FDI Activities, Exports and Manufacturing Growth in a Small Open Economy: An Industry-wise Panel Data Analysis. Ananda Jayawickrama 1.

Transcription

1 FDI Activities, Exports and Manufacturing Growth in a Small Open Economy: An Industry-wise Panel Data Analysis Ananda Jayawickrama 1 And Shandre M Thangavelu 2 1 Department of Economics and Statistics, University of Peradeniya, Peradeniya, Sri Lanka Tel: (94) anandaj@pdn.ac.lk and ajayawickrama@gmail.com. 2 Corresponding author: Shandre M Thangavelu, Department of Economics, National University of Singapore, AS , 1 Arts Link, Kent Ridge Crescent Singapore , Tel. (65) , Fax. (65) ecssmt@nus.edu.sg.

2 Abstract The paper examines the influence of FDI on manufacturing growth of Singapore in a panel data sample of 14 manufacturing industries over 30 years stretching from 1975 to By controlling for unobserved industry characteristics and time effects, we find a positive contemporaneous effect of FDI on the output growth of Singapore manufacturing industries where 1 percent increase in FDI tends to increase manufacturing output growth by nearly 0.4 percent. We also observed positive impact of FDI on manufacturing output growth using Arellano-Bond GMM estimator that controls for the endogeneity problems in the estimation. Keywords: FDI Inflows, GMM Estimation, Panel Data JEL: F21, O14, O53 C23

3 1. Introduction Singapore economy is an open economy that relies heavily on foreign investment to maintain its competitiveness and drive its economic growth. FDI in export-oriented industries, particularly the electronics industry, has been the key factor in driving the exportled growth in the Singapore economy. Although there are several studies highlighting the importance of FDI on the Singapore economy, most of them are focused on the impact at the aggregate economy and there is lack of empirical studies to study the impact of FDI at the manufacturing sector at a disaggregated level (Hu, 2004; Chang, 2005; Anwar, 2008; Low, 1999). This paper intends to fill this gap by examining the impact of FDI flows on the growth of the Singapore economy at the disaggregated manufacturing industry level. To our knowledge, this is the first paper to examine the impact of FDI on manufacturing output growth using a disaggregated industrial level data. The role of FDI on output growth of economies has been extensively analyzed in the literature. Traditional growth models as well as endogenous growth models highlight the importance of technology and efficiency improvements in stimulating economic growth and hence provide the framework to analyze the relationship between FDI and economic growth. These growth models highlight that FDI inflows lead to high output of the recipient economy by increasing investment and/or enhancing the labour productivity. 3 In an excellent survey of literature, De Mello (1997) lists two channels through which FDI inflows enhance economic growth: adoption of new technology in the production process through capital spillovers, and 3 FDI inflows augment domestic capital formation and expand the production capacity of the economy. As technological progress is a major factor in the endogenous growth models, FDI inflows could have a permanent impact on economic growth through technology transfer, diffusion and spillover effects. Findlay (1978) postulated that FDI inflows would promote economic growth through technological transfer and knowledge diffusion.

4 knowledge transfers through labour training and skill acquisition and better management practices. However, empirical evidence on these issues using single country time series or cross sectional study is rather inconclusive. Nair-Reichert and Weinhold (2001) note that while many studies argue that FDI inflows may have positive impact on economic growth of the recipient economy through technological diffusion and capital formation, others suggest that these positive effects may not be unconditional and points to the lack of technology transfer and spillover effects. Macroeconomic studies which examine the causality between FDI and growth using aggregate FDI inflows and growth data in a cross country framework, generally, suggest that FDI inflows positively affect economic growth. Zhang (2001) finds that FDI strongly Granger-cause GDP growth in a sample of 11 countries. Choe (2003) finds a bi-directional causality between economic growth and FDI in a sample of 80 countries over the period However, their results also show that the causality is rather more apparent from growth to FDI than from FDI to growth. In a sample of 32 countries that includes OECD and non-oecd countries and using a single-country time series regression framework, De Mello (1999) find that the long-term effect of FDI on growth is heterogeneous across countries. He does not find firm evidence for the positive effect of FDI on growth in a panel of non-oecd countries. Nair-Reichert and Weinhold (2001) find that FDI on average has a significant positive impact on growth though the relationship is highly heterogeneous across countries. There are many studies that attempt to draw conclusions on FDI-growth causation by controlling for human capital, openness of the economy and different stages of growth. Blomström et al. (1996) find that FDI inflows are an influence on growth rates for high income developing economies, but not for lower income ones as it depends more on

