Industry Characteristics and FDI Induced Technology Spillovers[◆]
Akinori Tomohara[*] Kazuhiko Yokota
University of CaliforniaLos AngelesThe International Center for the Study of
East Asian Development in Cooperation
Abstract
While the literature has explored the relationship between FDI and productivity, a consensus has yet to be reached regarding FDI’s impacts on the productivity of local companies, specifically with respect to vertical spillovers. Motivated by various results in the literature, this paper specifies the conditionsunder which industries enjoy horizontal, backward, or forward technology spillovers. Our analysis extends upon previous works andsheds light on the necessity of distinguishing industry characteristics when discussing potential benefits from FDI. The results of the analysis show that export-oriented sectors enjoy a higher degree of technology spillover than domestic-market oriented sectors do. Taken together with historical evidence, the results imply that a host government might benefit by attracting FDI into export-oriented industries toattain the goal of long-run economic growth.
JEL classification: F2; O1; O3
Keywords: Foreign Direct Investment (FDI); Productivity; TechnologySpillovers; Vertical Linkages
1. Introduction
Does Foreign Direct Investment(FDI) really assisthost countries in the economic development process? International organizations advocate access to the global economy via foreign direct investment, specifically for developing countries. Anti-globalization movements do not necessarily agree that foreign direct investment positively influences host countries. Self-interested, multinational companies may exploit a host country’s resources,impairingsubsequent development. Industrial policy regarding FDI is one of the major policy debates faced by the World Bank and IMF today.
The literature has explored if FDI benefits localestablishments in host countries via increased productivity (see Görg and Strobl (2001) for a survey). Caves (1974) on Australia, Globerman (1979) on Canada, and Blomström and Persson (1983) on Mexico are seminal empirical studies. Using cross-sectional analysis Kokko (1994) examinesMexico and Blomström and Sjöholm (1999) examineIndonesia. Others, including Haddad and Harrison (1993) on Moroccoand Aitken and Harrison (1999) on Venezuela,employ firm-level panel data analysis. Previous empirical studies primarily examined whether the presence of multinational companies affects the productivity of local companies operating in the same sector(i.e., horizontal spillovers). However, a number of theoretical works also discuss FDI’s technology spillovers via vertical linkages (Lin & Saggi, 2007; Markusen Venables, 1999;Rodoriguez-Clare, 1996). More recent empirical literature incorporatesvertical linkages across industries,(together with endogenous input decision-makingand establishment heterogeneity)(Blalock & Gertler, 2008; Javorcik, 2004; Kugler 2006; Lileeva, 2008;).[1]In spite of the multitude of studies conducted, the literature has yet to reach a consensus regarding the impacts of FDI on domestic companies’ productivity, specifically with respect to vertical technology spillovers.The story is more complicated if we decompose vertical spillovers into backward and forward ones.[2]
This research is motivated by various results about vertical spillovers in the late 2000s literature. Previous empirical works study technology spillovers using multiple countries with different characteristics. Policy planners wonder which results hold when the analysis is applied to a single country of interest. Wetry to identifyconditions under which FDI benefits localestablishments in host countriesusing Thailand as a case study. Thailandhas experienced success under a policy ofFDI-led growth, making the country relevant for analyzing the impact of FDI ontechnology spillovers. Our work confirmsthe results of studies that relied on data from multiple countries.
Our analysis presents the possibility that FDI will improve the welfare of host countries through increased productivity of local companies. However, the analysis underscores the need to distinguish industry characteristics when discussing FDI benefits. Export-oriented sectors such as office, accounting and computing machinery, radio, television, and communication equipmentenjoy forward spillovers. Horizontal and backward spillovers are observed in non-export-oriented(or domestic-market oriented) sectors such as food, paper, and other non-metallic mineral products. Spillover effects vary dependingon characteristics of potential beneficiaries in host countries.
