Note on the effect of FDI on export diversification in Central and Eastern Europe

Transcription

1 Note on the effect of FDI on export diversification in Central and Eastern Europe 1. Introduction Export diversification may be an important issue for developing countries for several reasons. First, a diversified bundle of export products provides a hedge towards price variations and shocks in specific product markets (Bertinelli et al., 2006; Levchenko and di Giovanni, 2006). Second, the type of products exported might affect economic growth and the potential for structural change (Hausmann et al., 2007; Hausmann and Klinger, 2006; Whang, 2006). Third, export diversification in the direction of more sophisticated products may be beneficial for economic development. Given these potential benefits of export diversification, an important policy question is what a country can do to diversify its exports. The purpose of this note is to examine whether foreign direct investment (FDI) can contribute to export diversification. Our preliminary analysis based on nine Central and Eastern European countries (CEECs) suggests a positive answer to this question. Firms undertake FDI in order to serve the market in the particular host country (market-seeking FDI) or to lower production costs (efficiency-seeking FDI). As in the latter case the output is not only meant for the host country, this kind of FDI should affect exports from the host country. If the foreign owned plant is producing different goods than other exporting firms in the host country, efficiency-seeking FDI affects the composition of the export bundle and may make it more diverse. However, if foreign-owned plants export only products that the host country already exports intensively, the efficiency-seeking FDI could lead to more specialized rather than more diversified exports. To empirically test the effect of FDI on export diversification we use data on nine CEECs: Bulgaria, Czech Republic, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia. During the past two decades, these nine countries have undergone a rapid transformation from being centrally-planned economies to following free-market principles and becoming members of the European Union. During the same time period they have also received large FDI inflows, aimed to a large extent at serving the European Union markets and they reoriented their exports from serving the former Soviet bloc to supplying mostly the European Union. This natural experiment makes CEECs a good setting for our study. In our empirical analysis, covering the period , we would like to investigate whether the level of export diversification increases more in sectors receiving considerable FDI inflows than in other sectors. Besides the lack of detailed sector-level information on FDI in developing countries, another obvious difficulty is that FDI inflows themselves can be affected by the level of export diversification in the host country. Therefore, rather than using information on actual FDI inflows we exploit the fact that investment promotion efforts in 3 countries in our sample have relied on sector targeting. Sector targeting, i.e. focusing FDI promotion efforts on particular sectors, is considered to be best practice by investment promotion professionals. Moreover, as shown by Harding and Javorcik (2007) there is robust evidence consistent with investment promotion efforts leading to higher FDI inflows. In their sample of almost 100 developing countries, Harding and Javorcik find that targeted sector receive on average more than twice as much FDI as non-targeted sectors or targeted sectors before targeting begins. Another benefit of using information on targeting rather than actual FDI flows is that the detailed sector-level data on the latter are not readily available. 1

2 2. Data and empirical approach We follow Hwang (2006) and measure export concentration using a Herfindahl index of export shares in sector i, in country c at time t: H = ( 2 s p p ) *100 s pc i is product p s share in the exports of industry i in country c at time t and is given by s = p x X p where x is the export value of p from country c sector i at time t and X is the total export value from country c sector i at time t. Our data on exports are from Feenstra et al. (2005) and the values of exports are measured in current US dollars. Sectors i are defined at the 3-digit level in the NAICS classification, which corresponds to 30 sectors (see Table 1). 1 H ranges from 0 to 100. In our empirical specification, we use (100 H ) to capture export diversification: (100 H ) = α + βsector_targeted + z π + Y θ + γ + γ + ε ct ci jt The indicator variable Sector_targeted equals one if investment promotion efforts country c target sector i at time t, and zero otherwise. This information comes from the 2005 Census of Investment Promotion Agencies described in Harding and Javorcik (2007). z is the total volume of exports of sector i in country c at time t and Y ct is a vector of time varying country specific variables. γ ci is country-industry fixed effects and γ jt is a sector-year fixed effect. The question of interest is whether targeted sectors, which tend to receive more FDI (assuming targeting works), are more diversified in their exports after targeting efforts start relative to other sectors or their own performance in the pre-targeting period. Note that time invariant characteristics that differentiate sectors chosen for targeting in a given country from other sectors (in the same or other countries) will be captured by country-sector fixed effects. These fixed effects will take out all unobserved, time-invariant characteristics specific to country-industry combinations. Shocks common to all sectors in a particular country in a particular year will be captured by time-varying country controls or (in a later specification) country-year fixed effects. Sector-time fixed effects will capture factors affecting worldwide shocks to exports (e.g. demand shocks) at a particular point in time. Three countries in our sample: Czech Republic, Lithuania and Slovenia have been engaged in targeting and provided detailed information on sectors targeted and start or stop (if applicable) dates of such efforts. 1 The trade data of Feenstra et al. (2005) are reported in the SITC Rev. 2 classification. We use NAICS (1997) as our sector classification. We used the concordance table between SITC Rev 2 and NAICS (1997) as found at 2

