The recent surge in cross-border financial flows has generated an intense debate among economists and policymakers on the benefits and costs of the ongoing international financial integration. In principle, financial integration can promote productivity growth by allowing a country to borrow to finance productive domestic investment and increasing access to foreign advanced technologies. However, it can be accompanied with macroeconomic volatility and even financial crises. Fuelling this debate, the empirical evidence on the effects of financial integration on growth has been mixed (Kose et al. 2006).
Yet, recent financial crises led policymakers to rethink appropriate policies for growth. It is widely believed that the short-term debt is the most volatile source of foreign capital and can contribute most to instability of financial markets at a time of crisis, whereas foreign direct investment (FDI) is the most stable form of capital inflows. Importantly, FDI has long been regarded to be a major vehicle of technology transfer. More than ever, countries seek to leverage FDI for development. FDI represents the largest share of external capital flows to developing countries (United Nations Conference on Trade and Development 2007). (1) This seems to be driven by the belief that FDI brings positive effects to the economy, such as technology transfer, introduction of new production processes, and advanced management practices. However, it still remains to be an important empirical question whether and how FDI affects productivity growth in countries at various stages of development.
In this article, we study total factor productivity (TFP) in relation to technology diffusion through FDI for a couple of important reasons. First, debates over the relative importance between factor accumulation and TFP in raising income per capita took a dramatic turn. Recent studies found that more than half of the cross-country variation in both income per capita and its growth results from differences in TFP and its growth, respectively (Caselli 2005; Easterly and Levine 2001; Hall and Jones 1999; Klenow and Rodriguez-Clare 1997; Parente and Prescott 2001). (2) This finding suggests that, in order to understand the growth of nations, it is important to develop a better understanding of the forces that shape TFP.
Technological change is an important determinant of TFP. This was Robert Solow (1957)'s original view as well as the view of many economists in the literature (Helpman 2004). Endogenous growth models provide rigorous theoretical frameworks for understanding the economic forces underlying technological change. The models have focused on two important types of technological change: (1) innovation through research and development (R&D) and (2) technology diffusion through assimilating and adapting advanced foreign technology (see Barro and Sala-i-Martin 2003; Coe, Helpman, and Hoffmaister 1997; Grossman and Helpman 1991; Romer 1990, 1992 among others). Many of the earlier empirical studies focused on the effects on growth of innovation (measured by R&D expenditure or the number of scientists). The evidence on the positive impact on growth of innovation, especially at the microlevel, is substantial (Helpman 2004).
The other channel of technological change, technology diffusion, has received relatively less attention in the empirics. In a typical model of technology diffusion, technological change of a less-developed country depends on the extent of adoption and implementation of new technologies that are in use in the advanced countries (technology diffusion). That is, technological change largely consists of assimilating and adapting foreign technology. FDI is an important way to access advanced foreign technology. Beyond adding more capital to a receiving country, FDI can be the conduit to the production technology, cutting edge of R&D, and management expert. International technology diffusion can also take place through import of capital goods embodied with high technology (Eaton and Kortum 2001).
However, empirical research on the role of FDI in economic growth is still in its infancy (albeit growing rapidly) and has focused on growth of income per capita (e.g., Alfaro et al. 2004; Blonigen and Wang 2005; Borensztein, de Gregorio, and Lee 1998), not the TFP growth that is of our interest in this article. Given the striking evidence on the importance of TFP in explaining the cross-country income differences and the preeminence of the technology development and diffusion as a key determinant of TFP in endogenous growth theory, it is surprising that there is no cross-country empirical study on the effect of FDI on TFP growth. (3) In a comprehensive review of the literature on financial globalization, Kose et al. (2006) also conclude that how different types of capital flows including FDI affect TFP growth is one of the important future research topics. Our article fills this important gap in the literature. (4)
The goal of this article was to provide a comprehensive econometric analysis on the effect of FDI on TFP growth in a large sample of countries (developed and developing) for the 1970-2000 period. While exploiting both cross-sectional and time-series dimensions of data, we also carefully address the robustness and consistency of the results in terms of data, samples (developed vs. developing countries), estimation methods (cross-country and panel regressions), and outlier problem (robust estimation), which often plague the standard ordinary least squares (OLS) regression analysis.
Our results indicate that FDI has a significantly positive direct effect on TFP growth. Various estimation methods and robustness check yield largely the same result. To the best of our knowledge, our article is the first one that presents the evidence of positive direct effect on TFP growth from FDI in a cross-country study. (5)
Yet, it is interesting to compare our finding with the existing empirical studies on FDI and international financial integration. As noted earlier, the existing empirical articles on FDI have focused on growth of income per capita rather than TFP growth. Despite the distinct difference between income per capita and TFP, theoretical implications for direct effects of FDI on both per capita income growth and TFP growth are rather straightforward. However, earlier studies on FDI and per capita income growth failed to find a statistically significant positive effect of FDI on income growth. Instead, some of the studies (but not all) reported that the positive effect of FDI on income growth is only conditional on other factors such as human capital (Borensztein, de Gregorio, and Lee 1998) and financial development (Alfaro et al. 2004). (6) Thus, it became a popular view that the effect of FDI on income growth is only contingent on the recipient country's capability to absorb foreign technology.