The present exposition is distinctfrom previous works in that we examine whether the existence of spillovers depends on characteristics of potential beneficiaries. The results show a co-existence between horizontal and backward spillovers, but forward spillovers tend to occur in isolation. The results are intuitive because the relationships with upstream and downstream sectors vary across industries. The literature during the 2000s shows country-wide spillover effects averaged with industry weight. We identify the industry characteristics that lend themselves to the presence of spillovers.
Results of our analysis suggest that industrial policies inviting FDI to export-oriented sectorsare effective engines of economic growth when our results are related to the infant industry argument of import-substituting industrialization. Historically, we observe that countries that have operated under policies of export promotion attained rapid economic growth. However, countries that have operated under import substitution developed sluggish economies (Akinlo, 2004; Krugman & Obstfeld, 2005). We conclude that policies encouraging the presence of foreign-owned multinational companies arepotential catalysts for economic growth in host countries if foreign capital is invested properly in export-oriented sectors.
The paper proceeds as follows. Section 2 summarizes the data used for the analysis. Section 3 describes the empirical model. Results of the analysis are presented in Section 4. Section 5 concludes the paper and suggests future lines of research.
2. Data
We use an establishment-level panel dataset from an industrial survey conducted by the National Statistical Office (NSO) of Thailand between 1999 and 2003. The NSO staff interviewed the owners of manufacturing establishments with 10 persons or more which were selected using a combination of stratified sampling and systematic sampling. The NSO stratified establishments in each province according to industry codes and the number of workers. Then samples were selected from each province-industry-worker stratum using systematic sampling. The samples cover nearly halfof theestablishments with 10 persons or more operating in Thailandandare thusrepresentative of Thai companiesfrom various industries and of different sizes. The survey provides information on ownership(whether an establishment has foreign capital), sales, labor(the number of employees), capital(book value of fixed assets), material and electricity costs, an export-output ratio, an imported material ratio, location(province), and industrial classification. The survey provides information on the prior years date (e.g., the 2001 survey provides 2000 data).
The analysis examines manufacturingin Thailand from 1998-2002. FDI inflowsincreased rapidly during thisperiod (see Figure 1), making this specific time period particularly relevant for analyzing technology spillovers. Thailandhasserved as a host country to FDI since the1960s. FDI inflows into Thailandbegan to increase dramatically in 1988. Average net FDI inflows were6.6 billion Baht (US$276 million) per year during 1980-87, but increased to47.5 billion Baht (US$1.9 billion) during 1988-95. It was during this time that the government shifted its trade policy from one of import substitution, as was typical in the 1960s and 1970s, toone of export promotion,which prevailed throughout the 1980s. Correspondingly, the economic growth rate increased from 5.9percent(the average rate between 1980 and 1987) to 9.1percent(the average rate between 1988 and 1995). Thailand experienced another large increase in FDI after the Asian Financial Crisis. Average net FDI inflows were 166.4 billion Baht (US$ 4.0 billion) per year during 1998-2002.[3]Thailand’s success under FDI-led growth policy makes it an ideal country to use in studying FDI induced technology spillovers.
The Bank of Thailand provides various price index data.[4]We deflate total sales using a producer price index (PPI) by product group, capital stock by a PPI of capital equipmentsat the stage of processing, intermediate inputs by a PPI of intermediate materials by product group, and electricity expenditures by a consumer price index. All variables are measured in 2000 Thai Bahts. Depreciation rates of capital are obtained from the Office of the National Economic and Social Development Board, Office of the Prime Minister(NESDB), Capital Stock of Thailand, 2004.[5] The analysis uses input-output tables in order to relate industries in upstream and downstream sectors. We obtain 1998 and 2000 input-output tables and relevant information from the NESDB. The original input-output (I-O) tables have 180 industry sectors. Among the 180 sectors, 90 sectors are manufacturingrelated. We classified the 90 manufacturing sectors into 22 sectors based on ISIC 2-digit codes(ISIC 15-36). We drop sector 37 (recycling)from the analysessince recycling does not appear in the I-O tables. Backward and forward spillovermeasures are calculatedusing the1998 I-O data to compile the 1998 and 1999 datasets, and2000 I-O data for the 2000 and 2002 datasets.