3 Table lists the sectors and the corresponding years. Some of the other countries in the sample reported sector targeting, but gave us incomplete timing. The targeted sectors with incomplete timing information are excluded from the estimation sample. We construct two different variables from the targeting information. The first one, called Sector targeting, is a dummy variable taking one in country-sector-years that were subject to targeting. The second, called Age of sector targeting, equals the number of years targeting has been in place in a particular country-sector-year combination. We include this variable in the log form (we add one before taking logs). Based on the findings of Harding and Javorcik (2007), which suggest that targeted sectors in developing countries receive double FDI inflows of non-targeted sectors (or targeted sectors in the pre-targeting period), we think of these two variables as proxies for higher FDI inflows. 3. Results Table 1 shows our baseline model of the effect of FDI on export diversification. As Herfindahl index measures the level of concentration, a high value of our dependent variable (100 - H) indicates a high level of export diversification. A significant positive coefficient on the sector targeting variable is interpreted as the presence of FDI positively affecting the level of export diversification. The coefficient of 2.8 in the first column of Table 3 suggests that sector targeting is associated with a 2.8 point increase in our export diversification index (which ranges from 0 to 100). This translates into targeted sectors having a 4.6 percent higher diversification (2.8/60 = 4.6 where 60 is the average level of the dependent variable in the sample) relative to non-targeted sectors or targeted sectors in the pre-targeted period. Note that when Sector targeting enters as a one- or two-period lag, the magnitude of the coefficient increases. It is, however, no longer significant in the case of a three-period lag. When we use the number of years the sector has been targeted (Age), we find that two extra years of targeting (relative to the mean value) are associated with an increase in export diversification of around 3 points, or about 4.5 percent. If investment promotion agencies take into account the level of export diversification when choosing sectors to be targeted, we may have a reverse causality problem. To examine this possibility, in Table 2 we include models with a dummy variable taking the value of 1 in the year (or two, three or four years) immediately preceding targeting and zero otherwise. Positive and statistically significant coefficients on these dummies would suggest that targeted sectors were already more diversified before targeting started. However, the coefficients on the dummies turn out to be insignificant in all four specifications. The results of the F-test suggest that in all but the first column, the coefficient on the targeting variable is significantly different from the coefficient on the pre-targeting dummy. The exercise gives us some confidence that FDI presence may be causing higher export diversification, rather than the other way around. Another challenge in our analysis is to distinguish the effect of FDI from other changes relevant for export diversification occurring at the same time. In an attempt to address this issue, we include country-year fixed effects which capture country-specific factors that may influence export diversification at a particular point in time. As the countries in our sample found themselves in a rather dramatic transition phase in the estimation period, the inclusion of countryyear dummies is important for the credibility of our results. The country-year dummies capture all changes due to transition from planning economies to market oriented economies to the extent 3

