Abstract 

This paper attempts to reconcile the often inconclusive evidence on the impact of FDI on growth by taking two into account the heterogeneity both among industries and among countries. Using a comparable database at the industry level for 35 countries in OECD, Asia and Eastern Europe from 1987 to 2002, we test both stage of development and FDI industrial pattern for the economic impact of FDI on growth. In certain industries and for the catching-up countries, a significant and positive relationship emerges when 

FDI interacts with investment or export orientation. 

CHAPTER ONE

1. Introduction 

While in theory the nexus between FDI and growth (in terms of output and productivity) is in general positive, the empirical literature is far less conclusive. Some studies find positive effects from outward FDI for the investing country (Van Pottelsberghe and Lichtenberg, 2001; Nachum et al., 2000), but suggest a potential negative impact from inward FDI on the host country. This results from a possible decrease in indigenous innovative capacity or crowding out of domestic firms or domestic investment. Thus, in their view and in line with the standard literature on the determinants of FDI (i.e. Dunning’s OLI paradigm, see Dunning 1988) inward FDI is intended to take advantage of host country (locational) characteristics instead of disseminating new technologies originating in the sending country. Other studies report more positive findings: Nadiri (1993) finds positive and significant effects from US sourced capital on productivity growth of manufacturing industries in France, Germany, Japan and the UK. Also Borensztein et al. (1998) find a positive influence of FDI flows from industrial countries on developing countries’ growth. However, they report also a minimum threshold level of human capital for the productivity enhancing impact of FDI, emphasizing the role of absorptive capacity. Absorptive capacity or minimum threshold levels in a country’s ability to profit from inward FDI is often mentioned in the literature (see also Blomström et al. 1996). Consequently the effect of FDI depends among other things to a large extent on the characteristics of the country that receives FDI. However, the resulting issue of cross-country heterogeneity has largely been neglected in the literature so far with few exceptions. Blonigen and Wang (2005) stress explicitly cross-country heterogeneity as the crucial factor which determines the effect of FDI on growth. Further, Nair-Reichert and Weinhold (2001) and Mayer-Foulkes and Nunnenkamp (2005) explicitly take up this aspect in their analysis. Our paper will follow their direction and introduce two forms of heterogeneity, differences between countries and differences between receiving industries. 

We argue that since host country heterogeneity plays a role, it is equally likely that the impact of FDI on the host economy differs greatly according to the receiving industry. FDI in constant returns to scale industries will have different effects than FDI in increasing returns to scale industries. Likewise, the effect of FDI may be related to the technology and human capital intensity of the industry and other factors. As a very intuitive example, heavy FDI in the extractive sector in Nigeria has not improved the country’s growth performance (Akinlo, 2004). Consequently, the potential for positive spillovers does not solely depend on a country’s overall absorptive capacity, but also on which sectors or industries in the economy receive FDI. Thus, the impact of FDI differs depending on country specific absorptive capacity or stage of development as well as on the sectoral and industrial structure and allocation of FDI. Since the two are in general related, this implies a relationship between the industrial pattern of inward FDI and its effect on the host country. The economy wide effect of industry specific FDI inflows will then further depend on the extent of intra-industry versus inter-industry spillovers. 

In this paper we use a unique panel data set at the industrial level for a range of industrialized and catching-up countries to investigate the magnitude of all these factors for the role played by FDI in different manufacturing industries in the economic development of the host economy. Due to measurement issues, interdependencies between various types of spillovers and their complexity, it is difficult to distinguish between different theoretically possible channels of technology transmission in empirical research. Therefore, we will focus on the overall effect of foreign sourced capital on manufacturing productivity growth in addition to the effects of traditional factors (domestic capital and labour) and controlling for other factors. What is new in our analysis is the focus on the industry-level of the economy. To our knowledge, there is very little empirical research on FDI at this level of disaggregation. Disaggregated data on FDI for a large and heterogeneous set of countries rarely exist in a comprehensive and comparable form. If these data exist, they are often plagued with two kinds of problems: On the one hand, the coverage of firms and flows which are recorded as FDI can differ between countries (problems are often caused by the exclusion of reinvested profits in some countries). On the other hand, the classification into industrial activities may differ between countries. We therefore established a comprehensive new data set in order to address adequately the role of industrial FDI patterns. 

