New Opportunities for Catching-up:
The Impact of Chinese FDI on Economic Growth in
Sub-Saharan Africa
Loïs M.B. de Haan
1University of Groningen
January 15, 2016
Supervisor: dr. T.M. Stelder
Abstract
As result of the recent increase in Chinese investment in Africa, this paper examines the impact of foreign direct investment from China on economic growth in a cluster of Sub-Saharan African countries. The empirical model includes generally accepted growth determinants and special attention is paid to the concept of absorptive capacity. In order to exploit the time-series nature of the relationships in the growth equation, the results are obtained with the use of panel data analysis over the period 2003 to 2012. The within estimator with both country and time fixed effects fails to find significant evidence for the role of Chinese investment as growth determinant of African real GDP per capita growth. Nonetheless, the benefits of South-South foreign direct investment should not be underestimated, especially for underdeveloped countries.
Keywords: Economic growth, technology diffusion, foreign direct investment, absorptive capacity,
panel data analysis.
JEL codes: O33, O47, O55, P33
1
Student of the University of Groningen, Faculty of Economics and Business, the Netherlands E-mail: l.m.b.de.haan@student.rug.nl
1. INTRODUCTION
Economic growth and its drivers have been at the focus of economic research for decades. It is widely accepted that productivity growth is the cornerstone of economic growth. Abramovitz (1986) states that productivity growth rates tend to vary inversely with productivity levels, resulting in convergence across countries. This is captured by the catch-up hypothesis which declares that being backward in level of productivity carries a potential for rapid advance (Abramovitz, 1986: 386). Developed countries are in possession of advanced technologies and, therefore, carry the role of “technological leader”. Accordingly, countries with lower levels of productivity can replace obsolete stock and exploit the technologies already employed by the leader. Low-income economies have relatively high potential to make a large leap because of the technological gap compared to the leader. This theory implies that developing countries, the “followers”, are fairly depended on the adoption of advanced technologies in order to economically grow. Subsequently, technology transfer is considered to be an explanation of convergences and thus, plays a central role in the process of economic development (Borensztein et al., 1998: p. 116).
Previous research has focused on a variety of channels through which new technologies and ideas are transferred from followers to leaders. The extent to which the benefits of technological progress spill across national frontiers is an interesting question. The focus of this paper is on technology diffusion through foreign direct investment (FDI). Technology diffusion is expected to boost long-run economic growth in the host country by means of technological upgrading and knowledge spillovers. The difficulty is that foreign technologies are not always suitable to the economic and social conditions of host economies. This is reflected by the inverse relation that would exist between a country’s distance to the technological frontier and its absorptive capacity. The absorptive capacity of the host economy, which has been measured by human capital and degree of openness in the bulk of studies, needs to be sufficient in order to adopt foreign, advanced technologies. Therefore, the effect of technology diffusion through FDI on economic growth will be studied in a model that includes the absorptive capacity of the host country (De Mello, 1997).
country. Research could support that investment policies should focus on attracting capital flows from comparable economies, as opposed to aim for learning from the best. Altogether, the objective of this paper is to contribute to the more limited research on the role of technology diffusion, through FDI as spillover channel, in the catch-up strategies of underdeveloped economies.
As indicated in Figure 1, FDI flows to developing economies have grown significantly and, now, constitute circa half of global flows. Among the developing economies, China has become the largest FDI host country. In the 1980s, China adopted an externally-oriented economic development strategy which included renewed focus on attracting FDI. The resulting increase in FDI into China has stimulated much growth in income that could almost surely have not been realized without these investments (Zhang, 2006). China has even been called a developing giant due to its remarkable high growth rate for over the past two decades. Today, China appears to be changing from recipient of FDI to global investor, with the country’s outbound investments growing at a higher rate than its inward FDI flows. China’s “going out” policy, that was introduced in 1996, focuses on securing natural resources and seeking markets for Chinese exports (Jenkins and Edwards, 2006).
China’s success story about FDI as a driver of economic growth makes an interesting case for Africa. Today, Sub-Saharan Africa is the most underdeveloped region of the world. Despite efforts of African governments, the record of attracting FDI has been disappointing over the past decades. Asiedu (2004) summarized Africa’s achievement with attracting FDI as an absolute progress but a relative decline compared to other developing countries. Still, findings point out that revised macroeconomic policies have improved the investment climate and FDI inflows have increased substantially in Africa. This is partially the result of the increasing Chinese interest in this continent. While China’s engagement in Africa is not a new phenomenon, the rapid growth in Chinese FDI flows towards Africa nourished the debate on their relation. FDI, trade and aid represent the key economic
Figure 1: Global FDI flows by destination from 1990 to 2014
0 1,000 2,000 3,000 4,000 5,000 6,000 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 In war d FD I fl o ws (i n US$ m ill io n s)
channels of interaction between China and African countries. Since 2000, China has emerged as Africa’s largest trading partner (Kolstad and Wiig, 2011). In terms of African destinations, Chinese overseas FDI flows everywhere, but with the focus on resource-rich countries (see Figure 2). In the reverse direction, i.e. FDI from Sub-Saharan Africa to China, FDI remains marginal (Pigato & Tang, 2015).
In light of potential channels for technology diffusion, this paper will focus on Chinese FDI in a cluster of Sub-Saharan African countries. The cluster of 21 countries has been selected based on the FDI flows from China to these countries and data availability. Figure 3 displays the net FDI flows from China into Sub-Saharan Africa over 2003 to 2012, which is the time span of this paper. Since bilateral flows in South Africa cause the peak in 2008, the development is also shown for the cluster excluding South Africa. Accordingly, a clear positive trend in FDI flows from China to Sub-Saharan Africa appears.
Figure 3: Chinese outward FDI into Sub-Saharan Africa from 2003 to 2012
Source: UNCTAD FDI/TNC database
Sub-Saharan Africa
excl. South Africa
Figure 2: China’s outward FDI stock in Africa, 2012
In the following section, economic theory on both exogenous and endogenous growth and the process of technology diffusion is outlined. In addition, the concept of absorptive capacity is introduced. Accordingly, the empirical framework is established in section 3. This section provides a discussion on the variables of interest and the expected relations between the dependent variable and the regressors. Section 4 elaborates on the sources that have been consulted to create a data set that could be used to test the growth equation. Subsequently, the results of the panel data analysis are interpreted in section 5. Finally, recommendations for future research are made and an overall conclusion is drawn at the end of this paper.
