Estimation of Effective Rate of Protection
Chapter 5: The FDI-Growth Nexus in the Thai Manufacturing Sector
This chapter probes the FDI-growth nexus in the manufacturing sector. As mentioned in Chapter 2, while FDI has been widely recognized as a growth-enhancing factor in developing countries, its effect is not automatic but depends on trade policy regimes in host countries, as postulated by the ‘Bhagwati hypothesis’. The growth enhancing effect of FDI is likely to be far less, or even negative, under an IS regime compared to a policy regime geared to EP. To examine the growth-enhancing effect of FDI, the growth equation is estimated by applying the co-integration technique to time series data for the period 1970-2002. Three alternative indices of trade openness, (i.e. the trade to goods GDP, the export-gross output ratio in the manufacturing sector and the incidence of applied tariff rates in the manufacturing sector), are used in order to test the sensitivity of results on these indices.
The analysis in this chapter is expected to provide a broad indicator of the impact of FDI on Thai manufacturing. In addition, despite the immense policy relevance, so far only a few studies have undertaken to test the role of the trade policy regime empirically (e.g. Balasubramanyam et al., 1996; Athukorala and Chand, 2000). While these studies generally provide strong support for the hypothesis in the context of inter-country cross- sectional analysis, the analysis is subject to two caveats. Firstly, there are also vast differences among countries with respect to the nature and quality of data, which makes cross-country comparison a rather risky business. Secondly and more importantly, the cross-sectional approach cannot capture the dynamic effects of a shift from an IS regime towards an EP regime. With the failure of an IS regime, developing countries moved towards an EP regime with different speed. Such a dynamic aspect cannot be properly captured by cross-sectional analysis.
This chapter is organized as follows: Section 5.1 presents the empirical model used to examine the FDI-growth nexus. It is followed by a discussion of the data and the econometric method in Section 5.2. The results are presented and discussed in Section 5.3. The final section presents key inferences and policy implications.
5.1 The Model
The empirical model in this chapter involves estimating a growth equation derived in the context of the new growth theory, which provides for capturing the impact of FDI interactively with economic openness on the growth-enhancing effect. The starting point of model formulation is the aggregate production function of the manufacturing sector.
Y= f (A, L, K, H) (5.1)
where Y = Manufacturing output L = Manufacturing employment
K = Physical capital stock of the manufacturing sector H= Human capital stock of the manufacturing sector.
A = Total factor productivity (TFP) of the manufacturing sector
In equation 5.1, manufacturing output is a function of factor inputs consisting of labour (.L), and two types of capital —human and physical capital, denoted by H and K, respectively. The effect of technological changes is aggregated and represented by total factor productivity (A).
As argued in Chapter 2, FDI could directly affect output through increasing K as well as creating an impact on A in the host country. To capture the effect of FDI on manufacturing growth, K is composed of domestic and foreign physical capital stock, denoted by KD and KF, respectively. Firstly, FDI inflows increase KF and enlarge the capital stock. This leads to output expansion. In addition, since FDI is associated with
advanced technology, an increase in KF potentially enhances the technological capability o f the manufacturing sector and positively affects A. All other things being equal, this will also enhance output growth.
Nevertheless, the net growth impact o f FDI depends on the nature o f the trade policy regime. The key hypothesis is that the more open the trade regime, the greater the impact o f FDI on output growth. In an IS regime, FDI (as well as domestic investment) takes place in sectors predominately characterized by high capital intensity in production where the host country does not have a comparative advantage. An increase in FDI inflows could result in immiserizing growth. Moreover, FDI becomes an avenue for foreign companies to maintain a market share and reap extra profit from economic rent, created by the highly-protected domestic market. Such a regime also provides incentives for rent seeking and directly unproductive profit seeking (DUPS) activities. These seems to be less FDI technology spillover under an IS regime. Under this regime, FDI inflows are directed to industries where proprietary assets are important.1 This creates barriers to entry for local firms and thus constrains technology and efficiency spillovers. Moreover, the protection generated by an IS regime is likely to limit local competition, which is an important factor in stimulating firms to update new technologies in both production and management, and enhances their own productivity. Furthermore, it is more likely that the entry o f an MNE affiliate creates a ‘market-stealing’ effect that adversely affects the productivity o f local firms.