The paper proceeds as follows: The next section describes the data set. Section 3 revises briefly the theoretical background of the FDI-growth nexus and introduces the estimating framework. The results are summarized in Section 4. Section 5 concludes. 

2. Industrial Patterns of Inward FDI 

Due to a lack of comparable data at the industry level, empirical research on the link between FDI and development has largely remained at the macro level, since comparable FDI data across countries are best available at this level. More recently, firm-level datasets have been released and, as a consequence, the number of studies using micro data has grown rapidly. However, in contrast to the macro-level analysis, which often takes a global perspective and analyses large cross-country data sets (in the cross-section dimension as well as in the panel dimension), many firm-level studies are limited to one country or a homogenous group of countries (like the EU) due to issues of data-availability and comparability. 

In order to get a good picture of the link between FDI and growth of individual industries, we collected indicators like output, gross fixed capital formation, inward FDI stock data and exports from several sources (UNIDO, UN COMTRADE, OECD, wiiw, ASEAN Secretariat, Timmer 2003, MOEA 1993). Woerz (2005) gives a detailed description of the dataset. In total, our data set contains more than 3000 observations for 28 to 35 countries, eight industries and 14 years (1987-2000). The data set is highly unbalanced, the number of countries varies over time, with data for 28 countries over the years 1987 to 1997 and data for 35 countries over the years 1998 to 2000. We collected information on output, employment, FDI inwards stock, gross fixed capital formation, exports and imports for a rather heterogeneous group of countries: OECD members, ASEAN members, other dynamic Asian countries and Central and Eastern European countries. The time series for CEECs 

start only in 1993 or later. 

The ratio of inward FDI stock to output varies along all dimensions, across industries, years and countries. For the complete sample, the FDI to output ratio ranges from far less than 1% in the textiles and wood industry in Japan to more than 100% in the industry group comprising fuel, rubber, plastics and chemicals in Indonesia. Also the variance is highest in the latter industry (see Figure 1). 

[insert Figure 1] 

It is further striking to see not only the rise in average FDI to output ratio, but also the rapid increase in variance over time. In some cases, the ratio of FDI to total industry output has increased to 100%. The general rise in FDI in relation to industry output clearly reflects the increasing internationalisation of production. The additional sharp increase in variance across countries tells us that this internationalization did not happen at equal rates for all countries and industries. While Asian countries on average show higher shares of FDI in total industry output, they also exhibit much more variation across individual countries than OECD members. Entering the picture at a much later point in time, CEECs show again substantially higher FDI to output ratios, yet with considerably less variation across countries. Thus, this region experienced a uniformly high inflow of foreign capital into manufacturing. On average, CEECs display higher FDI to output ratios than most other countries in the sample. Many of the former communist countries allowed and actively encouraged the inflow of foreign capital as a way to privatize the former state owned companies. Due to a general lack of domestic capital and the disruption of state owned companies, with many inefficient firms exiting the market, the share of foreign capital was especially high in transition countries. 

Table 1 below gives the weighted averages of inward FDI stocks to output ratios for different geographic regions towards the end of the observation period. The table illustrates nicely the two sources of heterogeneity, stemming from differences among countries as well as between industries. There are distinct differences between industries, with the highest ratio in general prevailing in the petroleum, chemical, rubber and plastic industry. CEECs are the only region with the highest FDI to output ratio in the transport industry. Apart from the strong role of FDI in petroleum, chemicals, rubber and plastics, all regions differ with respect to the importance of FDI in individual industries. Thus, the data exhibit very large disparities across regions as well as across industries, supporting our argument of the two sources of heterogeneity in the relationship between FDI and output or productivity. 

[insert Table 1] 

Structural developments have also been diverse between these regions, with more structural change in both groups of OECD countries as opposed to more stable FDI patterns in East Asian countries (including the four Tigers). There is further a low correlation between FDI patterns and output patterns in both groups of OECD countries and in Emerging Asia. The four Asian Tiger economies and CEECs seem to be different with closely matching FDI and output structures. In the case of the four Tiger countries, where the observation period extends over 20 years, the sequencing of industry patterns suggests that high FDI shares in electrical machinery have resulted in subsequently high output shares in the industry. For the CEECs, the time period is too short for any conclusions. However, FDI seems to play a more important role in these two regions and less so in others. 