2. LITERATURE
2.1 Exogenous Growth Theory
In order to determine the approach to examine the relation between technology diffusion and economic growth, growth theories have to be taken into consideration. In the history of economics, several major approaches to economic growth have been developed. The basics of modern theories of economic growth have been founded by classical economists. In the classical approach, the three fundamental production factors include capital (K), labour (L) and natural resources (f.e. land). As proposed by classical economist Adam Smith, machinery was considered to be a complement to labour rather than a substitute for it. Smith viewed ‘new improvements of art’ as result of specialized economies. This leaves technological progress as an exogenous phenomenon and, thus, rather unexplained. David Ricardo, however, revised Smith’s assumption of constant returns to inputs by proposing the law of diminishing returns. Following Ricardo, an economic system will grow at a diminishing rate in the absence of technological progress (Kurz & Salvadori, 2003). Furthermore, Ricardo introduced the economic concept of comparative advantage, which has later been acknowledged as a fundamental explanation of gains from trade. Countries have a comparative advantage in commodities which they could produce relatively cheap. Therefore, countries with technology differences, for example in terms of labour productivity or industrial specialization, should focus on producing those goods in which they have comparative advantage and, subsequently, take part in international trade (Kerr, 2013).
growth model within the neoclassical framework. At the core of the exogenous growth model is an aggregate production function. Assuming labour-augmenting technological progress, the Cobb-Douglas production function is (where α is the cost share of capital):
𝑌(𝑡) = 𝐹(𝐾(𝑡), 𝐿(𝑡), 𝐴(𝑡)) = 𝐾(𝑡)𝛼(𝐴(𝑡)𝐿(𝑡))1−𝛼
which expresses the total output (Y) as a function of capital input (K),labour input (L) and total factor productivity, often referred to as technological progress (A). The Solow-Swan production function exhibits constant returns to scale, diminishing returns to inputs and a positive and smooth elasticity of substitution between the inputs. By substituting 𝐿(𝑡) ∙ 𝐴(𝑡) with the effective amount of labour 𝑁(𝑡), the fundamental equation of the neoclassical model, which is the equation for capital accumulation, is obtained:
𝑘̇ (𝑡) = 𝑠 ∙ 𝑓(𝑘̇(𝑡)) − (𝑛 + 𝑔 + 𝛿)𝑘̇(𝑡)
where all variables are expressed in terms of per unit of effective labor and the dot notation denotes a time derivative. L and A are assumed to grow exogenously at rate n and g, respectively, 𝛿 is the constant rate of depreciation and s is savings rate. This implies that in the steady state (𝑘̇ (𝑡) = 0), the stock of capital is determined by the savings rate, population growth rate, depreciation rate and technological progress according to (Dowrick & Rogers, 2002):
𝑘̇∗= ( 𝑠
𝑛 + 𝑔 + 𝛿)
1/(1−∝)
Following the assumption of constant returns, cross-country differences can be explained by the equation for steady-state per capita income:
ln (𝑌(𝑡) 𝐿(𝑡)) = ln 𝐴(0) + 𝑔𝑡 + 𝛼 1 − 𝛼ln(𝑠) − 𝛼 1 − 𝛼ln(𝑛 + 𝑔 + 𝛿)
where 𝐴(0) reflects the initial state of technology, but also resource endowments, climate and, for example institutions (Piras, 2010).
own steady state with a convergence speed that is inversely related to the distance from the steady state. Assuming that countries are all able to adopt non-rival, i.e. freely available, technologies, differences in country growth rates are explained by countries’ distances from steady state and diminishing returns to capital (Dowrick & Rogers, 2002).
2.2 Endogenous Growth Theory
While today’s mainstream economics is predominantly neoclassical, alternative approaches to growth have been developed because of dissatisfaction with the use of exogenous factors to explain long-run growth. In the mid-1980s, a new wave of scholars introduced an alternative approach to economic growth which is called endogenous growth theory. In contrast to the neoclassical assumption that technology would be non-rival, endogenous theories have focused on two aspects of technology: (1) technology is non-rival in the sense that the marginal costs for an additional agent to use it are negligible; and (2) the return to technical investments is partly private and partly public. The latter aspect would result in continuing incentives to innovate as market power is generated to a limited extent. More specifically, the expected value of the return to any innovation is increased by imitation. So, potential profits could outweigh the technology spillovers, that are embodied in the non-rival aspect of knowledge, through which outsiders could benefit from private R&D investment (Baro & Sala-i-Martin, 2003). This was the foundation for the investment-based growth model of Romer (1986). He proposed the so-called AK model, where the assumption of diminishing returns to capital was dropped and replaced with constant returns to scale. This model failed to predict both absolute and conditional convergence whereas each economy was expected to grow at the same per capita rate, regardless of its initial position (Barro & Sala-i-Martin, 2003).
In more comprehensive endogenous growth models the assumption of diminishing returns to capital was dropped by extending the concept of capital. Lucas (1988), for example, aimed to provide a model of economic development that would account for observed diversity across countries. Lucas added human capital, in terms of general skill level, to the Solow-Swan model:
𝑌(𝑡) = 𝐹(𝐾(𝑡), 𝑁𝑝(𝑡)), where 𝑁𝑝(𝑡) = 𝐻(𝑡)𝐿𝑝(𝑡)
where 𝑁𝑝(𝑡) is effective labour used in goods production and 𝐿𝑝(𝑡) is the labour employed in the
rates to be different across countries as result of differences that were not systematically related to income level (Barro & Sala-i-Martin, 2003).
The initial wave of endogenous growth theories by, among others, Romer (1986) and Lucas (1988) lacked the fundamental incorporation of technological change. In their models, endogenous growth could continue as result of returns to investment in (the broad concept of) capital. Endogenous growth models that include technology diffusion, suggest that growth rates in the short run depend on the dynamics of technological catch-up while, in the long run, growth rates converge across economies (Dowrick & Rogers, 2002). Grossman and Helpman (1991), for example, emphasized the importance of knowledge externalities in a R&D-based growth model. More specifically, they argued technological progress would be a result from deliberate R&D activity. Grossman and Helpman assumed production know-how to be private and improvement know-how to be public. Therefore, R&D activity would be rewarded by some form of ex post monopoly power and was assumed to be the underlying engine for positive long run growth. Romer (1990), on the other hand, proposed to model endogenous growth with a varieties structure, where technological progress appeared in the form of the generation of new ideas. Romer suggested that a range of country specific factors, such as R&D polies and market structure, could affect the long run growth rate of technology (Dowrick & Rogers, 2002). By assuming non-rival technology and thus, costless technology imitation or spillover, both Romer and Grossman & Helpman contributed to an explanation for the conditional convergence hypothesis. Again, this convergence would imply that less developed countries are assumed to potentially stand the most to gain from their international relationships by imitating and learning from the large stock of knowledge capital already accumulated in the industrialized world (Barro & Sala-i-Martin, 2003).
2.3 The Process of Technology Diffusion
After outlining the development of growth theory over the past two centuries, the next step includes linking growth theory to technology diffusion. Recall that long run economic growth is analysed by means of unexplained, exogenous technological progress in (neo)classical economics. More recent research on endogenous growth, on the other hand, provided determinants for long run growth within a model that endogenized technological progress. The endogenous growth theory has highlighted the dependence of growth rates on the state of domestic technology relative to that of the rest of the world (Borensztein et al., 1998). A country’s technical advances would not only contribute to its own productivity, but also drive growth elsewhere. Consequently, international technology spillovers have been a major source of technological progress for both developed and less developed countries (Lai et al, 2006).
economies. Continuing on the framework of endogenous growth, also including innovation, Barro and Sala-i-Martin (2003) suggested that the technological frontier would be equal to the number of varieties of intermediate products that has been discovered by the technological leader. Here, the honouring of intellectual property rights across international borders has ensured proper incentives for discoveries of new goods and techniques in leading economies. The innovations are initially used to produce final goods in the leading country, but are imitated by the following country subsequently. Associated with imitation are the costs of adaptation to a different environment. Following the assumption that it is cheaper to imitate others than to innovate, the argument for convergence again appears: the typical follower has the potential to grow relatively fast and catch up to the technological leaders. The followers’ growth path, which is an increasing function of the distance to its steady state, exhibits diminishing returns. This is because the costs of imitation increases as result of a decreasing pool of uncopied ideas. The growth rate of technological leaders, on the other hand, is assumed to be constant over time. Therefore, this approach to growth gives a model that considers the process of technology diffusion as the driver of convergence. Altogether, in Barro & Sala-i-Martin’s approach (2003) to technology diffusion, backwardness would not enhance the discovery or implementation of new technologies, but result in imitation of and catch-up to advanced economies.
