The Contribution of FDI to Structural Change

(1)

MASTER THESIS UNIVERSITY OF GRONINGEN - INTERNATIONAL ECONOMICS AND BUSINESS

The Contribution of FDI to

Structural Change

The case of Sub-Saharan Africa

Evy P.M. de Groot

30-6-2014

Supervisor: A.A. Erumban Co-Assessor: R.C. Inklaar

Abstract:

This thesis studies the impact of growth in foreign direct investment inflows on structural transformation in Sub-Saharan Africa based on the Africa Sector Database developed by the GGDC. The results suggest that FDI is an important factor in the transition from low-productivity to high-productivity sectors. However, the contribution of FDI only holds when the host country has a minimum threshold of financial development, human capital, and institutional quality.

(2)

(3)

3

List of Figures

3.1. Sub-Saharan Africa’s GDP per capita is US$ ... 14

3.2. Sub-Saharan Africa’s FDI inflows as a % of GDP ... 14

3.3. Inward FDI flows in Sub-Saharan Africa as a % of GDP; Resource Rich vs Resource Poor Countries ... 15

3.4. Average Labor Productivity per Sector in Sub-Saharan Africa ... 17

3.5. Correlation between Change in Sectoral Productivity relative to Total Productivity and Change in Employment Shares in Sub-Saharan Africa ... 138

List of Tables

4.1. Decomposition of Productivity Growth ... 21

4.2. Correlation Matrix ... 29

5.1. Basic Model ... 39

5.2. Determinants of the Structural Change Term ... 33

(5)

5

1. Introduction

Structural change – the shift of resources from low productivity to high productivity uses – is considered as a key driver of economic growth (Kuznets, 1966; McMillan and Rodrik, 2011; Page, 2012; De Vries, Erumban, Timmer, Boskoboynikov, and Wu, 2012). The changes in structure during this development process are characterized by a transition from a low income agrarian economy to an industrial urban economy with higher income. The countries that have managed to pull out of poverty are those that have diversified away from agriculture and other traditional products and moved labor and other resources into modern economic activities (McMillan and Rodrik, 2011). In developing economies value added and output per worker generally differ significantly across sectors as compared to more developed economies. These large productivity gaps are to some extent indicative of inefficiencies in the allocation of resources that reduce overall labor productivity, and can therefore be used to identify opportunities for development (Chenery and Syrquin, 1989). Positive structural change takes place when labor and other resources are allocated more efficiently and move from less to more productive activities, allowing the economy to attain higher output and productivity growth even without productivity growth within sectors (McMillan and Rodrik, 2011). Such transformation of the structure of production has been considered as essential in the process of development for decades already (Chenery, 1982). In their cross-country analysis of structural change and economic growth, McMillan and Rodrik (2011) conclude, inter alia, that Sub-Saharan African countries have experienced growth-reducing structural change from 1990 to 2005. However, in a more recent study McMillan, Rodrik, and Verduzco-Gallo (2013) observe that most African countries in their sample have experienced expansions in the manufacturing sector since 2000 and consequently structural change has positively contributed to overall productivity growth.

Even though structural change is considered to be a major driver for developing economies to achieve faster productivity and economic growth, factors that influence structural change are hardly examined in the literature. McMillan et al. (2013) attribute most of the structural change patterns to local circumstances such as the level of employment rigidity and currency undervaluation. Nevertheless, an external factor that could play an important role in driving structural change is foreign direct investment (FDI), as FDI inflows in Africa have increased significantly as of the same year that structural change became positive for the region. Therefore, this study examines the impact of FDI on structural change and the influences of host country conditions on the effect of FDI on structural change.

(6)

6 countries, and for these economies FDI inflows are argued to have a larger growth enhancing effect than domestic investment (Borensztein, De Gregorio, and Lee, 1998). Hansen and Rand (2005) conclude that there is no threshold for the level of development of the host country to observe the positive relation between FDI and growth. Two main channels via which FDI contributes to economic growth in developing countries are technology spillovers and physical capital inflows (Johnson, 2006). FDI allows technology transfers that cannot be achieved by pure financial investments or trade in goods and services and contributes to human capital development through employee training and a positive influence on managerial skills and know-how (Alfaro et al., 2004; Loungani and Razin, 2001). Further, FDI encourages competition and the introduction of new processes in domestic markets resulting in productivity gains (Zilinske, 2010), and it makes international production networks and international markets more accessible for the host country (Alfaro et al., 2000). FDI may also help to limit the implementation of bad policies of the host government (Ciburiene and Zaharieva, 2006). In general, most governments are willing to attract foreign investments as a means for productive capacity building and sustainable development (UNCTAD World Investment Report, 2013)1.

Whereas Sub-Saharan Africa (SSA) has shown instable and even negative GDP per capita growth rates for a long period of time, GDP per capita has been positive and on a steady rise since 20012. The upsurge of FDI inflows to SSA may have contributed to this rise in GDP per capita. Africa experienced year-on-year growth of FDI inflows since 2000. Although this growth pattern was partly driven by FDI in extractive industries, investment in manufacturing and service industries has been expanding as well (UNCTAD World Investment Report, 2013). BRICS countries in particular are becoming significant investors in Africa and are now ranked among the top investing countries in this region. Especially Brazilian and Chinese FDI to Africa has risen in recent years, entailing positive growth effects for the continent (UNCTAD World Investment Report, 2013). Since structural change has been found to be positive and thus contributive to the economic growth of SSA after 2000, it is interesting to see whether the increasing FDI inflows played a role in enhancing structural change in Africa. As outlined previously, FDI inflows to developing countries enable technology transfers, human capital development, productivity gains, and limit the implementation of bad policies. All of these factors make it easier for the host economy to move towards more modern activities. Via this channel, FDI might positively influence structural change. Moreover, as FDI can create opportunities in more productive sectors, movement of workers and resources to such productive segments of the economy can contribute to positive structural change. The fact that FDI is rising in sectors other than the natural resource industry is indicative of expansion of productive sectors which might promote

1

However, investors face many difficulties and risks when entering a developing market like coping with the foreign legal system, corruption, underdeveloped infrastructure and political instability (Asiedu, 2006).

(7)

7 growth enhancing structural change. Because of these reasons, the main question addressed in this thesis is whether FDI inflows to SSA have positively contributed to structural change in this region. This thesis is based on the theories and methods of McMillan and Rodrik (2011). Their research however, following the existing literature on the topic of structural change patterns (Lin, 2010), concentrates on different host country factors that help determine whether structural change contributes to overall productivity growth. This thesis makes a contribution to the existing literature by looking into the effect of different host country conditions along with the effect of an external factor - the recent upswing of FDI inflows - on structural change patterns in SSA.