3. Theoretical Background 

3.1 The FDI-growth nexus in theory 

Economic theory has provided us with many reasons why foreign direct investment may result in enhanced growth performance of the receiving country. In the neo-classical growth literature, FDI is associated positively with output growth because it either increases the volume of investment and/or its productivity and thus puts the economy on a higher longterm growth path. In an exogenous growth model, FDI has only a level effect in the steady state and no permanent impact on the growth rate, except during the transitional dynamics to the new steady state. The potential role for FDI is much greater in endogenous growth models. In a neoclassical production function, output is generated by using capital and labour in the production process. With this framework in mind, FDI can exert an influence on each argument in the production function. FDI increases capital, it may qualitatively improve the factor labour (explained below) and by transferring new technologies, it also has the potential to raise total factor productivity. Further, as discussed in more recent theoretical growth models (e.g. by Grossman and Helpman, 1991) by raising the number of varieties for intermediate goods or capital equipments, FDI can also increase productivity (see Borensztein et al., 1998 for an empirical analysis of this channel). Thus, in addition to the direct, capital augmenting effect, FDI can also have additional indirect and thus permanent effects on the growth rate. Most importantly, FDI can permanently increase the growth rate through spillovers and the transfer and diffusion of technologies, ideas, management processes, and the like. 

The literature mentions different channels at the micro level which allow for technological spillovers from FDI to the host economy (Kinoshita, 2001, Halpern and Muraközy, 2005): through imitation, through the labour market (i.e. training of local workers by foreign firms), by inducing a competition effect and through backward or vertical spillovers. For all of these channels, host country characteristics – in particular the relative development level of the host country compared to the sending country in terms of technology, human capital etc. – play a relevant role. 

A potential problem at the micro-level, where the spillovers arise, is the evidence for selfselection bias. I.e. while there is general consensus that FDI increases the productivity of receiving firms, part of this effect is in fact due to FDI selecting better firms as targets for takeover (Bellak, 2004). At the more aggregate level, this translates to the imminent causality or endogeneity problem, faced by all empirical studies on the effects of FDI. 

Another crucial role in this context is played by the absorptive capacity of the host country. The importance of absorptive capacity – often captured by differences in the stage of development between donor and host country - has been a central finding in many empirical studies on the FDI-growth link (Blonigen and Wang 2005; Borensztein et al. 1998; De Mello 1999). There are also theoretical justifications for the importance of a certain amount of absorptive capacity. For example, Markusen and Rutherford (2004) develop a three-period model where they show that the speed and degree of positive spillovers from FDI is positively related to the absorptive capacity of the host country. In an earlier paper and using a new economic geography model, Rodriguez-Clare (1996) relates the developmental impact of multinational firms to the type of the linkages which they create. Positive linkage effects are stronger the more intensive the multinational is in the use of intermediate goods, the larger the costs of communication and trade between headquarters and local plants and the more similar home and host country are in terms of the variety of intermediate goods produced. This implies stronger linkages - and thus greater positive effects - if the developmental gap between donor and host country is smaller. 

To summarize, one can argue that positive spillovers will only occur in a suitable setting. If the host economy does not provide an adequate environment in terms of human capital, private and public infrastructure, legal environment and the like, many of the spillovers that may potentially arise from FDI can not materialize. Public infrastructure such as educational institutions, publicly funded R&D collaborations can significantly support potential spillovers. As a consequence, country specific effects have a strong influence on the effects of FDI on growth. Hence, cross country heterogeneity is one of the important aspects to be addressed in empirical research on the topic. In addition, different stages of economic development are characterised by specific industrial patterns. In line with the previous arguments, a high structural match between the donor and the host country would imply a proximity in stage of development and thus also a good precondition for the absorptive capacity of the receiving country to be high. In other words, the match between the industrial allocation of FDI and the host country’s stage of development as characterised by its industrial structure determines the effectiveness of FDI. We argue in this paper that the “optimal” pattern of FDI across industries varies with the stage of development. The effect of FDI in the same industry but in countries at different stages of development can be just as much different as the effect of FDI in one country but in different industries. Thus, we will address both, cross country and cross industry heterogeneity in this paper.