Going back to the foundations laid down for the models of technology diffusion, Nelson and Phelps (1966) contributed significantly with their revised approach to the relationship between technology progress and human capital. In terms of technology transfers, Nelson and Phelps suggested that the rate at which the gap between the technology frontier and the level of current productivity of the follower is closed, depends on the level of human capital. They used the technology level of the leading country as a measure for the technology frontier and the level of education as a measure for human capital. Consequently, the level of education was assumed to drive technology diffusion and thus long run growth. Benhabib and Spiegel (2005) exploited the Nelson-Phelps model by acknowledging the facilitating role of education in the process of technology diffusion. They introduced an interaction term between human capital and the distance to the technology frontier as a measure for backwardness. Benhabib and Spiegel proposed countries may even experience divergence in total factor productivity if the initial human capital level is not sufficiently high to ensure catching-up to the leading country.
2.4 Absorptive Capacity
been a key concern in developing growth policies. Dowrick and Rogers (2002) found that the rates of technological progress vary significantly across economies due to both systematic catch-up and country-specific factors. For developing economies, the main obstacle to technological improvement, and thus economic development, has not been the access to technologies, but the weakness of domestic skills and competencies. In literature, the barrier to technology adoption has been referred to as the absorptive capacity of a country. This term is similar to the concept ‘social capability’ introduced by Abramovitz (1986). As mentioned, Borensztein et al. (1998) claimed that a threshold of human capital would be necessary to absorb the spillovers of foreign advanced technology. Correspondingly, Borensztein et al. measured the absorptive capacity by human capital accumulation and thus, viewed FDI and human capital as being complementary in the process of technological spillovers. Following the concept of absorptive capacity, the potential to imitate advanced technologies is subjected to the learning effect in backward economies. Human capital in terms of the level of education could have an essential impact on economic growth due to its role in creating domestic technology and in adopting and implementing the technologies from abroad.
Alternative indicators of absorptive capacity include domestic R&D resources, the degree of openness and the development of the financial system. In terms of domestic R&D resources, many researchers viewed domestic R&D investment as necessary complement to the adoption of foreign technologies. This implies that a country could not just imitate technological leaders, without also invest in domestic knowledge (Keller, 2004). Two effects come into play when it comes to the degree of openness to trade. Grossman and Helpman (1991) believed the degree of openness would make it more likely that a country would imitate and learn from technological leaders due to the increased exposure to foreign technologies. On the other hand, the higher a country’s degree of openness, the higher the competitive pressure from abroad. This pressure might push the backward country to adapt itself to the competition by increasing, for example, domestic R&D expenditures. Furthermore, Hermes and Lensink (2003) argued the importance of the development of the financial system of the host country as a precondition for technology diffusion through FDI. They stated that the country’s financial system would enhance the efficient allocation of resources and in this way, would improve the absorptive capacity. Taking countries in Sub-Saharan Africa into consideration, the lack of regional integration, underdeveloped institutions and poor infrastructure could also entail bottlenecks to adopting technologies from abroad successfully (Pigato & Tang, 2015).
concept of absorptive capacity, developing countries should not imitate the most advanced technologies, which would be hard to absorb, but the technologies that are suitable to the initial domestic technology level. In this way, barriers to technology adoption are less restrictive and the largest benefits could be shared. Accordingly, Gelb (2005) stated that South-South technology transfers embody technologies and business models that are often more appropriate to the host country context. He characterized Chinese investment into Africa as resource-seeking, including activities related to extractive industries such as mining or the oil industry and the search for cheap labour as a factor input. Extractive instudries account for 30% of total FDI from China. However, Chinese FDI in finance, construction and manufacturing has increased, which provides more potential for economic transformation in Sub-Saharan Africa. With China’s growth slowing, it can be expected that the focus of overseas investments partly shifts away from the extractive industries to more high-value added activities. Ethiopia, for example, succeed to attract substantial investment in labour-intensive industrusies, with China as one of the top three job creators in its manufacturing sector. These investments were accompanied by infrastructure enhancement and workforce development (Pigato & Tang, 2015).
2.5 Spillover Channel
intermediate goods. The trade flows between China and Africa follow the concept of comparative advantage. Over the years, African firms have not been able to position themselves within Chinese value chains. Consequently, the trade relation between China and Africa appears to have a limited impact on economic transformation (Pigato & Tang, 2015).
The second, arguably more important channel for technology diffusion is FDI. FDI can transfer international technology embodied in goods and services, but also in human capital. In other words, FDI conveys not only production know-how, but also intangible assets such as managerial skills. In this way, advanced knowledge created by technological leaders can be transferred to a developing country through ‘learning by watching’, training of labour and other links. Potentially, the imported skills will increase the marginal productivity of the host capital stock and thus, promote growth (Balasubramanyam et al., 1996). At the firm level, new technologies may spillover to firms in the host country in four different ways. First, foreign technology is demonstrated by foreign firms and, accordingly, imitated by local firms. Second, the entrance of foreign firms through FDI would bear competitive pressure for the local firms to adjust their activities and to introduce new technologies. Third, foreign linkage effects would arise, implying that local firms are forced to use more efficient technologies after purchasing from foreign suppliers. Finally, foreign firms with local joint venture partners may train local employees to enable them to work with the advanced technologies (Hermes & Lensink, 2003). On the other hand, taking FDI as a channel for technology diffusion also brings uncertainties. Multiple researchers questioned the extent to which FDI has indeed generated substantial technological externalities for domestic firms. Negative externalities could arise as result of barriers to access technology and raised competition. Considering the investments from China into Africa, case studies give also ambiguous results. Chinese investment is becoming increasingly diversified and is reaching almost all African countries. However, despite of the rapid increase in the volume of Chinese FDI, it accounted for only 7% of global FDI in Sub-Saharan Africa in 2013. Moreover, Chinese firms have been tended to rely on their own low-cost labour and have appeared to not invest heavily in the training and education of African workers (Zafar, 2007). Other Chinese investments have been associated with the establishment of local capacity, technology transfers and increases in exports to several African States. An example is Zimbabwe, one of the major tobacco exporters of the world, where Chinese investors have supported to process tobacco into cigarettes, followed by assistance with the export of the finished value-added products (Renard, 2011).
2. ECONOMETRIC FRAMEWORK
accelerator. In the traditional Solow-Swan model, the long run effect of FDI, as a source of capital accumulation, on the output level is constrained by the existence of diminishing returns to capital. New endogenous growth models, on the other hand, allow for the examination of the role of FDI in explaining economic growth. Adopting the framework of endogenous models, FDI is expected to alter the long run growth rate through both capital accumulation and externalities that accelerate the technological progress (De Mello, 1997).