4 that they affected all sectors in the same way. The results, presented in Table 3, show that the augmented model leads to similar results. If anything, the effects of FDI on export diversification get somewhat larger. Missing in our specifications are measures of market access. During the period under study, the European Union opened its market to manufacturing imports from the countries in the sample. Unfortunately, at this point we do not have at our disposal time-varying tariff information for the full sample. However, the limited information available suggests a high correlation between sector-specific tariffs imposed by the EU on individual CEECs (the correlations vary from 0.5 to 0.99). Thus, sector-year fixed effects included in our model will to a large extent control for market access. 4. Conclusion The preliminary analysis included in this note is consistent with a positive relationship between FDI inflows and export diversification in the Central and Eastern European countries studied. In future work, we plan to extend the analysis to other developing countries and different measures of export outcomes. The alternative measures include indices of export sophistication proposed by Hausmann, Whang and Rodrik (2007), Hasmann et al. (2006) and Whang (2006). As FDI is typically undertaken by highly productive and sophisticated firms within their respective industries, it is plausible that FDI not only affects the degree of diversification of a host country s exports but also the degree of sophistication of the products exported. Given the positive effects of sophistication on economic growth and potential for structural change found in the papers mentioned above, FDI s effect on export sophistication appears as an important question for future research. 4

5 Table 1: Number of observations by sector NAICS97 NAICS97description Number of observations 111 Crop Production Animal Production Forestry and Logging Fishing, Hunting and Trapping Oil and Gas Extraction Mining (except Oil and Gas) Utilities Food Manufacturing Beverage and Tobacco Product Manufacturing Textile Mills Textile Product Mills Apparel Manufacturing Leather and Allied Product Manufacturing Wood Product Manufacturing Paper Manufacturing Printing and Related Support Activities Petroleum and Coal Products Manufacturing Chemical Manufacturing Plastics and Rubber Products Manufacturing Nonmetallic Mineral Product Manufacturing Primary Metal Manufacturing Fabricated Metal Product Manufacturing Machinery Manufacturing Computer and Electronic Product Manufacturing Electrical Equipment, Appliance, and Component Manufacturing Transportation Equipment Manufacturing Furniture and Related Product Manufacturing Water Transportation Motion Picture and Sound Recording Industries Professional, Scientific, and Technical Services 199 Total 5,726 Table 2: Sectors targeted NAICS97 NAICS97 description Total 315 Apparel Manufacturing Chemical Manufacturing Computer and Electronic Product Manufacturing Electrical Equipment, Appliance, and Component Manufacturing Fabricated Metal Product Manufacturing Furniture and Related Product Manufacturing Leather and Allied Product Manufacturing Machinery Manufacturing Nonmetallic Mineral Product Manufacturing Paper Manufacturing Petroleum and Coal Products Manufacturing Plastics and Rubber Products Manufacturing Primary Metal Manufacturing Professional, Scientific, and Technical Services Textile Mills Textile Product Mills Transportation Equipment Manufacturing Wood Product Manufacturing Total Note: Table shows the targeted sectors of the three countries in our sample (Czech Republic, Lithuania and Slovenia) for which we have full timing information. 5