The next section will provide an overview of existing research on the topics of structural change and FDI and growth. The economic development of SSA with a focus on structural change patterns and FDI in this region is discussed in section 3. Section 4 will provide the data and methods used for analysis. Section 5 contains the empirical results with a discussion of these results and robustness checks for some of the variables. The conclusions, limitations and suggestions for future research are presented in section 6.

2. Literature Review

Given the theoretical importance of structural change and FDI for economic growth, especially for developing economies, structural change patterns and the impact of FDI inflows on economic performance have been researched extensively. Since the present thesis focuses on the impact of FDI on structural change, the following subsection provides an overview of the most relevant literature on the topic of structural change, followed by a subsection devoted to relevant literature about the effect of FDI on economic growth.

2.1 Structural Change and Economic Growth

(8)

8 activities favors aggregate productivity performance, implying that structural change causes aggregate productivity growth.

Given the importance of structural change for economic progress and productivity growth in the process of development, many studies have examined the quantitative magnitude of the impact of structural change. The central idea that structural change is an important factor in the development process is based on classical dual economy models (Lewis, 1954). In an economy with unlimited labor supply, economic growth is accomplished by capital accumulation in more sophisticated industries, which in turn attracts the excessive labor from agricultural industries. Whereas this stream of literature is based on a two-component analysis, more recent studies have examined the topic by taking more detailed factors into account. Baily, Hulten and Campbell (1992) made an important contribution to the literature by developing a method to measure aggregate productivity levels, in which growth in productivity levels of individual firms and relative efficiency of firms were differentiated from each other. A further extension of this decomposition resulted in the most commonly used decomposition method nowadays (Isaksson, 2010), developed by Foster, Haltiwanger and Krizan (2001), introducing additional terms capturing simultaneous change in firm output share and entry and exit of firms. Using this decomposition over the period from the early 1990’s until the early 2000’s, it has been found that the effect of productivity growth within sectors dominates the structural change term independent of the stage of development of the economy (Foster et al, 2001).

(9)

9 additional support by an examination of structural change in BRIC countries, finding that China, India, and Russia have experienced positive aggregate productivity growth through reallocation of labor across sectors, whereas in Brazil the opposite has been the case (de Vries et al., 2012).

Factors enabling structural change to contribute to overall productivity growth and distinguishing Asia from Latin America and SSA are a relatively small share of natural resources in exports, the management ability to cope with downsides from import competition, competitive exchange rates and a higher level of labor market flexibility (McMillan and Rodrik, 2011). McMillan and Rodrik (2011) base these findings on a standard shift-share analysis, where aggregate labor productivity is decomposed into weighted sectoral productivity growth shares and a structural change component, weighed by end-of-period employment shares and beginning-of-period productivity levels. Dabla-Norris, Ho, Kochhar, Kyobe, and Tchaidze (2013) base their research on the same decomposition method and find additional evidence for the results of McMillan and Rodrik (2011). They distinguish between six regions which are either advanced, emerging, or low-income, and show that the contribution of structural change to aggregate productivity growth levels differs significantly across regions. Although labor productivity growth in emerging market and low-income economies between 1990 and 2008 has mainly been driven by within-sector productivity growth3 and reallocation of labor across sectors has been productivity-reducing for Latin America, SSA, and the Middle East and Northern Africa, structural change has been an additional driver of aggregate productivity gains for fast growing economies like emerging Asia. An important additional finding for SSA is that however, as expected, labor productivity in nonagricultural sectors is significantly higher than in agriculture, labor productivity in services is on average higher than labor productivity in manufacturing sectors in this region. This is in contrary to earlier findings that services generally show lower average productivity than manufacturing (Duarte and Restuccia, 2010) and indicates that labor allocation from agricultural to service sectors is potentially contributive to structural change in SSA as well. McMillan et al. (2013) also make use of the same decomposition method. They however find a turning point for SSA in 2000, after when structural change has been growth enhancing. A more stable macroeconomic and political environment combined with increasing agricultural productivity and rising global food and commodity prices made investors more willing to invest in agribusiness in the region. Rising labor costs in China also make Africa a more attractive location for labor intensive industries. The augmented attractiveness combined with the recent spread of democracy in the continent and increased bargaining power of African governments due to the global search for

3_{The decomposition is applied to agriculture, manufacturing, mining, construction, trade, transport and}

(10)

10 natural resources, makes it more likely that resources will be used more efficiently and enhance positive structural change (McMillan et al., 2013).

Using the recent and carefully constructed Africa Sector Database, de Vries, Timmer, and de Vries (2013) find that labor has been moving from lower productivity to higher productivity sectors in SSA over the period 1960 to 2010. A possible explanation for these positive results compared to previous negative findings lies in the inclusion of both formal and informal employment in the Africa Sector Database, whereas the UNIDO data used by previous researchers exclusively comprise formal employment. Nevertheless, the findings of de Vries et al. (2013) illustrate that sectors that expanded in terms of employment shares experienced negative productivity growth. The former effect is due to static reallocation contributing to aggregate productivity growth, whereas the latter, called the dynamic reallocation effect, measures productivity growth of employment absorbing sectors. Because of the negative contribution of the dynamic reallocation effect, the total reallocation effect in Africa is only small between 1990 and 2010.

(11)

11 Mello, 1999). After the Second World War, production financed by FDI was dominated by the United States because of their large technological capacity and a world shortage of US dollars, and foreign investments were made in a very limited amount of destination countries (Dunning, 1979). From the late 1960’s onwards, there has been a diversification of both origin and destination countries of foreign investments. Influenced by the importance of global value chains in today’s global economy and the wide reach of MNEs that enable integration of worldwide production through horizontal and vertical linkages (Buckley and Ghauri, 2004), worldwide FDI flows have increased particularly fast since the 1990’s (UNCTAD World Investment Report, 2013).

Although debates about the causality between FDI and economic growth exist (Choe, 2003), an extensive amount of literature finds FDI inflows to be positive for the host country and FDI has been more resilient over time than other forms of private capital during past decades. Especially developing countries might therefore prefer FDI over other forms of capital flows (Loungani and Razin, 2001).

(12)

12 important for capital accumulation, is associated with the development of new technologies, and is more important than domestic investment.