Contrary to this popular perception and some of the studies on FDI and per capita income growth, however, we do not find any significant evidence that the contribution of FDI to TFP growth is only contingent on the recipient country's capability to absorb foreign technology, regardless of how the absorptive capability is measured (human capital, financial market development, or institutional quality). (7) It seems that technology diffusion process through FDI flows affects the TFP growth differently. (8)
Our finding of positive effect of FDI on TFP growth is also in contrast with the empirical studies on the effect on income growth of international financial integration that often yield only ambiguous results. For example, Edison et al. (2004) examined various measures of international financial integration (such as volume of capital flows including FDI, equity, and debt) and confirmed lack of a robust relationship between financial integration and per capita income growth (see Kose et al. 2006 and references therein). (9)
The plan of the article is as follows. In Section II, we discuss the concept of TFP and our new data set on TFP. In Section III, we briefly discuss the empirical literature on FDI and economic growth and data on FDI. In Section IV, we present our econometric analysis of FDI's effects on TFP growth. In Section V, we address the robustness and consistency of the results in terms of reversed causality, outliers, and unobserved omitted variables, and then conclude in Section VI. Additional information on data is provided in the Appendix Tables 1 and 2.
II. TOTAL FACTOR PRODUCTIVITY
Consider a standard aggregate production function where aggregate output (Y) depends on physical capital (K), labor (L), human capital (H), and TFP or stock of knowledge (A):
(1) Y = Af(K, H x L),
where H x L is human capital augmented labor (i.e., labor in efficiency units). Growth of aggregate output will depend on the rate of change of those four factors. The growth rate of TFP, which is obtained as a residual in the growth accounting, is often ascribed to technological progress. TFP can change for many reasons. First of all, an increase in stock of knowledge about production methods. The endogenous growth theory focuses on technological progress that results from intentional industrial innovation through R&D activities in response to their expected profits such as monopoly rents (for seminal papers, see Grossman and Helpman 1991; Romer 1990). Both the costs of R&D and the rewards that innovators gain are influenced by conditions in product (including market size), factor (such as skilled labor) and capital markets, and government policies and institutions that govern these market conditions.
In the context of less-developed countries, technological change is related to the extent of adoption and implementation of new technologies that are in use in the advanced countries (technology diffusion). That is, technological change largely consists of assimilating and adapting foreign technology. FDI can provide an access to advanced foreign technology, such as production technology, cutting edge of R&D, and management expert, while boosting market competition and generating spillovers and externalities to local firms in the host economy. Thus, we expect a positive direct effect of FDI on TFP growth.
Data on TFP
We construct TFP growth rates and exploit both their cross-country and panel dimensions for 92 developed and developing countries in the period of 1970-2000. (10) National income and product account data and labor force data are obtained from the PWT version 6.2 of Heston, Summers, and Aten (2006). Taking a standard neoclassical approach, we assume a Cobb-Douglas production function:
(2) Y = A[(K).sup.[alpha]] [(H x L).sup.1-[alpha]],
where 1 - [alpha] is labor income share. (11)
To construct the labor quality index for human capital (H), we take average years of schooling in the population more than 15 yr old from an international data on educational attainment of Barro and Lee (2000). We follow Hall and Jones (1999) and Klenow and Rodriguez-Clare (1997) to give larger weight to more-educated workers as follows:
(3) H = [e.sup.[phi](E)],
where E is average years of schooling and the function [phi](E) is piece linear with slope of 0.134 for E [less than or equal to] 4, 0.101 for 4 < E [less than or equal to] 8, and 0.068 for 8 < E. (12) The rationale behind this functional form for human capital is as follows. The wage of a worker with E years of education is proportional to his human capital. Since the wage-schooling relationship is widely believed to be log linear, this would imply that human capital (H) and education (E) would have a log-linear relation as well, such as H = exp(const. E). However, international data on education-wage profiles (Psacharopulos 1994) suggest that in sub-Saharan Africa (which has the lowest levels of education), the return to one extra year of education is about 13.4%, the world average is 10.1%, and the OECD average is 6.8%. Thus, Hall and Jones's specification above reconciles the log linearity at a country level with the convexity across countries. (13) We also calculated the TFP growth using an alternative data on educational attainment from Cohen and Soto (2001). (14)
We estimate the capital stock, K, using the perpetual inventory method:
(4) [K.sub.t] = [I.sub.t] + (1 - [delta])[K.sub.t-1],
where [I.sub.t] is the investment and [delta] is the depreciation rate. Data on [I.sub.t] are from PWT 6.2 as real aggregate investment in purchasing power parity. (15) We further adjust these capital stocks for the portion of residential capital stock that is not directly related to production activity. (16)
Two batteries of consistency checks suggest that our estimates of TFP growth are reasonable. First, the correlation coefficients between the TFP growth estimate based on Barro and Lee (2000) human capital data and the TFP growth estimate based on Cohen and Soto (2001) human capital data is 0.98 for the 1970-2000 period. For our 10-yr decade panel data, the correlation between them is 0.97. Not surprisingly, the regression results are very similar, regardless of which human capital data we use to compute TFP. Thus, we mainly report regression results based on the largest sample in which we use Barro and Lee (2000). In our sample of 92 countries, Cohen and Soto (2001) have data points for 77 countries only. Second, the correlations between our TFP growth estimate and that from Bosworth and Collins (2003) are 0.89 (number of observations = 78) for the 1970-2000 period and 0.80 for the 10-yr decade panel data. (17)
III. FOREIGN DIRECT INVESTMENT
A. The Empirical Literature
Recently, economists have begun to examine whether FDI, a factor largely ignored in the empirical growth literature, has an independent direct impact on per capita income growth (e.g., Blonigen and Wang 2005; Melitz 2005; Kose et al. 2006). As noted earlier, there is no cross-country study that examines the effect on TFP growth of FDI yet.
Despite the straightforward theoretical implication for direct effect of FDI on per capita income growth, however, earlier empirical studies on FDI and income growth have failed to find statistically significant positive impact of FDI on income per capita growth. Instead, some of the studies, but not all, have reported the positive effect of FDI on income …