5 domestic factors such as secondary education, changes in labour force participation, and infrastructure. Balasubramanyam et al. (1996) find that FDI promotes economic growth in a sample of 46 developing economies during the period Their results further revealed that FDI inflows are more productive in countries with export promoting trade and investment strategies than with import-substituting strategies. Basu, Chakraborty and Reagle (2003) also emphasize trade openness as a crucial determinant for the impact of FDI on growth. By revisiting these findings in the context of more recent cross-sectional data, Greenaway et al. (2007) confirm the robustness of the impact of FDI on economic growth. The heterogeneity of the results of these macro level studies indicates different country specific effects and also points to various specification issues in models. There are arguments that these studies do not fully control for simultaneity bias, country-specific effects, and the lagged effects of dependent variables in growth regressions (Carkovic and Levine 2005). By addressing these issues in data, Carkovic and Levine (2005) find that the exogenous component of FDI does not exert a robust and positive influence on economic growth. Hansen and Rand (2006) using mean group estimator find a strong causation from FDI to GDP, and their results indicate that FDI appears to be growth enhancing much in the same way as domestic investment. Similar to country-specific macro studies, studies which uses micro-level data are also fail to draw strong conclusions on the effect of FDI on growth. While several studies find positive association between FDI and output growth (see for recent work by Branstetter 2006, Kneller and Pisu 2007), some studies find that FDI has negative impact on productivity (see for example Aitken and Harrison 1999, Konings 2001). 4 In this paper, we avoid the above country specific effects and heterogeneity by 4 See Görg and Greenaway (2004) for an extensive survey of the literature..

6 examining a disaggregated industry level data for a specific economy such as the Singapore economy. This study differs from the previous studies in terms of examining the impact of FDI inflows on manufacturing economic growth using disaggregated industry level data for the Singapore economy. In particular, we use the 2-digit industrial classification. In this paper, we examine the FDI influence on growth of Singapore s manufacturing sector at the 2-digit industrial classification using a panel data of 14 industries from The strong panel data will provide us with the framework to capture the dynamic relationship between FDI and output growth at the industry level. The results indicate that the contemporaneous influence of FDI on Singapore s manufacturing growth is quite high and significantly different from zero. The manufacturing output growth rate would increase by about 0.4 percent in response to one percent increase in FDI to output ratio from the previous year. Though the effect is highly significant, we believe that endogeneity biases may have boosted the effect of FDI on output growth. In a dynamic GMM model that controls for the endogeneity and omitted variable biases through the use of lagged independent and dependent variables as instruments, we find that one percent increase in the FDI ratio would increase the current output growth rate by nearly 0.2 percent. The effect is quite robust to the variation in the information set given the orthogonality condition among regressors. The rest of the paper is organized as follows. In section 2, we provide an overview of the FDI inflows into the Singapore manufacturing sector. Section 3 discusses the methodology and the data set of the study. Section 4 discuses the empirical results. Section 5 gives the policy conclusion of the paper.

7 2. Overview of FDI inflows in Singapore Singapore manufacturing sector has undergone significant changes over the last three decades. Its output share increased from 24 percent in 1975 to around 30 percent in recent years. As given in Table 1, the average growth rate of the manufacturing output is over 10 percent during the period The growth of manufacturing sector seems to be mostly export driven as direct exports of the sector closely follows the growth path of output. It is observed that Singapore manufacturing industries are highly export oriented as direct exports account for 63 percent of average manufacturing output. Further, direct exports of manufacturing industries reported more than 11 percent average growth rate over the last 30 years. Although it grew at a relatively slower pace, the average manufacturing employment also closely follow output fluctuations (see Figure 1). Importantly, manufacturing sector FDI inflows averaged 12 percent growth rate over the past three decades. Rising from around 20 percent in mid 1970s to about 50 percent in recent years, the ratio of manufacturing FDI inflows to output averaged 32 percent over the sample period. These statistics indicates that Singapore s manufacturing sector has attracted a large volume of foreign investment over the past decades and therefore its output growth is highly dependent on FDI inflows. As illustrated in Figure 1, there is a co-movement between output growth and FDI growth over time despite weak linkages in some periods. FDI inflows to the country s manufacturing sector reported large fluctuations in 1990s as compared to the previous decade and it seems to be more stable in recent years. Among the industry categories of concern (see Table 1), electronics, chemicals, fabricated metals, machinery, paper, transport equipments, and precision instruments are the high growth industries. Industries such as wood, textile, petroleum, food and basic metal productions reported relatively slower average rate of output