The dependent variable of our model is economic growth, which can be proxied by the real GDP per capita growth rate (Hermes & Lensink, 2003). FDI from China to Sub-Saharan Africa is included as a factor input to examine its potential role as growth determinant. Subsequently, the aggregate production function that describes the real output level (Y) can be expressed as a function of labour (L), domestic physical and human capital stock (K and H, respectively), foreign capital stock (F) and t as an index for time, given by:
𝑌(𝑡) = (𝐴(𝑡)𝐿(𝑡)1−𝛼−𝛽−𝛾𝐾(𝑡)∝𝐻(𝑡)𝛽𝐹(𝑡)𝛾) ( 1)
Which can be transformed to per capita denotation: 𝑦(𝑡) = (𝐴(𝑡)𝑘̇(𝑡)∝ℎ(𝑡)𝛽𝑓(𝑡)𝛾) ( 2)
Assuming that the augmented Cobb-Douglas function is linear in logs and adding a vector of conditional variables that have been generally accepted as determinants of growth, the output per worker is expressed by:
𝑦𝑖𝑡 = 𝛽0+ 𝛽1∙ 𝑘̇𝑖𝑡+ 𝛽2∙ ℎ𝑖𝑡+ 𝛽3∙ 𝑓𝑖𝑡+ 𝛽4∙ 𝑍𝑖𝑡 ( 3)
where Z includes the conditioning set of growth determinants, i is an indicator for the Sub-Saharan African countries in the cluster, t denotes the time period. Starting from this expression for output per worker, we can work towards a model for the growth rate of real GDP per capita.
Due to limited availability of data on capital stock in developing economies, the growth rate of the capital stock is approximated by the share of domestic investment – measured by the gross capital formation – as a percentage of GDP. From economic theory, a positive relation is expected between the growth rate of the domestic capital stock and economic growth (Choe, 2003):
H1: Domestic capital stock growth is positively related to the rate of growth of a particular economy.
and Levine (1993), a higher (initial) level of human capital is expected to be associated with faster subsequent real GDP growth rates:
H2: The level of human capital is positively related to rate of growth of a particular economy.
Correspondingly, the growth rate of foreign capital stock is replaced by the share of foreign direct investment inflows as percentage of GDP (Balasubramanyam et al., 1996). Since the interest is in the FDI flows from China to the Sub-Saharan African countries, these inflows should be included separately while the model also controls for FDI from the rest of the world. Despite contradicting arguments, the ratio of inward FDI from China to output is expected to reflect productivity-enhancing externality effects as it can be characterized as South-South investment. Therefore, the model will be used to test the following hypothesis:
H3: Chinese FDI in Sub-Saharan Africa is positively related to the rate of growth in the host country.
Subsequently, the growth equation can be extended using the economic theory outlined at section 2. The impact of FDI on economic growth might be conditional upon characteristics of the host country in terms of its absorptive capacity. Different measures for absorptive capacity have been used in the past. Following the bulk of studies, interaction terms including human capital are used as an indicator for the absorptive capacity of the Sub-Saharan African countries. The higher the level of human capital in the host country, the higher the effect of FDI on the growth rate of the economy. FDI and human capital would be complementary with respect to increasing the rate of economic growth (Borensztein et al., 1998: 121). The coefficient on the interaction term is thus expected to be positive:
H4: The higher the level of human capital in the host country, the higher the effect of FDI on the growth rate of the economy.
As human capital as restriction on the successful adoption of technology is assumed to be less stringent for South-South investment, a control group is used to test this assertion. The aim is to compare the estimates of the model including the FDI flows from China into Sub-Saharan Africa with the estimates of an equivalent model including FDI flows from advanced countries into Sub-Saharan Africa. Accordingly, the following hypothesis can be tested:
The second interaction term follows from the Nelson and Phelps (1966) hypothesis that human capital would affect the rate of technological catch-up: the higher the level of human capital, the higher the rate of technological catch-up. Following Benhabib and Spiegel (2005), the model should include an interaction term between human capital and the distance to the technology frontier as a measure for backwardness (gap). More specifically, the relative backwardness can be measured by the distance of the real income per capita of the Sub-Saharan African cluster to the real income per capita in the U.S.A. as a proxy for the technology frontier (Li & Liu, 2005):
𝑔𝑎𝑝𝑖𝑡 =𝐺𝐷𝑃𝑈𝑆𝐴,𝑡− 𝐺𝐷𝑃𝑖𝑡 𝐺𝐷𝑃𝑖𝑡
Holding the initial level of human capital constant, countries with a lower initial productivity level will experience faster rates of growth of total factor productivity (Benhabib & Spiegel, 1994). The corresponding hypotheses are:
H5: A country’s distance to the technology frontier is inversely related to economic growth.
H6: The higher the level of human capital, the higher the effect of the technology gap on economic growth.
Since the logarithm of the initial level of real GDP per capita is also a broad accepted measure to reflect the process of catch-up, this measure is used as sensitivity measure. Without the control for human capital, being backward in level of productivity (i.e. indicated by a low initial level of real GDP per capita) carries a potential for rapid economic growth:
H7: The initial level of GDP is negatively related to the rate of growth of a particular economy.
Finally, a measure for openness to trade should be included as control variable. Keller (2002) has found that countries that have adopted relatively open trade regimes have often grown substantially faster than more closed countries. According to Asiedu (2004), the most widely used measure of openness is the share of the sum of total exports and import in GDP. This measure follows from the difficulty if one aims to distinguish the effects of imports and exports on output given the high collinearity between the two series.
Altogether, the model of technological catch-up that expresses the real growth rate as a function of its growth is:
𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝑡 = 𝛽0+ 𝑐𝑖+ 𝛽1∙ (𝐺𝐷𝑃𝐼 ) 𝑖𝑡+ 𝛽2∙ ℎ𝑖+ 𝛽3∙ ( 𝐹𝐷𝐼𝐶𝐻𝑁 𝐺𝐷𝑃 )𝑖𝑡 + 𝛽4∙ ( 𝐹𝐷𝐼𝑅𝑂𝑊 𝐺𝐷𝑃 )𝑖𝑡+ 𝛽5∙ ℎ𝑖∙ (𝐹𝐷𝐼𝐶𝐻𝑁 𝐺𝐷𝑃 )𝑖𝑡+ 𝛽6∙ ℎ𝑖∙ ( 𝐹𝐷𝐼𝑅𝑂𝑊 𝐺𝐷𝑃 )𝑖𝑡+ 𝛽7∙ 𝑔𝑎𝑝𝑖𝑡 + 𝛽8∙ ℎ𝑖∙ 𝑔𝑎𝑝𝑖𝑡+ 𝛽9∙ 𝐺𝐷𝑃𝑖+ 𝛽10∙ ( 𝑡𝑟𝑎𝑑𝑒 𝐺𝐷𝑃)𝑖𝑡+ 𝜀𝑖𝑡 (4)
where growth indicates the real GDP per capita growth rate, i is an indicator for the Sub-Saharan African countries in the cluster, t denotes the time period and 𝜀 is a white-noise disturbance term which varies both over time and country. Moreover, in order to control for country-specific differences in technology, production and socio-economic factors, a time-invariant unobserved individual effect (𝑐𝑖) is added to the model (Balasubramanyam et al., 1996). This is an endogenous growth model where FDI as a channel for technology diffusion, the level of human capital and measures for the technological gap are included as important determinants of the growth rate of low-income economies.
3. DATA
Finally, data on the FDI flows from the control group, which include the advanced G7 countries, have been based on the International Direct Investments Statistics of the OECD and statistics reported by the African Development Bank.
Data for the other explanatory variables are taken from reliable sources, including the World Development Indicators (2014) for data on the annual real GDP per capita growth rate and gross capital formation as % of GDP, and IMF’s World Economic Outlook Database (2015) for data on the initial level of real GDP per capita and trade as % of GDP. Finally, in order to measure human capital, data are limitedly available for the Sub-Saharan African countries in the sample. According to Barro and Lee (1994), the average years of secondary schooling among males (age: 25+) is the one most significantly correlated with economic growth. This measure of educational attainment is available for 21 countries in the sample. Since this measure has been widely accepted within growth research (Borensztein et al., 1998; Benhabib and Spiegel, 2005), the preference is given to use this measure and thus, limit the cluster. In conclusion, the sample of this research consists of 21 Sub-Saharan African countries with data over the short time series from 2003 up to 2012. An overview of the countries included in the cluster can be found in Appendix A.