For developing countries in particular, previous work on the relationship between FDI and growth has shown that FDI has a positive impact on economic growth (Chowdhury and Mavrotas, 2006). Tsai (1994) found that, for the lowest developed countries, FDI is contributive to long-term economic growth through inter alia technology and knowledge spillovers. These spillovers are important because they contribute to higher productivity levels of capital and labor and effective allocation of these resources in the host country (Hermes and Lensink, 2003), nurturing movement of resources from low productive to high productive sectors. FDI spillovers can also contribute to labor productivity growth in low productive sectors like agriculture, influencing structural change positively. Because of the higher expected labor productivity and the flow of resources towards higher productivity sectors through FDI inflows, the first hypothesis is:

Hypothesis 1: FDI inflows have a positive effect on structural change in Sub-Saharan Africa.

However, for poor developing countries it appears difficult to derive the potential macroeconomic benefits from FDI (Nunnenkamp, 2004). They may experience difficulties in making efficient use of new technologies embodied in FDI-related capital accumulation or in assimilating capital and technology-intensive improvements (de Mello, 1999). In order for FDI to be contributive to economic growth, physical and capital resources should be allocated effectively so that the potential technological diffusion and associated productivity growth can take place. This will in turn lead to expansion of modern sectors and consequently to structural change and economic growth. The strength of the relationship between FDI, structural change, and growth might therefore be affected by institutions and be subject to a number of host country conditions.

(13)

13 level of human capital, well developed financial systems, and a sound institutional environment, and have no significant impact on structural change by themselves.

Hypotheses 2: The effect of FDI inflows on structural change in Sub-Saharan Africa depends on

host-country conditions like human capital, financial development, and institutional quality.

Before any of the potential benefits can be derived however, developing countries should in the first place be able to attract FDI flows. Outward FDI is likely when the benefits of own foreign production exceed those of inter-firm agreements (UNIDO, 2003). Following the eclectic paradigm explaining the patterns of foreign value added activities of firms in a globalizing economy (Dunning, 2000), foreign production and investment undertaken by MNEs is determined by the interaction of three interdependent, context specific and dynamic factors; ownership specific advantages, locational attractiveness of countries, and the benefits of internalization of cross-border intermediate product markets. To examine how developing countries can attract more FDI inflows, locational attractiveness is the main variable of interest. Foreign investors generally tend to attribute host country attractiveness to political stability, economic stability, international outlook, government regulations and attitude, and factors related to infrastructure, labor, and banking and finance (UNIDO, 2003). Bartels, Kratzch, and Eicher (2008) analyzed foreign investors and their reasons to invest in SSA countries4, and found that the factors ‘Political Economy of Investment Climate’ and ‘Legal Environment of Governance’ are the most important motivations. This finding is in line with the IMF Regional Economic Outlook on Sub-Saharan Africa of Aril 2014, suggesting that investors discriminate on the soundness of policy environment. Foreign investors thus mainly base their choice to invest in a particular location on political and economic stability, providing a sound investment climate and a transparent legal framework.

3. Economic Growth, FDI and Industrial Structure in SSA

After a period of economic downturn, SSA has experienced a threefold of GDP per capita since 2001 (figure 3.1). Economic growth has led to higher living standards, poverty reduction, and improved social indicators. Next to sound economic policies and stronger institutions, risen foreign investments in several industries under which mining, infrastructure for transport and communication, and energy production have been major contributors to growth (IMF Regional Economic Outlook, 2014).

4

(14)

14 Aseidu (2006) and Anyanwu (2012) both examine which host country conditions are necessary for SSA to attract FDI inflows by making use of cross-country regressions. They find that next to political stability and an efficient legal system, factors which were already found important for developing countries in general, openness to FDI and macroeconomic stability also play a significant role in promoting FDI inflows. Other determinants positively affecting the amount of FDI flows towards countries in the SSA region are availability of natural resources, the size of the market, good infrastructure, low levels of corruption, and higher levels of agglomeration (Aseidu, 2006; Anyanwu, 2012).

As presented in figure 3.2, overall FDI inflows to SSA as a percentage of GDP have increased significantly during the past two decades. Although FDI inflows remain partly driven by investments in resource extractive industries, recently there has been a rapid increase of FDI in consumer-oriented manufacturing and services, accounting for 23 percent of the total inflows in 20125. A number of factors, including the growing population and rising middle class, raise the prospect of an increasingly dynamic consumer market and might be playing a role in the diversification of the industrial composition of FDI inflows. Moreover, a population share of 25-year-olds or younger accounting for more than half of the residents, together with increasing urbanization, make the Sub-Saharan African demographic composition favorable for increasing investment opportunities in consumer oriented industries (UNCTAD World Investment Report, 2013). Figure 3.3 reveals that FDI flows, as a percentage of host country GDP, towards SSA countries with a lack of natural resources have over the period 1990-2005 on average been lower than FDI flows towards resource-rich economies, but have exceeded the latter during 2005-20106.

5

Data obtained from UNCTAD World Investment Report, 2013.

6_{Resource-intensive countries include Angola, Botswana, Cameroon, Central African Republic, Chad,}

Democratic Republic of Congo, Republic of Congo, Equatorial Guinea, Gabon, Ghana, Guinea, Mali, Namibia, Figure 3.1. Sub-Saharan Africa’s GDP per capita in US$

GDP in current prices and current exchange rates

Figure 3.2. Sub-Saharan Africa’s FDI inflows as a % of GDP GDP in current prices and current exchange rates

(15)

15 An important share of the increasing FDI inflows to SSA comes from the rising investments from emerging and developing markets in the region. Among these economies, Malaysia is the main investor in Africa with investments across all sectors, including significant investments in agribusiness and finance. Next to Malaysia the BRICS countries are the largest developing-country investors in Africa. Each of the BRICS countries has its own investment strategy in SSA. Brazilian FDI to Africa has been led by public financial institutions like the Brazilian Development Bank, and mainly flows into the new ethanol industry in countries like Angola, Ghana and Mozambique. The UNCTAD World Investment Report (2013) mentions that Russian investment in SSA has only expanded very recently and has been focused primarily on raw-material supplies and easing the accessibility of local markets. Whereas Indian FDI has mainly been concentrated in Mauritius because of historical links, investment in other countries such as Côte d’Ivoire, Ethiopia, Senegal and Sudan have nowadays become important as well. Over the past decade, China has built a network of trade, aid, and investment links with about 50 African countries and therefore Chinese FDI in SSA has attracted much attention (Zafar, 2007). Sudan, Nigeria, and Zambia have been the leading recipients of Chinese FDI, which did not only focus on extractive industries, but also on manufacturing and service sector in these countries. In contrary to the Western view (Kaplinsky, McCormick, and Morris, 2007), most African countries have a positive attitude towards China’s investments in the resource sector since they also give out loans for infrastructure, which mostly can be repaid with natural resources, or do not have to be repaid at all (Sautman and Hairong, 2007). In many African countries China has already contributed to the construction of railways, dams, ports, office buildings, hospitals, etc., which are necessary prerequisites for development (Sautman and Hairong, 2007; Zafar, 2007). Finally, South Africa’s FDI in Africa, attributable to reinvested earnings in the private non-banking sector, is particularly concentrated in Mauritius, Nigeria, Mozambique, and Zimbabwe. Overall, the share of BRICS in FDI inflows in Africa reached 25 percent of total inflows in 2013, with only 26 percent of the value of their projects in primary and resource extractive sectors

Niger, Nigeria, Sierra Leone, South Africa, Tanzania, Zambia, and Zimbabwe (IMF Regional Economic Outlook of Sub-Saharan Africa, 2014).