8 growth. In all industries, except for wood, textiles and basic metals industries, high output growth rate is associated with high FDI growth although the direction of causality is unknown. The scatter plot of rate of output growth and change in FDI ratio omitting few outliers illustrates that FDI and manufacturing output have strong positive correlation over the sample period. ===================== Insert Figure 1 and Table 1 ===================== ===================== Insert Figure 2 and Figure 3 ==================== 3. Methodology and Data We examine the impact of FDI on output in a production function framework, while controlling for other causes of growth. We assume that the production of manufacturing industries can be approximated by Cobb-Douglas technology with capital and labour. The incorporation of FDI into the growth equation is not unique among researchers. For example, the Nair-Reichert and Weinhold (2001) model regressed growth rate of output (first differenced of log output) on the growth rate of foreign investments (first differenced of log FDI), while Hansen and Rand (2006) shows that growth rate of output should depend on the first differenced of FDI ratio (FDI/Y). Given that the magnitude of FDI may be negative in some periods, we specify the relationship between output and FDI ratio in a semi-log framework. Ignoring the impacts of other variables for the moment, the relationship is given as log Y = f[( FDI / Y) ] where Y is output, i is industry category and t is time subscript it it

9 and f > 0. This allows the modeling of the growth rate of output (logyit log Yit 1) as a function of change in FDI ratio [( FDI / Y ) it ( FDI / Y ) it 1]. Thus, FDI ratio should have a positive impact on the growth rate of output given the influence of other variables. We specify industry-wise manufacturing output growth in a full regression model as follows: Δ y = βδ ( FDI / Y) + ( Δ x ) ψ + ε (1) it it it it where is the differenced operator, y it = logy, x = log X, X is an (n 1) vector of it it it predetermined variables other than (FDI/Y) and ψ is an (n 1) vector of coefficients. By incorporating the industry-wise heterogeneity, we write equation (1) as follows: Δ y = βδ ( FDI / Y) + ( Δ x ) ψ + Z α + ε (2) it it it i it where Z α embodies all the unobservable industry-effects or fixed effects in which Z is an (n 1) vector of unobserved industry characteristics and α is an (n 1) vector of coefficients. In equation (2), the coefficient β gives the contemporaneous correlation between Δ ( FDI / Y ) and the growth rate of industry output. However, many will argue that endogeneity bias may distort the estimated contemporaneous correlation coefficients (see Nair-Reichert and Weinhold 2001). The bidirectional causality between output and FDI invokes an identification issue in equation (2) as we might have failed to identify the exact influence of FDI on output since data generating processes of these variables are contemporaneously correlated. The issue of endogeneity bias may be serious in our model as some of the variables in vector X may also be co-determined with output. In a dynamic model of industry panels, we are, nevertheless, able to control for the endogeneity issue by using lagged values of endogenously determined dependent

10 variables as instruments. The inclusion of lag dependent variables will also solve the biases due to omitted variables. The dynamic specification of equation (2) is given as follows: Δ y = δ Δ y δ Δ y + % β Δ ( FDI / Y) % β Δ( FDI / Y) it 1 it 1 p it p 1 it 1 p t p +Δ ( x ) ψ% ( Δ x ) ψ% + Z α + % ε (3) it 1 1 it p p i it where p is lag length. In a panel data setting, coefficients of equation (3) can be estimated by GMM as proposed by Arellano and Bond (1991). The effects of FDI ratio on the manufacturing output growth is given by, p β i= 1 i % in equation (3) and it may differ from the contemporaneous correlation effect ( β ) in equation (1) or (2). In the cross-country panel data literature, the growth rate of output (GDP) is regressed on change in FDI ratio/growth rate of FDI by controlling for several other variables. The vector of control variables normally includes gross domestic capital formation, growth rate of exports, human capital measure, trade openness and policy variables such as rate of inflation. Since our focus is on the influence of FDI on output growth of manufacturing industries in a single country context, our control variables are obviously differs from cross-country panel data studies. As the testable equation is derived from the standard Cobb-Douglas production function, one of the controlled variables is the growth rate of industry-wise manufacturing employment. We expect growth rate of employment to have positive influence on output growth. Since Singapore manufacturing industries are highly export oriented, the growth rate of exports is expected to have positive impact on manufacturing output. We also used other variables such as growth rates of material inputs (value) and number of firms operating in the business as instruments to control for endogeneity in the model. The number of firms in each industry