In contrast, the main incentives for FDI under an EP regime in a given host country are the relatively low-labour costs and/or the availability o f raw materials. This allows foreign investors to operate in an environment that is relatively free from distortion, and leads to output expansion in internationally competitive and export- oriented product lines. Hence, FDI inflows are unlikely to result in immiserizing growth. Moreover, the production o f firms in an EP regime is not limited by the size o f the domestic market and has the potential to reap economies o f scale through international
'Proprietary assets are defined as those that can differentiate productivity between firms that own them and those that do not. They can generate profit; other firms cannot quickly or effectively imitate (Caves, 1996).
market penetration. Besides, an EP regime is more conducive to generating favourable spillover effects because FDI is mostly attracted to industries in which the country has comparative advantage. In such industries, there is a greater potential for local firms to catch up with foreign firms and achieve productivity improvement. This generates healthy competition among foreign and local firms that encourages them to keep maintaining their competitiveness in subsequent following periods. All in all, it positively affects output growth.
Thus, in order to capture the total impact of FDI on output growth, both KF and the interaction term between KF and a proxy variable for the openness of trade policy regimes (TP) are incorporated in the equation.
Y = F (L, H, Kd, Kf , Kf *TP) (5.2)
The impact of FDI on output growth depends on both KF as well as TP.
Since reliable data series on domestic and foreign capital stocks in the manufacturing sector are not available for Thailand, the ratio of gross fixed capital formation in the manufacturing sector, net of FDI (GFCFn) to GDP, is employed to represent KD in this study. This proxy variable has been used in numerous previous studies (e.g. Barro, 1999; Balasubramanyam et al., 1996). Similarly, KF is proxied by the ratio of manufacturing FDI inflow to manufacturing output. Owing to the lack of an appropriate direct measure of human capital stock, H is proxied by public education and research expenditure as a ratio of gross national income (GNI), as has been done in several empirical studies (e.g. McMahon, 1998; Sylwester, 2000). While there are alternative measures such as primary or secondary school enrolment ratios, the choice made here is constrained by data availability for the period under study., i.e. 1970-2002. In addition, since the size of the public sector in Thailand has remained more or less the
same around 17 per cent o f GDP over the same period, the ratio o f public education and research expenditure would be a reasonable proxy o f human capital development in Thailand.
As discussed in Chapter 3, there is no unique measure o f the openness o f the trade policy regime. This study uses three alternative proxies: (a) the ratio o f total merchandise trade (import + export) to goods GDP, which is total GDP net o f value added in construction and services sectors, (OPENl)\ (b) the ratio o f export to gross output in the manufacturing sector (OPEN2)\ and (c) the ratio o f incidence o f applied tariff rates, the proportion o f total tariff revenue to total imports o f the manufacturing sector (OPEN3). These three alternatives are introduced to examine the sensitivity of results due to the proxies for the trade policy regime.
The first measure is superior to the widely-used trade to GDP ratio, i.e. degree o f openness because the inclusion o f non-traded activities (construction and services) as part o f the denominator could lead to an under-estimation o f exposure to foreign trade o f the given economy (Rivera-Batiz and Rivera-Batiz, 1994). This point is particularly relevant in Thailand where construction and financial services recorded rapid growth during the latter part (from the late 1980s) o f the period under study. The second measure is based on the premise that greater openness is a prerequisite for successful world market penetration in manufactured goods. In other words, export success in manufacturing is likely to occur under a policy regime where policies are more neutral and allow the market mechanism to effectively indicate the country’s comparative advantage (Edwards, 1993). The third openness measure is simply the weighted average actual tariff in the manufacturing sector. The higher the tariffs, the lower the degree o f openness.
Therefore, the estimating equation used in the empirical analysis is
2 The size of public sector is measured here by the per cent of the sum of public consumption and investment to GDP. There is no significant variation between the 1970s, 1980s and 1990s.
3 For a succinct discussion of various measures of openness and a detailed listing of related references, see Edwards (1998).
Yt = a + ß ]Lt + ß 2KDt + ß,H, + ß AKF, + ß 5(KFl * 77>) (5.3)
where Y = Manufacturing output ( in log form)
L = (+) Labour force in the manufacturing sector (in log form) Kd = (+) Gross fixed capital formation (GFCFn) net of FDI of
the manufacturing sector as a percentage of manufacturing output
H = (+) Public education and research expenditure as a percentage of gross national income (GNI)
Kf = (+/-) Foreign direct investment in the manufacturing sector as a percentage of manufacturing output
TP = Openness of the trade policy regime, proxied alternatively by (+) (1) OPEN1 = Ratio of total merchandise trade to goods GDP (+) (2) OPEN2 = Export-gross output ratio in the manufacturing sector (-) (3) OPEN3 — Incidence of applied tariff rates of the
t = Time subscript.
s = Stochastic error term
The sign expected for the regression coefficient is given in brackets.