4. RESULTS
The purpose of this paper is to estimate the effect of Chinese investments on economic growth of developing economies and to estimate the role of human capital – as part of the concept of absorptive capacity – in the process of technology diffusion. Since the paper’s time span only includes ten years we are dealing with short (micro) panels. The sample is both estimated as one single cross-section, using the data averages over 10 years of growth, and as a panel of three or four-year growth rates. In case of the latter, the data on the 21 countries are averaged over three sub-periods, resulting in a total of 63 observations. The estimations are based on a balanced – but small – data set, implying for each country in the cluster exactly one observation is available.
5.1 Pure Cross-Sectional Regressions
The relation between the real GDP per capita growth rate and included growth determinants is assessed with the use of a pure-cross-sectional, Ordinary Least Squares (OLS) analysis. The Breusch-Pagan test for heteroskedasticity fails to reject the null of homoscedasticity which rules out the need to use robust standard errors. The descriptive statistics are summarized in appendix B1 and the correlation matrix is provided in appendix C1. The Chinese FDI flows appear to be positively skewed. The correlation matrix shows that especially the log of the initial GDP per capita is only weakly related to the dependent variable (𝜌 = −0.0241).
added in two times: model 2 includes 𝐻 ∗ 𝐹𝐷𝐼𝑟𝑜𝑤 and 𝐻 ∗ 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎, and model 3 also includes gap and
𝐻 ∗ 𝑔𝑎𝑝. The estimation results indicate that gross capital formation as percentage of GDP is positively related to the real GDP growth rate at the 10% significance level. The estimates of the coefficient on
I/GDP vary from 0.141 to 0.147. This implies that an 1 unit increase in the domestic capital stock as
percentage of GDP results in at least an increase of 0.141 in the growth rate of real GDP per capita (%). Second, the regressions give a robust and positive coefficient on 𝐹𝐷𝐼𝑅𝑂𝑊 at the 10% significance level. This coefficient appears to be unstable ranging from 0.153 to 0.304. In model 3, for example, an increase of 1 unit in 𝐹𝐷𝐼𝑅𝑂𝑊 is associated with an increase of 0.304 in the growth rate of real GDP per capita. Accordingly, Wald tests have been performed after the regression of model 2 and 3 to check if the individual coefficients on 𝐹𝐷𝐼𝑅𝑂𝑊 are jointly significant with the interaction term 𝐻 ∗ 𝐹𝐷𝐼𝑟𝑜𝑤. These tests failed to reject the null hypothesis. Therefore, we are unable to interpret the marginal effect of 𝐹𝐷𝐼𝑅𝑂𝑊, possibly depending on the initial level of human capital, on the growth rate. Finally, the FDI inflows from China do not enter the growth regressions significantly, neither individually nor as part of an interaction term.
Table 1: Estimation Results of Pure Cross-Sectional Regressions based on data averaged over the
time period 2003-2012
Explanatory variables
(1) FDI China (2) FDI China (3) FDI China (4) FDI G7
Coeff. St. Err. Coeff. St. Err. Coeff. St. Err. Coeff. St. Err.
I/GDP 0.142** (0.0652) 0.147* (0.0696) 0.141* (0.0726) 0.136* (0.0665) Log(𝐺𝐷𝑃0) -0.262 (0.569) -0.308 (0.584) -1.068 (1.108) -1.910 (1.406) Log(𝐻0) -0.0137 (0.691) 0.998 (1.207) 0.710 (2.215) 1.471 (3.336) Trade/GDP -0.0168 (0.0221) -0.0191 (0.0236) -0.0246 (0.0250) -0.0360 (0.0264) 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 -1.669 (1.818) -2.476 (2.736) -0.191 (4.130) 𝐹𝐷𝐼𝑅𝑂𝑊 0.153* (0.0823) 0.291* (0.149) 0.304* (0.155) 0.310* (0.156) 𝐹𝐷𝐼𝐺7 0.473 (1.839) 𝐻 ∗ 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 0.789 (3.960) -0.263 (6.349) 𝐻 ∗ 𝐹𝐷𝐼𝑟𝑜𝑤 -0.186 (0.151) -0.181 (0.156) -0.187 (0.180) 𝐻 ∗ 𝐹𝐷𝐼𝐺7 -0.200 (2.562) gap -0.0283 (0.0244) -0.0437 (0.0244) 𝐻 ∗ 𝑔𝑎𝑝 -0.00363 (0.0283) -0.00601 (0.0244) Cons 2.198 (3.610) 1.927 (3.682) 8.865 (7.861) 16.34 (10.38) Observations 21 21 21 21 R2 0.373 0.446 0.511 0.548 Adjusted R2 0.104 0.0770 0.0214 0.0956
In Table 1, also a fourth model has been estimated. This model includes the FDI flows from the G7 countries into the Sub-Saharan African cluster instead of the FDI inflows from China. This model is added to assess potential differences in the role of human capital in the process of technology diffusion through South-South versus North-South FDI. The estimation results can be compared with the results corresponding to model 3. The coefficient on 𝐹𝐷𝐼𝑟𝑜𝑤 is again positive and significant at the
10% level. It should be noted that model 3 and 4 differ in terms of the countries that are included in the “rest of the world”. For some countries in the sample the inward flows from the advanced G7 countries comprise the bulk of their total FDI inflows. As this still implies uncertainty regarding what is included in the “rest of the world”, the two coefficients on 𝐹𝐷𝐼𝑅𝑂𝑊 given by regression 3 and 4 cannot
be compared. Furthermore, the coefficient on 𝐹𝐷𝐼𝐺7 does not enter the results significantly. Note that the sign and magnitude between 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 and 𝐹𝐷𝐼𝐺7 differ notably. 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 would be negatively related
to the economic growth of the host countries, whereas the 𝐹𝐷𝐼𝐺7 is associated with a positive
coefficient of 0.473. However, due to lack of significant evidence we are unable to make a statement about the difference in the relation between South-South FDI – proxied by the FDI inflows from China – and growth, and North-South FDI – proxied by the FDI inflows from the G7 countries – and economic growth in underdeveloped economies.
Finally, considering the goodness of fit measures for the different models, the R2 appears to increase when extending the model with interaction terms. However, it could be argued that the increasing R2 is the result of the increase in explanatory terms instead of better fit. This can be seen from the adjusted R2 which measures the explanatory power of the regressions while controlling for the number of predictors. This measure is actually the highest for the simple model that excludes the interaction terms of interest.
5.2 Regressions on Three Sub-Periods
An alternative to the pure cross-sectional regressions is the use of shorter time periods to bring out possible structural variations over time. This implies a move from cross-sectional to panel data as the data set now contains repeated observations on the cluster of 21 countries. As stated by Islam (1995), the smallest length to consider is yearly time spans. However, yearly growth rates are likely to be subjected to short-term disturbances. To cover structural variations in the growth rate, see Figure 4, and acquire periods that include approximately the same number of years, data are averaged over three time periods. The time periods are: 2003-2006, 2006-2009 and 2009-2013. This implies the data set contains 63 observations. 𝐿𝑜𝑔(𝐺𝐷𝑃0) is now measured by the real GDP per capita at the start of each
period (t=0 at year=2003, year=2006 and year=2009). Moreover, the level of human capital is equal to the Barro and Lee (1994) estimate for 2000 in period 1, for 2005 in period 2 and for 2010 in period 3.
span. Second, panel data models control for country-specific fixed effects, which are not included in the conditioning set of a pure cross-sectional regression (Carkovic and Levine, 2002). And third, panel data exploit the time-series nature of relationships of interest, which enables researchers to model individual dynamics (Verbeek, 2008).