11 could be a very good instrument as it will account for the change in the production capacity in the industry. Our sample includes 14 manufacturing industries: (1) food, beverage and tobacco products (2) textiles, wearing apparel and leather products (3) woods and wood products (4) paper and paper products, printing and publishing (5) chemicals and chemical products, pharmaceutical and biological products (6) petroleum and petroleum products (7) rubber and plastic products (8) basic metals (9) fabricated metal products (10) machinery and equipment (11) electronic products and components, electrical machinery and equipment (12) transport machinery and equipment (13) instrumentation, photographic and optical goods and (14) other manufacturing products. The time period of the study covers 31 years stretching from 1974 to Foreign direct investment inflows at the manufacturing industries are given in the annual reports of Foreign Equity Investment in Singapore published by the Singapore Department of Statistics. Data on industry-wise output, employment, material inputs, direct exports, number of firms in operation and gross fixed assets are obtained from the Annual Census of Manufacturing Activities conducted by Economic Development Board of Singapore. We aggregate SITC two-digit level industry data of the census to comply with our industry classification and match this with the foreign direct investment inflows from the Foreign Equity Investment survey. The first differenced of gross fixed assets is treated as the gross fixed capital formation (GCF). Gross domestic capital formation deflator and manufacturing price indices from the Yearbook of Statistics are used to deflate gross fixed capital and industrial output respectively. We deflate all variables to 2000 prices. The current values of FDI and GCF are deflated by gross domestic capital formation deflator. Real values

12 of industry-wise output, direct exports, and material inputs are obtained deflating by the most suitable manufacturing price index given above. 5 Growth rates (percentage) are computed using industry-wise real values of the respective variables. Industry-wise FDI ratio and GCF ratio give industry-wise FDI inflows and GCF as a percent of industry-wise output. 4. Results We first estimate the fixed effect model with and without time effects and provide the results in Table 2. Our modeling approach is general-to-specific starting with all possible information at first and later controlling for the most important information. In the first regression, growth rate of employment, exports and number of firms have statistically significant and positive impact on growth rate of manufacturing output. However, the effects of FDI ratio, growth rate of material inputs and GCF ratio are statistically insignificant. The coefficient on GCF ratio is statistically insignificant but negative casting doubts on possible co-linearity between FDI and GCF ratios. As Figure 3 illustrates, there is a clear positive correlation between FDI ratio and GCF ratio, while other pair of variables do not show such a systematic relationship. Although dropping material input variable adds no difference, the model without GCF variable is significantly different from the previous ones. The impact of the co-linearity between FDI and GCF ratios is clearly seen in the regression of equation (3), as the coefficient on FDI ratio declines by the same magnitude as that of the GCF ratio in the previous regression. Most importantly, it makes the coefficient of FDI ratio statistically significant without the loss of model predictive power. Thus, our results suggest 5 The appropriate price index by industry is obtained from the manufacturing producer price index as given in the Singapore Yearbook of Statistics.

13 that FDI have a significant positive impact on the growth rate of manufacturing output along with employment, exports and firm expansion. Table 2 also provides results of fixed effects regression with time effects. Time dummies would capture the impact of any unobserved macroeconomic changes that affect any industry at any given time. The results indicate that the influence of FDI on output growth becomes highly significant when GCF ratio is dropped. However, the inclusion of time effects distorts the impact of other variables. The effect of employment has decreased markedly and become insignificant. Although it is still highly significant, the effect of export growth has also declined. Further, the inclusion of time effects affects the level of significance of the firm expansion. The impact is statistically different from zero only at the 10 percent level of statistical significance. This signals that time dummies may capture significant part of the fluctuations of the output growth which can otherwise be explained by labour, exports and expansion of firms. ================ Insert Table 2 ================ Another fact that we observe in the fixed effect regressions (both with and without time effects) is the low correlation between the explanatory variables and the disturbance term. The correlation coefficient between regressors and the disturbance term does not exceed 0.1 suggesting that heterogeneity across panels is very low and thus both the fixed effect and random effect models may produce the same results. Table 3 provides the random effect estimates. Since material input is consistently insignificant in our previous regressions, we have dropped it in subsequent estimations. Again, the influence of GCF ratio on output growth becomes negative and insignificant when FDI ratio was included in the regression.