Figure 3.3. Inward FDI flows in SSA as a % of GDP; Resource Rich vs Resource Poor Countries GDP in current prices and current exchange rates

(16)

16 and the remainder mainly in manufacturing and services projects (UNCTAD World Investment Report, 2013).

Besides developing and emerging countries, developed countries also play a significant role in FDI flows to SSA. These flows have been at a high in the years before 2007, but tightened due to the global financial crisis. The decline in FDI outflows from developed countries accounted for almost all of the decline in global outflows in 2012 (UNCTAD World Investment Report, 2013). Nevertheless, Africa is an attractive investment location for developed economies compared with other developing destinations. Whereas returns on United States’ FDI were 20 percent in Africa in 2010, they were only 14 percent and 15 percent in Latin America and Asia, respectively (UNCTAD World Investment Report, 2012). Africa’s growth prospects are expected to be positively influenced by mature economies through their stabilizing economic situation and improved economic outlook (The World Bank, Africa’s Pulse Team, 2013).

Another feature of Africa’s recent economic growth is the changing structure of the economy. From the Africa Sector Database (de Vries, Timmer, and de Vries, 2013) it follows that personal services, government services, and agriculture are the sectors with the lowest labor productivity relative to total labor productivity in the eleven countries included in the database, whereas mining, utilities, and business services are the sectors with the highest relative labor productivity7. As shown in figure 3.4, the overall tendency of this distribution has remained stable over time, with an upsurge of productivity in the construction industry, now belonging to the top three most productive sectors. The general trend over the period 1990-2000 has been that agriculture, mining, manufacturing and utilities experienced a loss in employment share, while trade services, business services, and government services have gained the largest employment shares. Over the period 2000-2010 however, employment shares moving out of the low productivity agricultural sector doubled compared to the previous period, while employment shares in high productivity sectors like utilities, transport services, and business services increased significantly. Not only has labor increasingly flown towards higher productivity sectors over time, but as follows from the sectors that gained employment shares combined with figure 3.4, they have increasingly flown towards sectors that experienced high gains in productivity levels. This is presented in figure 3.5 by the changes in relative sectoral labor productivity and changes in employment shares. The steeper slope of the linear trend line over the period 2000-2010 compared to 1990-2000 indicates a structural change trend positive for economic development of the region.

(17)

17 Figure 3.4. Average Labor Productivity per Sector in Sub-Saharan Africa

Labor productivity measured at constant 2005 prices (thousands)

It is now evident that economic growth in Africa has been improving in the recent years, along with increasing inflows of FDI and changing economic structure. This strengthens the importance of understanding whether increasing FDI promotes structural change in Africa. The improved allocation of foreign investments together with increasing interest of foreign investors to invest in manufacturing and other high productivity sectors are both factors that have most probably positively influenced the contribution of FDI inflows to structural change. The next section will outline the model and date used to empirically test whether the growth in FDI inflows are indeed positively related to structural change and whether specific host country absorptive capacity conditions influenced this relation.

(18)

18 2000-2010

Source: Africa Sector Database (de Vries et al., 2013) Note: Abbreviations are as follows: (Agr) Agriculture; (Min) Mining; (Man) Manufacturing; (Utt) Utilities; (Con) Construction; (Trd) Trade services; (Trs) Transport services; (Bus) Business services; (Gov) Government services; (Psn) Personal services.

Figure 3.5. Correlation between Change in Sectoral Productivity relative to Total Productivity and Change in Employment Shares in Sub-Saharan Africa

(19)

19

4. Model and Data

4.1 Model

The following model considers different factors that are expected to have an effect on structural change, and will be used in order to determine what the impact of FDI on structural change is in SSA:

where denotes the county and denotes time. FDI is the main variable of interest, and is further specified in section 4.3.1. X is a vector of control variables that can influence structural change, including the variables identified by McMillan and Rodrik (2011) and variables denoting absorptive capacity as derived from the existing literature. IT represents the interaction terms between FDI and selected control variables, in order to test whether the control variables are possible prerequisites for FDI to be positively related to structural change.

The relationship between FDI and structural change is analyzed using data on eleven Sub-Saharan African countries included in the Africa Sector Database (de Vries et al., 2013), which are: Botswana, Ethiopia, Ghana, Kenya, Mauritius, Malawi, Nigeria, Senegal, South Africa, United Republic of Tanzania, and Zambia. The Africa Sector Database (de Vries et al., 2013) provides recent and reliable long-run data on these countries for value added and persons employed in the ten main sectors of the economy. The time span used is from 1990 to 2010 with five-year intervals, because both FDI inflows and structural change terms have changed considerably during this time period. A last note about the model is that Kenya and Zambia have been excluded from the regressions. Whereas Kenya shows, as opposed to the other countries in the sample, significantly lower structural change after 2000 as before and a declining GDP per capita between 1990 and 2005, Zambia’s economic growth has stagnated from 2000 to 2005 in terms of GDP per capita, but also in terms of institutional and financial development. These trends are not in line with the general tendency among the other countries and negatively affect the regression results, leading to the decision to omit both of the countries from the model. In what follows the expected relationship between structural change and various indicators which are used in the model are explained, along with the interaction terms examining host-country conditions which might influence the effect of FDI on structural change.