The panel data have been examined with three different models. The dependent variable is the three or four year growth rate of real GDP per capita (%). The descriptive statistics and correlation matrix can be found in Appendix B2 and C2, respectively. Compared to the descriptive statistics corresponding to the data average over ten years, the skewness and kurtosis values increased significantly. Moreover, while almost all signs of the correlation coefficients remained the same, the magnitude substantially changed.
The first estimator is the pooled OLS regression. The Breusch-Pagan test for heteroskedasticity provides highly significant evidence that the null of homoskedasticity should be rejected. Therefore, heteroskedasticity robust standard errors are applied to all regressions. Overall, the standard errors are smaller compared to the pure cross-sectional regressions. The estimates of the pooled regression from the panel three to four-year span data are summarized in Table 2. As in the cross-sectional regressions, four models have been estimated.
The regressions indicate that the coefficient on I/GDP is positive and highly significant. The estimates vary from 0.205 (when the FDI inflows from China are replaced with the inflows from the G7 countries) to 0.227 (when the model excludes gap and 𝐻 ∗ 𝑔𝑎𝑝). These estimates confirm the expected relation between the domestic capital stock as percentage of GDP and economic growth proposed before. Furthermore, the three regressions that include the FDI inflows from China indicate that the FDI inflows from the rest of the world are positively related to the economic growth of the Sub-Saharan African countries. Considering the regressions that include the Chinese investments, the coefficients vary from 0.203 in the simple regression to 0.229 and are significant at the 5% level.
0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 GDP g ro wth ( % ), F D I (as % o f G D P)
Source: The World Bank Average growth (SSA)
FDI
As model 3 and 4 also include interaction terms with FDI and the level of human capital, calculating their marginal effects could provide a more reliable estimate of their effect on economic growth than the individual coefficients indicate. Therefore, after the regression of model 3, Wald tests are performed including 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 and 𝐻 ∗ 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 and 𝐹𝐷𝐼𝑅𝑂𝑊 and 𝐻 ∗ 𝐹𝐷𝐼𝑅𝑂𝑊, respectively. In contrast to the first, the latter indicates that the null hypothesis of both coefficients being equal to zero should be reject at the 5% significance level. Based on the estimation results of model 3, the marginal effect of 𝐹𝐷𝐼𝑅𝑂𝑊 is equal to:
𝜕𝑔𝑟𝑜𝑤𝑡ℎ
𝜕𝐹𝐷𝐼𝑟𝑜𝑤 = 𝛽𝐹𝐷𝐼𝑟𝑜𝑤+ 𝛽𝐻∗𝐹𝐷𝐼𝑟𝑜𝑤∗ 𝐿𝑜𝑔(𝐻0) = 0.229 − 0.0153 ∗ 𝐿𝑜𝑔(𝐻0)
The marginal effect shows that the effect of the FDI inflows from the world, excluding China, depends on the level of human capital in the host economy. Initially, the marginal effect appears to be positive stating that FDI is associated with positive external spillovers; i.e. evidence for technology transfers. However, other than expected, this effect declines with the initial level of human capital. If the average years of male secondary schooling increases with 10%, the marginal effect decreases with 0.0229 ∗
Table 2: Estimation Results of Pooled OLS regressions based on data averaged over three
sub-periods: 2003-2006, 2006-2009 and 2009-2012.
Explanatory variables
(1) FDI China (2) FDI China (3) FDI China (4) FDI G7
Coeff. St. Err. Coeff. St. Err. Coeff. St. Err. Coeff. St. Err.
I/GDP 0.216*** (0.0766) 0.227*** (0.0800) 0.219*** (0.0791) 0.205*** (0.0540) Log(𝐺𝐷𝑃0) -0.118 (0.419) -0.113 (0.421) -1.214 (1.020) 0.0150 (0.870) Log(𝐻0) -0.280 (0.610) -0.782 (0.757) 0.242 (1.680) -0.884 (1.596) Trade/GDP -0.0136 (0.0134) -0.0179 (0.0132) -0.0114 (0.0143) -0.00730 (0.0128) 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 0.0785 (1.903) -1.855 (1.826) -1.462 (1.773) 𝐹𝐷𝐼𝑅𝑂𝑊 0.203** (0.0881) 0.229** (0.0974) 0.229** (0.100) 0.224*** (0.0728) 𝐹𝐷𝐼𝐺7 -0.558 (1.389) 𝐻 ∗ 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 3.762* (2.240) 3.876* (2.304) 𝐻 ∗ 𝐹𝐷𝐼𝑟𝑜𝑤 -0.0140 (0.110) -0.0153 (0.106) -0.0126 (0.0978) 𝐻 ∗ 𝐹𝐷𝐼𝐺7 4.208** (1.575) gap -0.0173 (0.0187) 0.00390 (0.0159) 𝐻 ∗ 𝑔𝑎𝑝 -0.0109 (0.0153) 0.000581 (0.0138) Cons -0.778 (3.350) -0.542 (3.343) 7.490 (7.813) -2.112 (7.166) Observations 63 63 63 63 R2 0.272 0.316 0.341 0.427 Adjusted R2 0.194 0.214 0.214 0.317
log(1.10) = 0.00218. This change is rather small, suggesting that the role of the level of human capital could be ignored in the process of technology diffusion.
Furthermore, the fourth model gives a positive and significant coefficient on 𝐻 ∗ 𝐹𝐷𝐼𝐺7 of 4.208.
The Wald test on 𝐹𝐷𝐼𝐺7 and 𝐻 ∗ 𝐹𝐷𝐼𝐺7 indicates that the null of no joint significance should be
rejected. Therefore, the marginal effect for 𝐹𝐷𝐼𝐺7 is calculated by:
𝜕𝑔𝑟𝑜𝑤𝑡ℎ
𝜕𝐹𝐷𝐼𝐺7 = 𝛽𝐹𝐷𝐼𝐺7+ 𝛽𝐻∗𝐹𝐷𝐼𝐺7∗ 𝐿𝑜𝑔(𝐻0) = −0.558 + 4.208 ∗ 𝐿𝑜𝑔(𝐻0)
Equivalently, the Wald test on 𝐹𝐷𝐼𝑅𝑂𝑊 and 𝐻 ∗ 𝐹𝐷𝐼𝑅𝑂𝑊 also rejects the null hypothesis (P<0.01). The marginal effect based on the estimation results of regression (4) follows similar reasoning as was found by regression (3):
𝜕𝑔𝑟𝑜𝑤𝑡ℎ
𝜕𝐹𝐷𝐼𝑟𝑜𝑤 = 𝛽𝐹𝐷𝐼𝑟𝑜𝑤+ 𝛽𝐻∗𝐹𝐷𝐼𝑟𝑜𝑤∗ 𝐿𝑜𝑔(𝐻0) = 0.224 − 0.0126 ∗ 𝐿𝑜𝑔(𝐻0)
It can be noticed that the marginal effects of the two aggregate FDI flows suggest opposite relation with respect to the economic growth of the underdeveloped, African countries. Initially, the marginal effect of the FDI inflows originating from the G7 countries are negatively related to the real GDP per capita growth rate of the host economies. Accordingly, a 10% increase in the initial level of human capital will increase the marginal effect by 4.208 ∗ log(1.10) = 0.401. The marginal effect of the FDI inflows from the advanced economies will become positive if the average years of male secondary schooling is higher than 1.133 in the host economy. This would imply that, initially, host countries in Sub-Saharan Africa are faced with the negative externalities of FDI, likely due to barriers to technology adoption and increased competition. However, a significant level of skills in the host country would enhance the potential for spillovers of foreign advanced technologies. Considering 𝐹𝐷𝐼𝑅𝑂𝑊, human capital as indicator for the absorptive capacity of the host country, again, appears to
have a limited role.