14 We observe that the growth rate of output rises by nearly 0.3 percent in response to 1 percent increase in FDI ratio. Compared to the marginally insignificant influence in the fixed effect model with time effects, growth rate of employment has highly significant positive effect on the growth rate of output. Manufacturing output will grow by 0.4 percent in response to 1 percent growth in the manufacturing employment. The growth rates of exports and expansion of firms have highly significant positive effects on the growth rates of output. Again dropping GCF ratio improves the significance of the FDI coefficient indicating the high multicollinearity between the variables. Since OLS residual based Breusch-Pagan LM test statistic (= 3.36) exceeds the 90 percent critical value for chi-squared with one degree of freedom (= 2.71), the classical linear regression model with single constant term is inappropriate for these data. That is, the Breusch-Pagan LM test tends to favour the random effect estimation. Further, the Hausman test was performed to see whether the explanatory variables are correlated with individual effects by using fixed effects models without and with time effects. 6 The computed Hausman chi 2 test statistic is 2.32 when the time effect is excluded from the fixed effect model. As the Hausman chi 2 test statistic is far below the chisquared critical value (= 13.3), the hypothesis that the individual effects are uncorrelated with other regressors in the model cannot be rejected. Thus, based on the Breusch-Pagan LM test and the Hausman test results, random effect model emerge as the better choice for our data. However, the null hypothesis of uncorrelated individual effects with other regressors is clearly rejected when the time effect is included in the fixed effect model (Hausman chi 2 test statistic is 24.4). This suggests that time dummies have resulted in an increased correlation 6 Hausman test is based on the idea that under the null hypothesis of no correlation between individual effects and other regressors both fixed effects and random effects estimates are consistent but fixed effect estimates are inefficient. However, under the alternative hypothesis of correlation, fixed effect estimates are consistent but random effect estimates are not (see Greene 2003: 301).

15 between individual effects and the other regressors which could distort the estimated coefficients. Since the low power of the Breusch-Pagan LM test in rejecting the null hypothesis of no correlation and the failure of the Hausman test to reject the hypothesis that the individual effects are uncorrelated with the other regressors when time effects are present, we opted to combine these two effects which is known as the mixed fixed and random effects (MFR) estimation in panel data analyses. However, we find that MFR estimates are very close to our fixed effect without time effects and random effect estimates, and it again suggests that the cross-sectional heterogeneity is not a significant issue in our sample. ============ Insert Table 3 ============ Since the time dimension is larger than the cross-sectional dimension in the data set, one may argue that our model may suffer from serial correlation and thus, previous estimates would be biased. In order to address this issue and to check whether our previous estimates are biased, we estimate the contemporaneous correlation of (1) using feasible GLS method. In the estimation, the following conditions are used: (i) industry panels are heteroskedastic but uncorrelated individually and the individual industry will have AR(1) type serial correlation, (ii) industry panels are heteroskedastic but correlated individually and individual industry will have AR(1) type serial correlation. Table 4 gives the estimated results of the contemporaneous correlation regression. The model selection criteria favour the model that allows for heteroskedastic panels with cross-section correlation while accounting for AR(1) serial correlation. The results of the FGLS estimates are very much similar to previous estimates of fixed effect without time effects and random effect models. Results again reveal