4.2 Dependent Variable

4.2.1 Structural Change

The dependent variable SC is the structural change term, derived by making use of the Africa Sector Database (de Vries et al., 2013). Whereas previous studies on resource reallocation have mainly been

(20)

20 focused on OECD and other developed regions because of a lack of data on developing economies8, this recently developed database contains large-scale sectoral data on output and productivity trends in Africa. Based on the method of McMillan and Rodrik (2011), change in labor productivity is decomposed as:

∑ _{)) )} ∑ _{)) )}

where refers to economy-wide labor productivity levels and refers to sectoral labor

productivity levels. is the share of employment in sector at time , and represents the change

in productivity or employment shares between and . As already stated before, productivity can grow within economic sectors through capital accumulation, technological change, or reduction of misallocation across plants, which is denoted by the first term of the decomposition. The second term of the decomposition is the term of interest, namely the structural change term. This term captures the movement of labor across sectors and is positive when changes in employment shares are positively correlated with productivity levels, meaning that labor moves from low-productivity sectors to high-productivity sectors. In this way structural change increases economy-wide productivity growth.

The method used to calculate the structural change in equation (2) differs slightly from the technique used by McMillan and Rodrik (2011). Instead of using labor productivities at t-k to calculate the within term and sectoral employment shares at t to calculate the between term, here the averages of labor productivity and employment share over the considered time period are used. In this way, overestimation of the contribution of one of the terms is hampered (Timmer and de Vries, 2009). The different decomposition method together with making use of another dataset seems to affect the results. Whereas McMillan and Rodrik (2011) show that the average structural change in SSA has been negative in the period before 2000, making use of the ASD gives positive structural change terms for each of the four time periods (1990-1995, 1995-2000, 2000-2005, and 2005-2010) taken into account. The value of the term doubles from 1990-1995 to 1995-2000, increases by almost 20% in the next period, and drops slightly in the last period. Table 4.1 provides the aggregate values of both decomposition terms and labor productivity growth for the countries included in the model over the examined time periods. The same pattern holds when the structural change terms are calculated by means of the decomposition method used by McMillan and Rodrik (2011) and it can thus be concluded that the differences found are due to the use of different databases. McMillan and

8_{Although there are exceptions, for instance the examination of structural change in BRIC countries by de}

Vries et al. (2012), analysis on Africa is still very limited.

(21)

21

Period Change in Labor

Productivity Within Between 1990-1995 0.74% 0.19% 0.55% 1995-2000 2.08% 1.20% 0.88% 2000-2005 2.66% 1.37% 1.29% 2005-2010 2.93% 1.95% 0.97% Total 8.41% 4.72% 3.69%

Table 4.1. Decomposition of Productivity Growth in Sub-Saharan Africa (in percentages)

Source: Africa Sector Database (de Vries et al., 2013)

both decomposition terms and labor productivity growth for the countries included in the model over the examined time periods. The same pattern holds when the structural change terms are calculated by means of the decomposition method used by McMillan and Rodrik (2011) and it can thus be concluded that the differences found are due to the use of different databases. McMillan and Rodrik (2011) mainly make use of the 10-Sector Productivity Database from the Groningen Growth and Development Centre supplemented with data derived from national accounts and comparable data sources for the African countries in their sample. The Africa Sector Database (de Vries et al., 2013) however provides more accurate and reliable data on the African output and productivity trends because of the construction method which ensures intertemporal, international, and internal consistency9. Individual structural change patterns for all countries and aggregate structural change terms across the four time periods are presented in appendix B.

4.3 Independent Variables

4.3.1 Foreign Direct Investment

The first independent variable, FDI, in model (1) is the main variable of interest in the analysis. As hypothesized in the literature review, FDI inflows are expected to have a positive effect on structural change through technology diffusion and the generation of opportunities in high productivity sectors, nurturing movement of labor and other resources from low productive to high productive sectors. The existence and magnitude of this effect may however depend on specific characteristics of the host environment.

FDI is measured by the average growth rate of real FDI inflows over the time periods 1990-1995, 1995-2000, 2000-2005 and 2005-2010. Data on FDI inflows, measured in US Dollars at current prices and current exchange rates, are defined as “an investment involving a long-term relationship and

reflecting a lasting interest in and control by a resident entity in one economy (foreign direct investor or parent enterprise) of an enterprise resident in a different economy (FDI enterprise or affiliate

9_{More information on the general and detailed sources and methods used to construct the Africa Sector}

(22)

22

enterprise or foreign affiliate), involving both the initial transaction between the two entities and all subsequent transactions between them and among foreign affiliates” (UNCTAD Statistics, 2014), and

obtained from UNCTAD Statistics on Foreign Direct Investment Flows. To obtain real FDI inflows, these data are deflated using the UNCTAD Statistics Consumer Price Index. On average, growth in FDI inflows to the countries in the sample was positive for 1990-1995, 1995-2000, and 2000-2005, but turned negative for 2005-2010. Growth in FDI inflows per country are presented in appendix C. 4.3.2 Control Variables

In this section, host country characteristics that are likely to affect structural change and the relationship between FDI inflows and structural change are presented. The host-country characteristics are derived from existing literature on factors driving structural change and factors necessary for low developed countries to grow economically or to be able to derive positive influences from FDI inflows. The variables identified by McMillan and Rodrik (2011) are taken into account, which include the share of employment in the agriculture sector, also called initial structural gap (ISG), raw materials share in exports (RawMat) and currency undervaluation (Underval). McMillan and Rodrik (2011) also include ‘employment rigidity’, which is omitted from this analysis due to lack of data on rigidity of employment in SSA over the examined time periods. The independent variables are supplemented with a set of indicators that can influence a country’s absorptive capacity of FDI inflows. Countries with a better developed financial and institutional environment and higher levels of human capital are likely to allocate capital inflows more effectively and to promote modern, higher productivity sectors. Therefore financial development (FinDev), human capital (HumCap), and institutional quality represented by rule of the law (RuleOfLaw) are included in the model. The impact of openness to trade, measured as the share of trade in GDP, has been examined as well, but is excluded from the model because it was found not to have any significant impact on the dependent variable. A possible explanation of this insignificant relation is that the openness of an economy in terms of either trade or financial openness might not explain the effectiveness of resource allocation in the economy.

4.3.2.1 Initial Structural Gap

(23)

23 investment opportunities and the benefits of absorbing foreign technologies (Dabla-Norris et al., 2013). Therefore the size of the initial structural gap is expected to have a positive impact on structural change.

The size of the initial structural gap may also impact the relationship between structural change and FDI. The expectation is that a higher share of FDI flows into low productivity sectors when a larger share of employment is active in low productive industries, making these sectors more attractive and capable to employ workforce, which in turn will prevent the movement of labor to higher productivity sectors (McMillan and Rodrik, 2011). On the other hand, higher FDI inflows to low productivity sectors can increase the productivity of these sectors, resulting in a lower demand for labor for the same amount of output. It is however expected that the surplus labor cannot be reallocated towards modern sectors because of the low development and absorptive capacity of those industries (Redding, 1996; Nict-Chenaf and Rogier, 2009). Consequently, it can be argued that the size of the initial structural gap is negatively related to the effect of FDI on structural change. To capture this effect, an interaction term of FDI and ISG (FDI*ISG) is included in the model.