Table 1 presents the sample’s summary statistics. We have 24,248 observations (about 6,000 plants in each year) after eliminating outliers and establishments with missing variables. The sample includes 22 industries at the 2-digit ISIC level. A comparison ofour sample to Javorcik’s (2004) study of Lithuania, a particularly influential study in the area,yields some notable differences. FDI trends differ by sector (Table 2). In Lithuania, the food and textile sectors attract a large share FDI. In Thailand, the majority of FDI flows into the computing and electronic machinery and radio and television sectors. Our data also indicate that the presence of foreign affiliates within the same industry is more common in Thailand than in Lithuania.[6]Similarly,local establishments in Thailand sold more of their goods to foreign affiliates in downstream industries relative to local establishments inLithuania.[7] Local establishments in Thailand have a closer relationship with foreign affiliates in upstream sectors relative to local establishment in Lithuania.[8]
3. Model
We use the following model to examine the impacts of FDI on localestablishments’ productivity:
. (1)
Output,, is the real output of establishment in industry sector in region at time . Output is calculated by deducting from sales changes in inventories of finished goodsand taxes. The first of the four input variables iscapital, , measured as the value of fixed assets at the beginning of the year. The second input,labor, , is the number of workers. Materials, ,is the value of material inputs and, is establishment’s electricity expenses. measures the potential forintra-industry spillovers. Specifically, the average foreign presence in sector at time is calculated by using the weight of foreign establishments’ output to total output in the sector. The weight captures the magnitude of foreign establishments’ effects on local establishments in the same sector:
.
measures the relationship between localestablishments in sector and foreign establishments in downstream sectors that buy intermediate goods from the industry sector:
,
where is the share of sector ’s output supplied to sector . This measure excludes goods supplied for final consumption, imports of intermediate goods, and inputs supplied within the sector. measures the relationship between localestablishments in sector and foreign establishments in upstream sectors that sell intermediate goods to the industry sector:
.
In the equation above, is the share of inputs that industry bought from industry among sector ’s total input purchases. Inputs purchased within the sector are not included.
The coefficients of interest are, , and . If these coefficients are positive and significant, it can be taken as evidence that a closer linkage between local and foreign establishments contributes to an increase in the productivity of local establishments. This would be evidence of technology spillovers from foreign establishments to local establishments.
FDI affects local establishments’ productivity through two different channels. The first is knowledge spillovers. Domestic companies learn how to employ superior technologies already used by foreign establishments. The second is an efficiency improvement via structural changes in the market. The entry of multinational companies will cause more competition in the host country, which may induce local establishments to operate more efficiently. The literature shows that market competition is positively correlated with productivity (Nickell, 1996). We include the Herfindahl index, , which measures industry concentrationin order to separate the effects of changes in the market structure from knowledge spillovers. The index is calculated as the sum of squared market shares of the four largest producers in a given sector.
Additional terms attempt to capture unobservable factors that may influence output levels. Year fixed-effects, , are time varying elements that affect all regions and industries in a given year. Regional fixed-effects, , are time and sector invariant elements that differ across regions. For example, higher quality infrastructure in a particular region would be controlled for with a regional fixed effect. Industry fixed-effects, , capture time and region invariant elements that differ across industries. Our analysis controls for time invariant unobserved characteristics, , using fixed-effects at the establishment level because omitted establishment heterogeneity such as management quality and financial conditions may affect productivity.