16 that the growth rate of manufacturing output will increase by 0.4 percent in response to 1 percent increase in the FDI ratio. 7 ===================== Insert Table 4 and Table 5 ===================== Nair-Reichert and Weinhold (2001), however, argue that contemporaneous correlation across the cross-section does not indeed imply the causation as these models suffer from endogeneity biases and the lack of proper instruments denies the satisfactory remedy to the issue. Endogeneity bias could be serious in our model as the joint-determination is possible not only between output and FDI but also between output and other variables such as employment and exports. That is, growth rates of employment and exports are dependent on the growth rate of output while they remain as causes of growth. A dynamic panel data model which has the ability to lag explanatory variables would be able to control for this issue of endogeneity. The lagged values of endogenous variables are used as instruments in explaining the current growth rate of output. Further, dynamic models will also be able to account for the omitted variable biases through the inclusion of lagged dependent variables. Table 5 gives estimates of a dynamic model of output growth as given in equation (3) using Arellano-Bond GMM estimation method. The co-linearity between FDI ratio and GCF ratio is also present in the dynamic model as well. Once the GCF ratio is dropped, the impact of lag FDI ratio on current growth rate of output turned out to be statistically significant and the magnitude of the effect is significantly less than that of the contemporaneous effect. The 7 We also estimated the contemporaneous correlation by the Prais-Winstein method which provides panel corrected standard errors and found no difference between the FGLS estimates. Although the Prais-Winstein regression results are not provided due to brevity reasons, they are available from authors upon request.

17 growth rate of manufacturing output increases by about 0.2 percent in response to 1 percent increase in lagged FDI ratio. The effect of lagged growth rate of employment is also dropped sharply and turned out to be only marginally significant. With two lags of the dependent variable, the impact of employment growth in the previous period becomes clearly insignificant. One may see this result as more meaningful since the current growth rate of output depend more on current growth rate of employment than on growth of employment in the past. But the absence of the lagged growth rate of employment may cause omitted variable biases in the model as implied by the sharp drop in the regression specification F test statistic. It is also observed that the effect of export growth in the previous period on output growth is much higher than in the contemporaneous effect. The manufacturing output growth rate increases by more than 0.4 percent in response to 1 percent growth of exports in the previous year. The result indeed supports the export-led growth hypothesis: the higher the foreign demand, the higher would be the output growth. The dynamic impact of expansion of firms is as same as in the contemporaneous impact. 5. Conclusions and Policy Implications Singapore is an example of a country where economic growth is largely driven by foreign investments. It relies on foreign investments to maintain its competitiveness and global trade through the access to foreign technology and network. In this paper, we examined the influence of foreign investment on output growth of 14 manufacturing industries of Singapore over the period of It is observed that high growth industries are associated with high absorption of foreign investments. Controlling for other growth enhancing factors such as employment, exports, material inputs, expansion of firms,

18 capital stock, we found that FDI inflows have significant positive impact on the growth of manufacturing industries in Singapore. However, recent evidence suggests that FDI inflows in the ASEAN region have declined after the Asian Crisis in Although we observe strong economic growth in the ASEAN region in recent years, FDI inflows are not converging to the same levels as to that of the pre-crisis level. There is a shift in FDI towards the North-east region of China, Korea and Taiwan. This shift in FDI activities indicate that there is a urgent need to re-align its investment and value-chain strategies to emphasis more value-added activities and to develop more indigenous technology to maintain the FDI activities in the Singapore and ASEAN region. This is the main challenge for the Singapore and ASEAN. The ability to align with the global activities is not just purely the activity of the firms, but also includes growing government investment in the improvement of physical infrastructure and education to develop human capital. Recent evidence suggests that the high flow of FDI into developed countries such as Europe and the US is mainly due to the strong fundamentals in technology, infrastructure and human capital (Lipsey and Feliciano, 2002; Balasubramanyam et. al., 1996). Further, study by Smarzynska and Wei (2002) highlights that strong institutions that clearly define the property rights that enables the efficient operations of the financial markets and with high intellectual property content tends to attract high quality foreign investments in knowledge and technology. High quality foreign direct investments have greater impact on output growth if the domestic capacity could complement the foreign technology of the multinational corporations.

19 While the initial emphasis was on labour intensive manufacturing, over the years the focus in the Singapore has shifted to encouraging inflows in higher value-added areas and skill-intensive manufacturing activities as well as knowledge-based professional services service sector activities such as financial services, ICT services and offshore services. Businesses are also encouraged to establish Research and Development (R&D) facilities in the City-State as well as to use the country as an international or regional headquarters. The emphasis of Singapore s FDI promotion has always been on developing key strategic clusters. Thus, the government s targeted policy helped develop chemical, electronics and engineering clusters, all of which became key economic engines for Singapore. More recently emphasis has been on product development, biomedical research, educational and health care services. Further studies on the impact of FDI on the output growth of Singapore economy could focus on these strategic industries in the future.