ISG is measured by the employment share in agriculture plus the employment share in sectors with a lower labor productivity than agriculture at the beginning of the time period, as derived from the Africa Sector Database (de Vries et al., 2013). The size of the initial structural gap differs significantly across the countries in the sample. Whereas Botswana, Mauritius, and South Africa show initial gaps of less than 40%, Ethiopia, Malawi, Nigeria, and Tanzania show initial gaps of 80% and higher.

4.3.2.2 Raw Materials

In countries with a large share of export in raw materials, globalization discourages diversification towards modern manufacturing activities and strengthens traditional specialization patters. Thereby, extractive industries have, although they are often highly productive, a very limited capacity to employ workforce (McMillan and Rodrik, 2011). The share of raw materials in a country’s exports is therefore expected to reduce the pace of structural change.

(24)

24 Therefore, it is hard to claim that the share of raw materials in a country’s exports would be related to the effect of FDI on structural change.

The share of a country’s exports that is accounted for by raw materials, as represented by RawMat, is measured as the share of exports of inedible crude materials (excluding fuels)10 and mineral fuels, lubricants and related materials11 in the total amount of exports, using data from UNCTAD Statistics Merchandise Trade Matrix. Because data are only available from 1995 onwards, the share of exports accounted for by raw materials at the end of each period is included. Since the values remain quite stable over time, this is not expected to lead to biased results. The two countries with the most extreme values for this indicator are Mauritius with an average of 0.8% of its exports accounted for by raw materials, and Nigeria with an average of 95% of its exports accounted for by raw materials. The reason for this is that, as follows from the data, for SSA in general a very large amount of the share of raw materials in exports is accounted for by mineral fuels, lubricants and related products, and only a small amount is accounted for by inedible crude materials. Countries rich in natural resources like Nigeria thus have a considerable higher share of raw materials in their exports.

4.3.2.3 Currency Undervaluation

Another independent variable used by McMillan and Rodrik (2011) is a measure of the undervaluation of a country’s currency and presented as Underval. Undervaluation increases the relative wealth of foreign firms and leads to capital inflows (Anyanwu, 2012). Overvaluation damages in particular the modern industries because it tightens tradable industries (McMillan and Rodrik, 2011). Undervaluation is therefore expected to affect modern industries positively, which can in turn generate more employment opportunities. Countries with an undervalued currency are thus likely to experience labor movement from traditional sectors to modern sectors, having a positive impact on structural change.

When modern industries are squeezed due to an overvalued currency, it also seems less likely for foreign investors to invest in these industries. Since undervaluation on the other hand promotes both capital inflows and modern industries, currency undervaluation is expected to influence the effect of FDI on structural change positively. For this reason the interaction term FDI*Underval is included in the model.

10

This segment includes Hides, skins and fur-skins, raw; oil seeds and oleaginous fruits; crude rubber; cork and wood; pulp and waste paper; textiles fibers and their wastes; crude fertilizers and crude minerals; metalliferous ores and metal scrap; crude animal and vegetable materials.

11_{This segment includes mineral coal, coke and briquettes; petroleum products and related materials; gas,}

(25)

25 The currency valuations are derived from Rodrik (2008)12. These data are only available until 2005, and therefore the valuations for the period 2005-2010 are estimated by assuming a stable growth trend over time, which generally holds for the previous periods. Rodrik (2008) measures undervaluation as the difference between the real exchange rate, calculated with yearly exchange rates and purchasing power parity conversion factors, and the Balassa-Samuelson-adjusted exchange rate13_{. For positive currency valuations the currency is undervalued, while for values below zero the}

currency is overvalued. Botswana, Nigeria, and Tanzania have overvalued currencies for all time periods, while Mauritius and South Africa have undervalued currencies for all time periods. For all other countries the variable fluctuates around zero.

4.3.2.4 Financial Development

Efficient financial systems reduce risks of investment projects, allocate resources better, monitor and screen investment projects, mobilize savings and facilitate the exchange of goods and services (Levine, 1997). The development of financial systems will also determine to what extent domestic firms can realize their investment plans and whether foreign firms will be able to borrow in order to carry out their innovative activities in the host economy (Hermes and Lensink, 2000). Especially the development of financial systems allocating more credit to private firms is important for productivity growth and economic development (King and Levine, 1993), which in turn contribute to growth enhancing structural change. It is therefore expected that financial development has a positive effect on structural change.

Because of the risk reducing effect of financial development on investment projects, investors are expected to be more willing to invest in high-risk projects when financial development is higher. Riskier investments are generally related to more modern industries. Also, better developed financial systems allocate resources more efficiently, implying that financial development is important for a country to be able to benefit from FDI spillovers and that it promotes economic growth (Levine, 1997; Wurgler, 2000). Alfaro et al. (2004) and Hermes and Lensink (2000) find that countries with higher levels of financial development gain more from FDI, as their financial systems enhance the process of technological diffusion. Following this line of argument, it is expected that financial development influences to the effect of FDI inflows on structural change positively.

Financial development, FinDev, is measured as domestic credit to the private sector as a percentage of GDP, referring to “financial resources provided to the private sector by financial corporations, such

12_{Database available at http:// http://www.hks.harvard.edu.} 13

(26)

26

as through loans, purchases of non-equity securities, and trade credits and other accounts receivable, that establish a claim for repayment” (World DataBank, 2014), using data from World DataBank

World Development Indicators. This measure is based on earlier findings that financial systems which allocate more credit to private firms are more important for productivity growth and economic development than other financial systems (King and Levine, 1993; Levine, 1997; Hermes and Lensink, 2000; Anyanwu, 2012). In our sample, Mauritius and South Africa have a relatively high level of financial development, while Ghana has a relatively low level of financial development.

