opposed to minority ownership Only 11% of the surveyedfirms are minority ownership with local partners This trend is consistent with other survey findings
in Ethiopia (World Bank, 2012) and other African countries that show that Chinesefirms tend to be wholly Chinese owned (Shen, 2013)
To identify the causal relationship between FDI presence and domesticfirm productivity, we instrument our measure of FDI pres-ence using sectorial measure of sector targeting by the EIA The study ofHarding and Javorcik (2012)was thefirst to utilize informa-tion on investment promoinforma-tion efforts to attenuate endogeneity concerns So doFarole and Winkler (2012)for a cross-sectional setting analysis of FDI and spillovers The choice for this instrument is based on two assumptions First, this variable must be correlated with
15 To maintain confidentiality, the dataset that was provided to us has no firm identifiers However, by following the information in the survey data on sector activity of firms, location (region, city/town, district, house no.) of firms, and phone address of firms, one is able to identify and determine firms' ownership.
Table 4
Foreign firms by ownership type.
Source: authors’ computations based on data from the Central Statistical Agency (CSA) of Ethiopia, Large and Medium Manufacturing and Electricity Industries Survey 2012.
Ownership type
Note: Minority ownership (b50).
Trang 9FDI presence Second, it must be uncorrelated with domesticfirm productivity We believe that sector targeting by the EIA is well suited as an instrument, as it is likely to meet both assumptions
First, it is reasonable to assume that sector targeting by national investment promotion agencies (IPAs) is correlated with FDI presence, because the primary objective of such policy tool is to identify and promote investment opportunities in host countries
to attract FDI (UNCTAD, 2001; UNIDO, 2011; Wells & Wint, 1990) Sector targeting is deliberated to be the best policy tool for invest-ment promotion activities, because more intense efforts focused on a few priority sectors are likely to lead to greater FDI inflows than less intense across-the-board efforts to promote FDI (Loewendahl, 2001; Proksch, 2004) Employing a difference-in-differences approach,Harding and Javorcik (2011)show that targeting a particular sector by a national IPA leads to more than doubling of FDI
inflows into the sector, with significant time lag effects between these two So doBobonis and Shatz (2007)andCharlton and Davis (2006)
Second,Harding and Javorcik (2012)show that the choice of sector targeting by IPAs is less likely to be driven by the quality of domesticfirms or industries in a host country, particularly in the context of developing countries like Ethiopia Instead, IPAs choose targeting a particular set of sectors in the hope and expectation that greater FDI inflows into these sectors can make a positive contribution to economic growth by generating jobs, bringing additional capital, and transferring new technologies and expertise
to host-countries
In the past decade, under different national strategy plans (Plan for Accelerated and Sustained Development to End Poverty [PASDEP] and Growth and Transformation Plan [GTP]), the Government of Ethiopia and with the help from the World Bank and other multilateral institutions has identified a particular set of priority sectors for investment aimed at attracting greater FDI inflows into these sectors Consequently, the Ethiopian government has identified leather and leather products, textile and garment, and agro-processing industries as priority sectors to potential international investors Also, under the GTP, the list of priority sectors was extended to include metal and engineering, and chemicals and pharmaceuticals
The primary selection criterion for these sectors is to enhance their competitiveness internationally and increase domestic value addition and sales in line with the country's latent comparative advantage (World Bank, 2011) For instance, the country has a strong latent competitive advantage in producing leather and leather products due to the outstanding quality and quantity
of local leather and extremely low labour costs Moreover, Ethiopia's large population (about 92 million), cheap and abundant labour and a fast growing economy (averaging 10.6% per year over the past decade) add to its attraction as a FDI host economy in textile and garment, and food processing industries (World Bank, 2012) The reason for the inclusion of metal and engineering, and chemicals and pharmaceuticals as priority area for investment is to reduce the country's heavy dependent on imports of these products (World Bank, 2011)
These industries have continued to receive special interest and extensive support programs from policymakers Accordingly, the Government of Ethiopia has implemented various reforms and continuously provided a basket of incentives, such as tax holidays and tariff-free policies for FDI equipment imports to attract and nurture investments in these priority sectors As can be seen from our survey data (Table 2), the sectoral distribution of foreignfirms is concentrated on these priority areas, which suggests that the survey provides a good characterisation of the general trends of FDI in Ethiopia
Table 5
Impact of foreign ownership on productivity.
Dependent variable: ln Y
L
of domestic and foreign firms
(0.074)
Notes:
1 Numbers in the parenthesis are the heteroscedasticity robust standard errors.
2 The symbols ***, ** and * indicate 1%, 5% and 10% significance levels, respectively.
3 All regressions were performed with 11 2-digit industry dummies, the inclusion of which was based on the F-statistics The excluded dummy corresponds to the last category of wood and furniture industries.
4 Please refer to Appendix A for the definition of lnKI, lnMI, lnL and CFOR variables.
43
M Seyoum et al / China Economic Review 33 (2015) 35–49
Trang 10FollowingHarding and Javorcik (2012), we instrument our foreign output share variable with two sectorial measures of sector targeting by the EIA Thefirst measure is an indicator variable (referred to as Sector targetedST) that equals one if the sector has been targeted by EIA in a certain year, and zero otherwise Since our foreign output share variable reflects the presence of foreign firms over a longer period of time, the second measure is constructed by aggregating the dummies over the period 2007–2011 to obtain a continuous variable, referred to as Length of sector targetingLST The sum may vary from 0 (no targeting over the period under study) to 5 (continuous targeting over the period under study) To control for non-linearities, we use the second measure in log form (adding 1 before taking the log)
We believe that utilizing information on sector targeting by EIA is well suited as an instrument, as it is uncorrelated with domestic firm productivity,16
but is correlated with foreign output share variable, especially since there is a time lag between these two measures InSubsection 5.2, we report thefirst stage statistical results for instrument validity
5 Empiricalfindings
In this section, we introduce the results of the empirical analysis All regression results follow various estimations of Eq.(4)and include industry dummies at the two-digit level to control for productivity differences across industries All reported standard errors are robust to heteroscedasticity We start by presenting productivity differences between foreign and domestic ownedfirms, and proceed showing the results for productivity spillover effects employing ordinary least squares (OLS) and instrumental variables two-stage least squares (IV 2SLS) procedures
5.1 Estimation results
Table 5presents the estimation results for the labour productivity differences forfirms in the Ethiopian manufacturing sector by ownership Column 1 is the regression of labour productivity on all independent variables used in the estimation, except the foreign
16 It is possible that the government of Ethiopia has been selective in allowing FDI in sectors, which are more or less productive than others If so, our instruments may have possible biases because the exclusion restriction will probably not be satisfied But due to data limit, we are not able to fully address this issue in the current paper.
Table 6
FDI presence and productivity spillovers.
Dependent variable: ln Y
L
of domestic firms only
Notes:
1 Numbers in the parenthesis are the heteroscedasticity robust standard errors.
2 The symbols ***, ** and * indicate 1%, 5% and 10% significance levels, respectively.
3 All regressions were performed with 11 2-digit industry dummies, the inclusion of which was based on the F-statistics The excluded dummy corresponds to the last category of wood and furniture industries.
4 Length of sector targeting LST as instrument for FDI jYin columns 4 to 6 Another instrument variable, Sector targeted ST (not reported) was also used as instrument for FDI iY The results obtained were consistent to the ones reported.
5 The estimations were also run by combining all of the variables in columns 1–3 and 4–6 to check for robustness The results obtained (not reported) are consistent with the results in each column.
6 Please refer to Appendix A for the definition of each variable.