20 References Anwar, S Foreign Investment, Human Capital and Manufacturing Sector Growth in Singapore, Journal of Policy Modeling 30(3): Aitken, B. and A. Harrison Do Domestic Firms Benefit from Foreign Direct Investment: Evidence from Venezuela, American Economic Review 89(3): Arellano, M. and S. Bond Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, Review of Economic Studies 58(2): Balasubramanyam, V. N., M. Salisu and D. Sapsford Foreign Direct Investment and Growth in EP and IS Countries, Economic Journal 106(434): Basu, P., C. Chakraborty and D. Reagle Liberalization, FDI, and Growth in Developing Countries: A Panel Cointegration Approach, Economic Inquiry 41(3): Blomström, M., R.E. Lipsey and M. Zejan Is the Fixed Investment Key to Economic Growth?, Quarterly Journal of Economics 111(1): Branstetter, L. (2006). Is Foreign Direct Investment A Channel of Knowledge Spillovers? Evidence from Japan s FDI in the United States, Journal of International Economics 68(2): Carkovic, M. and R. Levine (2005). Does Foreign Direct Investment Accelerate Economic Growth, in Theodore H. Moran, Edward M Graham and Magnus Blomstorm (eds.) Does Foreign Direct Investment Promote Development. Washington DC: Institute for International Economics: Chang, P.L Trade, Foreign Direct Investment and Regional Competition: The Case of Singapore, in W.T.H. Koh and R.S. Mariano (eds.). The Economic Prospects of Singapore, Boston: Addison Wesley. Choe, J. I Do Foreign Direct Investment and Gross Domestic Investment Promote Economic Growth?, Review of Development Economics 7(1): De Mello, L.R Foreign Direct Investment in Developing Countries and Growth: A Selective Survey, Journal of Development Studies 34(1):1-34. De Mello, L.R Foreign Direct Investment-led Growth: Evidence from Time Series and Panel Data, Oxford Economic Papers 51(1):

21 Findlay, R Relative Backwardness, Direct Foreign Investment, and the Transfer of Technology: A Simple Dynamic Model, Quarterly Journal of Economics 92(1):1-16. Görg, H. and D. Greenaway Much Ado about Nothing: Do Domestic Firms Really Benefit from Foreign Direct Investment?, World Bank Research Observer 19(2): Greenaway, D., D. Sapsford and S. Pfaffenzeller Foreign Direct Investment, Economic Performance and Trade Liberalization, World Economy 30(2): Hansen, H. and J. Rand On the Causal Links Between FDI and Growth in Developing Countries, World Economy 29(1): Hu, A. G Multinational Corporations, Patenting, and Knowledge Flow: The Case of Singapore, Economic Development and Cultural Change 52(4): Kneller, R. and M. Pisu Industrial Linkages and Export Spillovers from FDI, World Economy 30(1): Konings, J The Effects of Foreign Direct Investment on Domestic Firms: Evidence from Firm Level Data in Emerging Economies, Economics of Transition 9(3): Lipsey, R. and Z. Feliciano Foreign Entry into US Manufacturing by Takeovers and the Establishment of New Firms, NBER Working Paper 9122, Cambridge, M.A. U.S. Low, L Singapore: Towards a Developed Status. Oxford: Oxford University Press. Nair-Reichert, U. and D. Weinhold Causality Tests for Cross-country Panels: A New Look at FDI and Economic Growth in Developing Countries. Oxford Bulletin of Economics and Statistics 63(2): Singapore Economic Development Board Annual Census of Manufacturing Activities. Singapore: Economic Development Board. Singapore Department of Statistics Foreign Equity Investment in Singapore. Singapore: Department of Statistics. Smarzynska, B. and S. J. Wei Corruption and Cross-Border Investment: Firm-Level Evidence, World Bank Working Paper, World Bank. Zhang, K. H Does Foreign Direct Investment Promote Economic Growth? Evidence from East Asia and Latin America. Contemporary Economic Policy 19(2):

22 % Output FDI Direct exports Employment Figure 1: Growth rates of Singapore manufacturing output and FDI inflows, Output growth rate Change in FDI ratio Figure 2: Scatter plot of FDI inflows and Singapore manufacturing growth