4.3.2.5 Human Capital

The relationship between education and growth is analyzed by two different theoretical frameworks. In one line of literature, growth is the outcome of human capital accumulation (Lucas, 1988). However, a threshold of human capital may exist below which human capital accumulation cannot promote growth because externalities fail to materialize below this threshold (Nicet-Chenaf and Rougier, 2009). Nevertheless, a rapid increase in human capital in an economy with a low degree of modern sector development makes that the availability of skilled labor can exceed the demand for skilled labor. Under these circumstances, human capital gets inappropriately allocated across sectors, which hampers productivity enhancement and growth (Redding, 1996; London et al., 2008). In developing countries it has been widely observed that high skilled labor is employed in sectors with low skill requirements. Véganzonès-Varoudakis (2005) examined the relation between human capital and growth in Middle East and North African countries between 1960 and 2000, finding that human capital in these countries has been misallocated over this time period, diverting employment from growth-enhancing activities. In a simple two-sector model of a small open economy, Nicet-Chenaf and Rougier (2009) have shown that the effect of education on growth is more significant if the country has already entered into structural change, raising the demand for skilled labor. Until the sectoral structure of production generates sufficient demand for skilled labor, misallocation of employment is expected to occur, resulting in decreasing returns to human capital (Ventura, 1997; Nelson and Pack, 1999). Although SSA experienced positive aggregate structural change from 1990 to 2010, with an employment share of more than 50 percent in the agricultural sector, modern sectors are still in a very low stage of development14. Therefore, it is expected that the increased share of educated labor force in SSA is misallocated across sectors, having no or even a negative effect on structural change in the region.

The second approach concentrates on the explicit relationship between human capital and technological change, in which human capital positively affects the ability to innovate or absorb

(27)

27 technologies from developed economies (Nelson and Phelps, 1966). Spillovers and technology transfers are not a direct consequence from FDI but depend on the absorptive capability of the country to which the investments flow. Borensztein et al. (1998) argue that in developing countries, more advanced technologies require higher levels of human capital since the absorptive capacity of a country is largely determined by the level of human capital. A higher educated labor force is expected to allocate capital inflows more efficiently. Zhang (2001) also finds that for developing countries, improved education and thus higher levels of human capital is a necessary precondition for a positive influence of FDI on economic growth. Another way in which human capital can affect the impact of FDI on resource allocation is suggested by Redding (1996) and Nelson and Pack (1999), who conclude that an increase in skilled labor makes investors more willing to invest in modern sectors rather than traditional industries. FDI is thus expected to contribute to structural change when the absorptive capacity determined by human capital is relatively high. To account for this indirect effect, an interaction term of FDI and human capital (FDI*HumCap) is included in model (1). Human capital is measured as the initial gross enrolment ratio in secondary school of both sexes as a percentage of the total population (Alfaro et al., 2004) and obtained from UNESCO Institute for Statistics, data on education. However tertiary school enrolment ratios or the share of educated persons in the labor force would have been more preferable measures of human capital, these data are not available for all countries in the sample over 1990-2010. Two alternative measures for human capital are included in the robustness section and have not been found to lead to different results. For all countries in our sample, human capital increased significantly from 1990 to 2010 with an average increase of almost 40%15.

4.3.2.6 Institutional Quality

“Third World countries are poor because the institutional constraints define a set of payoffs to political/economic activity that do not encourage productive activity” (North, 1990). A better

developed institutional framework is likely to push an economy towards more modern activities and is therefore an important contributor to economic growth and structural change. Several different measures of institutional quality can be found in the literature on economic growth and the relationship between FDI and growth, including indicators of property rights enforcement, regulation, and effectiveness of rule of law (Durham, 2004; Asiedu, 2006; Anyanwu, 2012). These indicators are positively related to growth and make FDI inflows more effective. Corruption on the other hand is expected to depress growth, productivity, structural change, and FDI inflows, and has been observed to be particularly important for Africa (Asiedu, 2006). In the remainder of this thesis,

(28)

28 institutional quality is proxied by rule of law16 (RuleOfLaw), capturing “perceptions of the extent to

which agents have confidence in and accept the rules of society, with in particular the quality of contact enforcement, property rights, the police and the courts, as well as the likelihood of crime and violence” (World DataBank, 2014). In particular contract enforcement and property rights

enforcement are important for the development of modern, high productivity industries. Rule of law is therefore expected to have a positive effect on structural change.

It can be argued that an efficient legal system will make investors more willing to invest in more risky activities that are usually present in higher productivity sectors. Following Durham (2004), it is additionally expected that managers allocate FDI more effectively given a sound institutional environment. As a result a better developed institutional environment is predicted to enhance the ability of a country to absorb the potential spillover effects from FDI inflows. This implies that rule of law influences the effect of FDI inflows on structural change in a positive way, and to capture this indirect effect an interaction term between FDI and rule of law is included in model (1).

The data on rule of law are obtained from the World DataBank World Development Indicators. The variable is a percentile rank indicating a country’s rule of law quality amongst all countries covered by the aggregate indicator, where 0 represents the lowest rank and 100 the highest rank. In the country sample, Nigeria has the lowest rule of law rank of 12%, whereas Mauritius has the highest rank for this variable with an average of 78.9%.

4.3.2.7 Correlation of the Independent Variables

In Table 4.2, the correlation matrix for the independent variables is presented. As shown, the employment share in low productivity sectors is highly and negatively correlated with financial development, human capital, and rule of law. This is not that surprising since a large initial structural gap is generally an indicator for low economic development, and low developed countries are expected to have a lower level of financial development, human capital, and institutional quality. To test whether the high correlation affects the results provided in the next section, the regressions have also been conducted excluding the indicator for initial structural gap. This does not result in any substantial differences, and it can thus be concluded that the high correlation between the previous mentioned independent variables does not lead to biased results. The high correlation between financial development and human capital is controlled for in the same way. When human capital is excluded from the model, the effect of financial development on structural change maintains

16_{Other institutional variables discussed in this section are tried as well, but were not found to have a}

(29)

29 essentially the same. The high correlation between human capital and financial development are therefore also not expected to influence any of the results negatively.

Table 4.2. Correlation matrix

ISG RawMat Underval FDI FinDev HumCap RuleOfLaw

ISG 1.000 RawMat 0.3032 1.000 Underval -0.2984 -0.3343 1.000 FDI 0.0757 0.0214 0.0296 1.000 FinDev -0.7602 -0.1169 0.3213 0.0158 1.000 HumCap -0.8846 -0.2445 0.3711 -0.1243 0.7589 1.000 RuleOfLaw -0.7035 -0.6091 0.3864 -0.0473 0.3578 0.5955 1.000

5. Results and Discussion

5.1 Basic Model with variables in McMillan and Rodrik

At first model (1) is estimated only including the

(30)

30 expectations, currency undervaluation does not seem to have a substantial effect on structural change. Although the variable enters the regression with a positive coefficient, the effect is not significant. This insignificant effect of undervaluation is not in line with the results of McMillan and Rodrik (2011), who find a positive and highly significant effect of undervaluation on structural change. In column (2) the model is extended with the main variable of interest, where FDI enters the regression negative and insignificant. This suggests the importance of supportive host country conditions that could enhance the effect of FDI on structural change in SSA17.