A limitation to the Pooled OLS estimator is that the model does not control for individual heterogeneity. The Pooled OLS estimator assumes a constant and normally distributed error term and thus, ignores differences in country specifics like resource endowments and institutions. Due to the repeated observations on the 21 countries, inferences based on the OLS estimates will be misleading because the error term will differ across economies and could be correlated with the explanatory variables (Piras, 2010). A more suitable estimator that exploits the time dimension of panel data is the Fixed Effects estimator. The Fixed Effects model controls for the unobserved country-specific effect by including a specific intercept term for each country. Al together, the Fixed Effects estimator focuses on differences within the countries included in the cluster. Table 3 gives an overview of the estimation results of the Fixed Effects models.
Since the modified Wald test finds significant evidence for heteroskedasticity, white errors have been applied in all regressions. Moreover, time fixed effects have been added to control for year-specific shocks that are common across the cluster (Islam, 1995). Comparing the results of the Fixed Effects models with the pooled OLS regressions, the first interesting difference is the robust significance of the coefficients on the log of the initial level of per capita GDP. Whereas the pooled OLS regressions failed to find significant evidence for the relation between the initial level of GDP per
Table 3: Estimation Results of Fixed Effects models (with country and time fixed effects) based on
data averaged over three sub-periods: 2003-2006, 2006-2009 and 2009-2012.
Explanatory variables
(1) FDI China (2) FDI China (3) FDI China (4) FDI G7
Coeff. St. Err. Coeff. St. Err. Coeff. St. Err. Coeff. St. Err.
I/GDP 0.263 (0.168) 0.281 (0.165) -0.0886 (0.0915) 0.154 (0.0943) Log(𝐺𝐷𝑃0) -13.37** (5.489) -11.84** (4.557) -38.65*** (9.725) -26.04*** (8.931) Log(𝐻0) -1.636 (3.548) 0.789 (3.584) -8.918 (5.526) -5.962 (4.057) Trade/GDP -0.00876 (0.0469) -0.00790 (0.0381) -0.111* (0.0556) 0.0339 (0.0415) 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 0.402 (2.604) -1.075 (2.494) 0.157 (2.187) 𝐹𝐷𝐼𝑅𝑂𝑊 0.223 (0.158) 0.208 (0.126) -0.138* (0.0793) 0.0793 (0.0653) 𝐹𝐷𝐼𝐺7 0.984 (1.594) 𝐻 ∗ 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎 3.256 (2.777) 1.023 (2.789) 𝐻 ∗ 𝐹𝐷𝐼𝑟𝑜𝑤 0.0632 (0.146) 0.0674 (0.125) 0.110 (0.0929) 𝐻 ∗ 𝐹𝐷𝐼𝐺7 1.516 (1.685) gap -0.398*** (0.0885) -0.236** (0.0995) 𝐻 ∗ 𝑔𝑎𝑝 0.00435 (0.0423) 0.0242 (0.0337) Cons 85.54** (34.29) 74.85** (28.59) 284.7*** (69.11) 188.9*** (65.04) Observations 63 63 63 63 R2 0.636 0.668 0.813 0.823 Adjusted R2 0.337 0.357 0.614 0.634 rho 0.967 0.953 0.994 0.988
capita and the growth rate, the Fixed Effects model indicates the two are negatively related. The coefficient is, however, unstable as it varies from -11.84 (in model 2) to 38.65 (in model 3). Acknowledging the fit of model 3, an 1% decrease in the initial level of GDP per capita will increase the growth rate, which is expressed in percentages, by 0.3865%. This is in accordance with the catch-up hypothesis: the log of the initial level of GDP per capita, as proxy for the technological gap, is inversely related with a country’s growth rate.
Furthermore, the regressions on the full models give a negative and significant coefficient on gap and on Trade/GDP. As the latter is a proxy for the openness of the host economy, the finding of a negative coefficient is in contrast with the proposed hypothesis (H7). As an increase in openness is likely to result in increased competitive pressure from abroad, which could push firms in a country to adapt itself to the competition. Firms in an underdeveloped economy, however, might not be able to adapt and thus, compete with foreign firms. This could be an explanation for the negative effect of trade on a country’s economic growth. In regard to the significant coefficients on gap, a Wald test indicates we should assess the relation between the measure for backwardness and economic growth by means of its marginal effect. On the basis of the regression that includes the Chinese FDI inflows (model 3), the marginal effect is given by:
𝜕𝑔𝑟𝑜𝑤𝑡ℎ
𝛿𝑔𝑎𝑝 = 𝛽𝑔𝑎𝑝+ 𝛽𝐻∗𝑔𝑎𝑝∗ 𝐿𝑜𝑔(𝐻0) = −0.398 + 0.00435 ∗ 𝐿𝑜𝑔(𝐻0)
Holding the initial level of human capital constant, the marginal effect states that the level of backwardness is negatively related to economic growth. From economic theory, however, we would expect that an increase in the technological gap would imply an increase in the per capita growth rate. Moreover, the influence of the level of human capital on the rate of technological catch-up appears to be very small. If the average years of secondary schooling increases with 10%, the marginal effect increases with: 0. 00435 ∙ ln(1.10) = 0.000414.
conclusion the Fixed Effects model is more efficient compared to the Pooled OLS estimator, especially when estimating the full model (which is the aim of the paper)
The third and final estimator that could be used to evaluate the data is the Random Effects model. This model assumes that the error term of the growth equation is uncorrelated with the explanatory variables. In order to test if the Random Effects model is preferred over the Fixed Effects model, a test of overidentifying restrictions (i.e. orthogonality conditions) has been carried out. This test gives a Sargan-Hansen statistic of 23.418 with a P-value of 0.0093. This implies the null hypothesis stating that both the fixed and random effects estimator are efficient should be rejected at the 5% significance interval. In terms of overidentifying restrictions, rejection of the null hypothesis suggests that the additional orthogonality conditions proposed by the random effects estimator – the independent variables are uncorrelated with the group-specific error term – are not valid as instruments. Additionally, the Breusch-Pagan Lagrange Multiplier (LM) test is performed to test for random effects. All post-estimations fail to reject the null stating that variances across the Sub-Saharan African countries are zero. In conclusion, the random effects estimator is assumed to be inconsistent, implying that the estimation results of the Fixed Effects model are preferred for the evaluation of the relations proposed by the empirical model. The Random Effects model could be inconsistent because it treats the individual effects as random and uncorrelated with the other explanatory variables. The estimation results obtained using Random Effects models are summarized in Table 4 in Appendix D.