23 FDI ratio log Labour log Exports log Firms 0 50 GCF ratio Figure 3: Scatter plot matrix of regressors

24 Table 1 Singapore manufacturing growth by industries Industry Category Average Growth Rates Category Description FDI Output Emp Exports Food Food, beverage and tobacco Textiles Textiles, wearing apparel and leather products Wood Wood and wood products Paper Paper and paper products, printing and publishing Chemical Chemical and chemical products, pharmaceutical and biological products Petroleum Petroleum and petroleum products Rubber Rubber and plastic products Basic metals Basic metals Fab. metals Fabricated metal products Machinery Machinery and equipment Electronics Electronic products and components, electrical machinery and apparatus, Transport Transport machinery and equipment Instruments Instrumentation, photographic and optical goods Other Other manufacturing products Total Total manufacturing Note: Emp = number of persons employed, Exports = value of direct exports.

25 Table 2 Fixed Effect Regression Results Dep. Variable: log(y) No Time Effects Time Effects Variable (1) (2) (3) (1) (2) (3) (FDI/Y) (0.21) (0.21) 0.320* (0.16) 0.441* (0.24) 0.450* (0.24) 0.361** (0.16) log(emp) 0.403** (0.15) 0.405** (0.15) 0.404** (0.15) (0.18) (0.19) (0.18) log(exports) 0.128*** 0.128*** 0.128*** 0.241*** 0.240*** 0.242*** (0.04) log(firms) 0.123*** (0.03) log(matinputs) (0.001) (GCF/Y) (0.12) (0.04) 0.123*** (0.03) (0.11) (0.04) 0.123*** (0.03) (0.06) 0.101* (0.05) (0.001) (0.11) (0.05) 0.099* (0.05) (0.11) (0.06) 0.098* (0.05) R 2 overall Reg. F test Rho Observations Notes: indicates the first differenced series. The estimated constant term is not reported and the heteroskedastic robust standard errors are given in parenthesis. *, ** and *** stand for statistically different from zero at 10%, 5% and 1% significance levels.

26 Table 3 Random Effect Regression Dep. Variable: log(y) Variable (1) (2) (FDI/Y) 0.433** (0.20) 0.387*** (0.16) log(emp) 0.435*** (0.15) 0.439*** (0.15) log(exports) 0.131*** 0.133*** (0.04) log(firms) 0.118*** (0.03) (GCF/Y) (0.11) Constant 3.352*** (0.93) (0.04) 0.118*** (0.03) 3.538*** (0.83) R 2 overall Wald Chi Observations Note: Figures given in parenthesis are heteroskedastic robust standard errors of estimated coefficients and ** and *** stand for statistically different from zero at 5% and 1% significance levels. Table 4 Contemporaneous Correlation Feasible GLS Estimates Dep. Variable: log(y) Heteroskedastic panels with no cross-sectional correlation, Panel-specific AR(1) Heteroskedastic panels with cross-sectional correlation, Panel-specific AR(1) correlation Variable correlation (FDI/Y) 0.335*** (0.08) log(emp) 0.453*** (0.06) log(export) 0.149*** (0.02) log(firms) 0.122*** (0.01) Constant 2.999*** (0.66) Wald Chi Log likelihood SIC BIC Note: ***stands for statistically different from zero at the 1% significance levels *** (0.05) 0.429*** (0.05) 0.178*** (0.01) 0.123*** (0.01) 2.112*** (0.59)

27 Table 5 Dynamic Panel Data Regression: Arellano-Bond GMM Estimator Dependant Variable: log(y) Variable With One Lag of Dependent Variable With Two Lags of Dependent Variable (1) (2) (1) 3 (FDI/Y) (t-1) (0.17) 0.154* (0.08) (0.17) 0.163*** (0.05) log(emp) (t-1) 0.196* (0.11) 0.192* (0.11) (0.14) (0.14) log(export) (t-1) 0.415*** (0.06) 0.419*** (0.06) 0.436*** (0.06) 0.436*** (0.06) log(firms) (t-1) 0.120*** (0.02) 0.121*** (0.02) 0.125*** (0.02) 0.126*** (0.02) (GCF/Y) (t-1) (0.09) (0.09) log(y) (t-1) (0.03) (0.02) (0.04) (0.03) log(y) (t-2) (0.05) (0.05) Constant *** (0.07) *** (0.07) (0.12) (0.13) Wald Chi Observations Note: Heteroskedastic robust standard errors are given in parenthesis. *, ** and *** stand for statistically different from zero at 10%, 5% and 1% significance levels.