5.2 Introducing Interaction Terms

In column (3) of table 5.1, interaction terms between FDI and variables in the basic McMillan and Rodrik (2011) model in column (1) are introduced in order to examine whether these variables are complementary to the effect of FDI on structural change. As is evident from column (3), FDI*ISG and FDI*Underval enter the regression with (highly) significant coefficients. Because, as expected, FDI*RawMat does not have a significant effect, this variable is excluded from the model (column 4). In addition, the main effects of both FDI and currency undervaluation turn highly significant when the interactions are included. This points out that FDI and undervaluation only affect structural change under specific complementary conditions. To explain the interconnected effects of the variables, at first the effect of FDI on structural change will be examined by means of the following equation, representing the results in column (4) of table 5.1:

) )

Without interactions terms in the model, would be the effect of FDI on the dependent variable. However, when interaction terms are included the interpretation of the coefficients changes. The effect of FDI on structural change does not remain limited to anymore, but is now represented by everything in the model that is multiplied by FDI. Using the regression results in column (4), this means that the effect of FDI on structural change is indicated by ( )

). The most straightforward interpretation is that without an initial structural gap, so with no employees working in the lowest productivity sectors , and no under- or overvaluation of the currency, the effect of FDI on structural change would be 0.04. With all employees in the lowest productivity sectors and no under- or overvaluation, the effect of FDI on structural change would drop by 0.06 and thus become negative. With an undervaluation of 1 and no initial structural gap, on the other hand, the effect of FDI on structural change tightens, but remains positive. Nevertheless, in

17_{All regressions in section 5.1, 5.2, 5.3, and 5.4 have also been carried out with robust standard errors. This}

does, most likely because of the small sample size, not change any of the main results significantly.

(31)

31 our sample neither the initial structural gap nor undervaluation ever take on values of zero or their maximum value. Since our interest lies in the circumstances under which FDI positively influences structural change, and whether this effect holds for SSA, it is examined how the variables can balance each other in order to derive a positive effect. ( ) ( )

gives all ‘equilibrium values’ for which FDI does not affect SC. The negative coefficients of both interaction terms indicate that when ISG is being kept constant, undervaluation has to be lower than its equilibrium value so that the equation exceeds 0 and FDI positively influences SC. Given the mean of ISG of 0.60, undervaluation should on average take on a value below 0.06 to make the effect of FDI on SC positive. Similarly, when undervaluation is being kept constant, ISG has to be lower than its equilibrium value in order to obtain a positive effect of FDI on SC. The mean of Underval is 0.20, indicating that ISG should take on a value lower than 0.53 to make FDI positively related to SC. If evaluated at both of the average values (ISG = 0.60 and Underval = 0.20), the unique effect of FDI is -0.0044, suggesting that the current mean levels of the initial structural gap and undervaluation make the effect of FDI on structural change trivial. A possible explanation for this simultaneous effect is that although higher undervaluation in the host country attracts more foreign capital inflows, if ISG is large at the same time FDI might flow into low productivity sectors because these host largest part of the labor force and are therefore more attractive. This either prevents labor to move towards more modern industries, or tightens the labor demand in the agricultural sector because of increased productivity, while the modern sectors do not have the capacity to employ the surplus labor, forcing labor allocation to even lower productivity sectors. The presumed positive influence of undervaluation on attracting investments in modern industries thus diminishes given a high ISG. Nonetheless the mean values of ISG and undervaluation indicate a negative effect of FDI on SC over 1990-2010 in SSA, conditional on the simultaneous values of ISG and undervaluation, FDI can potentially be positively related to SC in SSA.

The main effects of undervaluation and ISG on SC can be interpreted as follows. Using equation (4), the effect of undervaluation is presented by ). Because here the impact of the

variable of interest is only dependent on one other variable, the equilibrium value becomes a ‘threshold value’, which in this case is 0.33 for FDI. For FDI values higher than this threshold value, undervaluation is negatively related to SC, whereas for lower values of FDI undervaluation positively influences SC. Since the mean value of FDI is 0.019 and exceeds the threshold value of 0.33 only very rarely, it can be concluded that, in line with our expectation, undervaluation has a positive effect on the dependent variable. So, overvaluation tightens modern industries, while undervaluation promotes labor flows towards these industries. For ISG the effect is presented by ).

(32)

32 lower values of FDI the effect of ISG on structural change is positive, meaning that for 0.02, the mean value of FDI, ISG positively influences SC in the sample countries. It can thus be confirmed that a higher employment share in low productivity sectors at the beginning of the period increases the opportunity for labor to flow towards high productivity sectors.

5.3 Extending the model with Absorptive Capacity

Extending the model in table 5.1 column (1) with the absorptive capacity variables outlined in section 4 results in the outcomes presented in table 5.2. Because RawMat does not enter any of the regressions significantly in this case and on top of that contracts the significance of the other variables, this variable is excluded from the model. In the basic model (column 1), all other variables apart from FDI enter with the expected and significant sign, confirming our expectations outlined in the previous section; the initial share of labor force in low productivity sectors, currency undervaluation, financial development and institutional quality are all positively related to structural change, whereas human capital has a negative impact on the dependent variable. Like before, FDI enters the regression negative and insignificant, indicating that without additional requirements FDI does not positively influence SC. In columns (2)-(6) the interaction terms are introduced individually. As shown in column (2), FDI*ISG enters the regression negative and significant and additionally FDI turns positive and significant. Like before, the effect of FDI on SC can be explained by ). The threshold value of ISG above which the variable negatively influences the effect of FDI on SC is 0.475. Because the mean value of ISG is 0.602 it can be concluded that on average ISG negatively influences the effect of FDI on SC in SSA, which is consistent with the assumption that when large structural gaps are present at the beginning of the period, FDI will flow to lower productivity sectors. The threshold value for FDI under which ISG positively influences SC is 1.10. Since all values of FDI are below this threshold – except for TZA over the period 1990-1995 – it can again be concluded that ISG positively influences SC.

FDI*Underval enters the regression positive but highly insignificant, while the main coefficients of FDI and Underval also turn insignificant (column 3). This indicates that neither undervaluation alters the relationship between FDI and SC, nor does FDI affect the relationship between undervaluation and SC. Thus, the hypothesized positive impact of undervaluation on the effect of FDI on structural change is not confirmed by the data.

The Contribution of FDI to Structural Change