5. DISCUSSION
The empirical work on cross-country growth differences has been closely related to developments in growth theories. From the conditional convergence hypothesis, we assumed that low-income countries could achieve economic growth by means of, among others, external spillovers from FDI. An empirical model has been established to analyze whether FDI flows from China, the developing country that was successful in attracting and utilizing inward FDI, can be considered as growth determinant for underdeveloped economies in Sub-Saharan Africa. Unfortunately, the estimation results obtained by means of panel data analysis did not provide significant coefficients on the relation between Chinese investments in Africa and the real GDP per capita growth rate of the host economies. The pooled OLS regressions did find a positive and significant coefficient on the interaction term including the level of human capital and 𝐹𝐷𝐼𝐶ℎ𝑖𝑛𝑎, which suggests the first plays a role in the adoption
The estimation results obtained by means of the Fixed Effects model, however, do not give a robust and positive relation between inward FDI and economic growth. In contrast, this estimator emphasizes the relation between the initial GDP per capita and the future growth rates. Whereas no significant relation on this explanatory variable was found in the OLS regressions, a negative coefficient is provided in the Fixed Effects model. This would correspond to assuming that a lower level of initial GDP per capita results in an increase in the growth rate, and is in accordance with the catch-up hypothesis. Additionally, the proxy for distance to the technology frontier (gap) was also associated with a highly significant coefficient in the Fixed Effects model. However, for similar levels of human capital, an increase in backwardness would result in lower economic growth. This finding is thus, in contrast with the finding on the log of the initial level of GDP per capita.
Comparing the estimation results from the pooled OLS regressions with the Fixed Effects models indicates another interesting difference. The magnitude and significance of the intercept increased substantially when moving from pooled OLS to Fixed Effects estimation. As noted, the pooled OLS regression does not control for unobserved heterogeneity and thus, treats country-specific effects as part of the error term. However, if the unobserved individual effect is correlated with explanatory variables and a determinant of the dependent variable, OLS estimates will suffer from omitted variable bias. The Fixed Effects model, on the other hand, does control for country characteristics. The considerable change in the magnitude and significance of the coefficients (including the intercept) gives rise to the importance of unobserved factors that do influence economic growth of the underdeveloped economies. We should note that Africa is a continent of differences, where political stability and relevant institutions have been missing in many countries (Pigato & Tang, 2015). In addition, economic growth is a comprehensive subject, where previous research has considered a variety of variables as growth determinants. For example, factors that have been argued to explain economic growth, which have been excluded from the empirical model, include the level of the financial system, government consumption, inflation (as a proxy for macroeconomic risk), black market premium, a measure of political instability and a measure of quality of institutions. The risk of omitted relevant variables is spurious causality.
emphasized that the increased Chinese interest in investing in Sub-Saharan Africa is a recent trend. The global economic crisis in 2008/2009 has been marked as the beginning of the rapid expansion in China’s engagement with Sub-Saharan Africa in terms of investment (Pigato & Tang, 2015). The impact of the Chinese FDI flows on economic growth might become feasible in several years. Therefore, future research could deliver more valuable insights regarding the relation between FDI and economic growth in developing economies due to more reliable and extensive data.
A final limitation is the potential issue of endogeneity, which has been acknowledged by several researchers when examining the role of FDI in determining economic growth. Blomstrom et al (1994) and Choe (2003), for example, found evidence that FDI Granger-causes economic growth. A potential solution to the issue of endogeneity is the use of the Generalized Method of Moments estimator that has been proposed by Arellano and Bond (1991). This dynamic panel model allows for the lagged dependent variable to be included as regressor. As this estimator is designed for situations with a small number of time periods and a dynamic dependent variable that depends on past realizations and fixed individual effects, it could provide a valuable alternative to the static panel estimators used in this paper. On the downside are the difficulty in finding valid instruments and the understanding of statistical tools required for obtaining estimation results.
Presently, despite the lack of significant evidence found in this paper, real life examples do support the belief in beneficial FDI externalities among developing economies. African countries have been willing to cooperate with China over European or American partners, as Chinese have appeared to be less averse to risk (Renard, 2011). Historically, Chinese FDI has been focused on the infrastructure sector in Africa. One recent project offers a great example of the way technology can be transferred from China to Africa. At the beginning of September 2014, Chinese Bridge and Road started the construction of the Standard Gauge Railway in Kenya. The railway will connect Kenya, Uganda, Rwanda and South Sudan and is financed by Chinese funds for 85%. The project is a cooperation between Chinese workers and locals. In August 2015, over 25,000 locals were hired to build the railway, together with circa 2,000 Chinese workers. Knowledge is transferred by training more than 15,000 skilled workers and 400 engineers and technicians. Moreover, for construction material has been demanded from local manufacturers. The establishment of a future railway is even under consideration to continue the training of key Kenyan staff in railway construction, operation and maintenance. Recently, the Chinese Ambassador to Kenya announced that the project is running successfully and that its progress is ahead of schedule.2
On the other hand, Chinese investment faced challenges that were similar to challenges Western investors had to overcome. In the context of Africa, poor infrastructure, insufficient institutions and political instability are examples of potential factors that influence the success of technology diffusion.
2
“Kenya: Standard Gauge Railway construction ahead of schedule,”
This implies that future research on the absorptive capacity of Sub-Saharan Africa might not only include the level of initial human capital, but could also consider other social economic conditions as constrains to FDI and growth in general. Al together, the increase in FDI flows among developing economies bears new opportunities for low-income economies to adopt technologies that are appropriate to their specific conditions.
6. CONCLUSIONS
While a close relationship between China and Sub-Saharan Africa is not a new phenomenon, the recent increase in Chinese FDI flows to African countries has raised new interest in their relation. Foreign investors could be a growing force in the economic transformation of Sub-Saharan Africa as these economies are depended on sources from abroad to achieve technological progress. The increase in investments from developing economies raises new perspectives on the potential for growth and thus, catch-up for undeveloped economies. While the results found in this paper could not underpin the relative benefits of Chinese investments over FDI from industrialized economies, the first do provide an alternative to African dependence on Western countries. African countries will be faced with an increase in options to attract investments, which could potentially increase their bargaining power. Subsequent research should be conducted to investigate the importance of technology transfers as important driver of economic growth in African economies, with minimal restriction imposed by their absorptive capacity. While FDI from China is still small relative to total flows into Africa, its rapid growth and shifting focus towards manufacturing raises questions about what the future will bring.
REFERENCES Abramovitz, M. (1986). Catching up, forging
ahead, and falling behind. The Journal of
Economic History, 46(02), 385-406.
Adams, S. (2009). Foreign direct investment, domestic investment, and economic growth in Sub-Saharan Africa. Journal of Policy
Modeling, 31(6), 939-949.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The review of economic studies,
58(2), 277-297.
Asiedu, E. (2004). Policy reform and foreign direct investment in Africa: Absolute progress but relative decline. Development Policy
Review, 22(1), 41-48.
Balasubramanyam, V. N., Salisu, M., & Sapsford, D. (1996). Foreign direct investment and growth in EP and IS countries. The
economic journal, 92-105.
Barro, R. J., & Lee, J. W. (1994, June). Sources of economic growth. In
Carnegie-Rochester conference series on public policy
(Vol. 40, pp. 1-46). North-Holland. Barro, R.J. and Sala-i-Martin, X. (2003).
Economic Growth, Second Edition.
Cambridge: The MIT Press
Benhabib, J., & Spiegel, M. M. (1994). The role of human capital in economic
development evidence from aggregate cross-country data. Journal of Monetary economics,