The coefficients ßi, ß2 and represent output elasticity with respect to labour, (domestic) physical capital, and human capital, respectively. Hence, they all are expected to be positive. The impact of FDI on growth (Y) is given by the partial derivative of Y in (5.4) with respect to KF, i.e.
dKF ß t + ß s * TP (5.4)
To test the relevance of the hypothesis, the statistical significance of ß 5 is examined. Under the Bhagwati hypothesis, the sign of ß 5 is expected to be positive for
will be an increasing function of OPEN1 and OPEN2, and a decreasing function of
OPEN3. The sign of ß A is ambiguous and can be positive or negative, depending on the
nature of the trade policy bias over the entire sample period. Even when ß 4 is negative, it does not imply that the FDI contribution is negative. Whether its contribution is negative or not depends on the size of the coefficient of the interactive term of FDI and the trade policy regime,ß 5, compared to ß4.
The model is estimated using annual data for the period 1970-2002. The full data set are reported in Appendix 10. Data on manufacturing output and gross fixed capital formation (GFCF) are obtained from the National Income Accounts, National Economic and Social Development Board (NESDB) of Thailand. These data are in real terms at 1988 prices. FDI inflows, exchange rates, tariff revenue, and international trade are from
the Bank o f Thailand Quarterly Bulletin, the Bank of Thailand (BOT). Data series of
FDI inflows are deflated to in terms of 1988 prices by investment price deflators.
The data on the work force come from the Key Indicators o f Developing Asian
and Pacific Countries, the Asian Development Bank (ADB). The OPEN1 variable, the
ratio of total merchandise trade to goods GDP, and the percentage of public education and research expenditure to Gross National Income (GNI) are obtained from World
Development Indicators, The World Bank.
The statistical summary and correlation matrix of these variables are given in Table 5.1 and 5.2, respectively. Output and FDI seem to exhibit a high correlation in the manufacturing sector. The correlation coefficient between manufacturing output and the share of manufacturing FDI is 0.64. Nevertheless, when manufacturing output and FDI are plotted together in Figure 5.1, it clearly indicates the correlation between output and
FDI is likely to be increasing as the trade policy regime in Thailand become more liberal, as postulated by the Bhagwati hypothesis.
Statistical Summary of Data used in the Regression Analysis Measurement units
Mean Min Max SD
Manufacturing value added (Y) (log) million baht
12.8 11.3 14.0 0.9
Manufacturing employment (L) (log) 1,000 workers
7.8 6.5 8.5 0.6
FDI as a percentage of Y ( K F) [log( 1+proportion)]
0.03 0.008 0.076 0.02
Gross domestic capital formation net of FDI as a percentage of Y ( K d )
0.126 0.012 0.199 0.052 Public education and research
expenditure as a percentage of Gross National Income (H)
3.0 2.3 3.5 0.4
Merchandise trade to goods GDP (O P E N 1 )
114.9 62.5 212.0 43.9
Export-output ratio (O P E N 2) per cent
20.7 0.9 57.0 18.4
1 Incidence of applied tariff in manufacturing (O P E N 3 )
12.9 4.5 20.6 4.9
Notes: Mean = simple average; Min = minimum; Max = maximum; SD = standard deviation
Source: Author’s calculation; see full data set in Appendix 10
Correlation Matrix of Data used in the Regression Analysis
Y L K F K D H O P E N 1 O P E N 2 O P E N 3 Y 1.00 L 0.97 1.00 K F 0.64 0.59 1.00 r D -0.15 -0.16 -0.42 1.00 H 0.87 0.86 0.53 -0.31 1.00 O P E N 1 0.92 0.88 0.79 -0.42 0.79 1.00 O P E N 2 0.91 0.86 0.74 -0.42 0.83 0.95 1.00 O P E N 3 -0.94 -0.92 -0.72 0.40 -0.86 -0.95 -0.93 1.00
Note: see variable notation in Table 5.1