Estimating equation (1) assumingexogenous production inputscauses a simultaneity problem. Decisions regarding input usage are endogenous if the levels of inputs used vary with establishment-specific characteristics. Establishments may use more inputs if establishments experience positive productivity shocks. Previous works often use the semiparametric estimation procedure proposed by Olley Pakes (1996) in order to handle the simultaneity problem. However, Olley and Pakes’ method is only applicable to establishments reporting non-zero-investmentbecause they use investment to control for correlation between input levels and unobserved firm-specific productivity shocks in estimating the parameters of the production functions. Unfortunately, nearly two-thirds of the establishments report zero investment in our sample. Thus, our analysis uses semiparametric estimation similar to that of Levinsohn Petrin (2003). Levinsohn Petrin (2003) propose an alternative methodusing an intermediate input such as electricity to address the simultaneity problem. From the production functionestimated, we calculate a measure of total factor productivity (TFP) as the difference between the actual output and predicted output:, where all terms are measured in logarithm units, and regress the TFP measure on the spillover variables in Equation (1).[9]
We further extend our analysis by distinguishing several establishment characteristics. The analysis examines whether there are any different vertical linkage effects depending on export orientation, imported materials, age, and size. For example, export oriented establishments may be required to satisfy higher quality standards than domestic market oriented establishments. Increased competitive pressures may affect technology spillovers. We stratify the sample into two sub-samples: export oriented establishments and non-export oriented establishments. We classify establishments as being export oriented if an establishment exports at least 70 percent of its products. Otherwise, the establishment is classified as being non-export oriented.[10] Similarly, we stratify the sample into two sub-samples: establishments with a high imported materials ratio and establishmentswith a low imported materials ratio. We classify establishments as being establishments with a high imported materialsratio if an establishment imports at least 70 percent of its intermediate materials and otherwise classified as being establishments with a low imported materialsratio. Additionally, we stratify the sample into two sub-samples: old establishments (operative years are at least 10 years) and young establishment (otherwise); large size establishments (more than 27 employees) and small size establishments (otherwise). The threshold levels are the median values of the establishment’s years in existence and the establishment’snumber of employees.
4. Results of the Analysis
The results of the analysis are summarized in Table 3. For each table, column (1) is the results obtained when estimation uses the whole sample. Columns (2)-(9) present the results using sub-samples, where stratification of the sample is based on local establishment characteristics.The analysis corrects standard errors for clustering within industry-year cells. We study the effects of aggregate variables (the time-variant, sector specific horizontal and vertical index variables) on micro units (the real output of the individual establishment). Previous works show that analysis without correcting for correlation among observations within the same group understates standard errors of coefficient estimates and, thus, leads to overestimated t-statistics (e.g., Moulton, 1990).
The results show some interesting patterns regarding spillovers. There are three groups where each group has the same signs on the estimated coefficients regarding horizontal and vertical spillovers. Non-export oriented establishments have the same estimated coefficient signs as establishments with a low import material ratio. Positive backward spillovers are observed only in the two types of establishments. Similarly, old establishments have the same estimated coefficient signs as small establishments. These establishments enjoy positive horizontal spillovers. Additionally, positive forward spillovers are observed among export oriented establishments with a high import material ratio as well as young and large size establishments. Observations within each group do not necessarily overlap. How do we interpret these results?
The analysis indicates that localestablishments could benefit from FDI depending industry characteristicsbecause establishment characteristics turn out to be related to industrial classifications. Using Table 2, we match establishment characteristics to corresponding industrial sectors. Industries such as food products and beverages, tobacco products, and publishing, printing and reproduction of recorded media include relatively old and small size establishments. These industries comprise what we will call the “Industry H” group. Industry H plus industries such as paper and paper products and other non-metallic mineral products include relatively non-export oriented establishments and establishments with a low import material ratio. Industry H plus the additionally mentioned industries comprise the“Industry B” group. Industry H and Industry B are composed of similar industrialsectors. Industries such as wearing apparel, tanning and dressing of leather, luggage handbags, saddlery, harness and footwear, office, accounting and computing machinery, radio, television and communication equipment and apparatuses, and furniture include establishments that tend to be export oriented and relatively young and large size. These are denoted as the“Industry F” group.