6 Table 1: Specification with sector-year and country-sector fixed effects, sector exports and country level controls Sector targeting 2.762** [1.180] L. Sector targeting 3.195*** [1.205] L2. Sector targeting 2.900** [1.312] L3. Sector targeting [1.621] Age sector targeting 2.489*** [0.920] L. Age sector targeting 2.757*** [1.058] L2. Age sector targeting 2.770** [1.338] L3. Age sector targeting [1.967] L. Export value *** *** *** *** *** *** *** *** [4.018] [4.016] [4.018] [4.024] [4.015] [4.016] [4.021] [4.026] L. GDP per capita [1.506] [1.514] [1.507] [1.503] [1.509] [1.511] [1.505] [1.502] Population *** *** *** *** *** *** *** *** [20.874] [20.895] [20.853] [20.836] [20.846] [20.841] [20.828] [20.852] Inflation * * * * * * * [0.153] [0.152] [0.152] [0.152] [0.152] [0.152] [0.152] [0.152] Observations Number of group(code NAICS97) R-squared R-squared overall Note: Standard errors are reported in brackets. ***, **, * denotes significance at the 1, 5 and 10% level, respectively. The dependent variable is Herfindahl index of export shares. The estimation sample consists of sectors at the 3-digit aggregation level for the nine new EU member countries (Bulgaria, Czech Republic, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia) for the years LX means lagged X periods. Export value is at the sector level and measured in current USD. GDP per Capita is measured in current US dollars. Age of sector targeting, Export value, GDP per capita and Population are all in logs. Table 2: Specification with sector-year and country-sector fixed effects, sector exports and country level controls Sector targeting 3.224** 2.504* 2.874* [1.274] [1.419] [1.653] [2.056] 1 year before sect. targ [1.656] 1 and 2 years before sect. targ [1.366] 1, 2 and 3 years before sect. targ [1.349] 1, 2, 3 and 4 years before sect. targ [1.561] L. Export value *** *** *** *** [4.019] [4.025] [4.028] [4.023] L. GDP per capita [1.510] [1.513] [1.510] [1.506] Population *** *** *** *** [20.893] [20.881] [20.896] [20.984] Inflation [0.153] [0.153] [0.154] [0.153] Observations Number of group(code NAICS97) R-squared Test coeff F Test coeff p Note: Standard errors are reported in brackets. ***, **, * denotes significance at the 1, 5 and 10% level, respectively. The dependent variable is Herfindahl index of export shares. The estimation sample consists of sectors at the 3-digit aggregation level for the nine new EU member countries (Bulgaria, Czech Republic, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia) for the years LX means lagged X periods. Export value is at the sector level and measured in current USD. GDP per Capita is measured in current USD. Export value, GDP per capita and Population are all in logs. 6

7 Table 3: Specification with country-year and country-sector fixed effects, sector exports and country level controls Sector targeting [1.614] L. Sector targeting 3.188* [1.664] L2. Sector targeting 3.866** [1.829] L3. Sector targeting [2.276] Age sector targeting 2.255* [1.352] L. Age sector targeting 3.519** [1.547] L2. Age sector targeting 3.896** [1.931] L3. Age sector targeting [2.703] L. Export value ** ** ** ** ** ** ** ** [3.743] [3.736] [3.727] [3.727] [3.737] [3.729] [3.725] [3.727] Observations Number of group(code NAICS97) R-squared Note: Standard errors are reported in brackets. ***, **, * denotes significance at the 1, 5 and 10% level, respectively. The dependent variable is Herfindahl index of export shares. The estimation sample consists of sectors at the 3-digit aggregation level for the nine new EU member countries (Bulgaria, Czech Republic, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia) for the years LX means lagged X periods. Export value is at the sector level, measured in current USD and in logs. 7

8 References Bertinelli, Luisito, Veronique Salins and Eric Strobl, Export Diversification and Price Uncertainty in Sub-Saharan Africa and Other Developing Countries: A Portfolio Theory Approach, mimeo (Universite de Luxembourg, Universite de Paris X, Ecolé Polytechnique). Feenstra, Robert C., Robert E. Lipsey, Haiyan Deng, Alyson C. Ma, Hengyong Mo, World Trade Flows: , Working Paper 11040, National Bureau of Economic Research. Harding, Torfinn and Beata Smarzynska Javorcik, Developing economies and international investors: Do investment promotion agencies bring them together?, mimeo, World Bank, Washington D.C. Hwang, Jason, Introduction of New Goods, Convergence and Growth, mimeo, Department of Economics, Harvard University. Hausmann, Ricardo, Jason Hwang and Dani Rodrik, What You Export Matters, Journal of Economic Growth, forthcoming. Hausmann, Ricardo and Bailey Klinger, 2006a. Structural Transformation and Patterns of Comparative Advantage in the Product Space. CID Working Paper, Harvard University, 128. Hausmann, Richardo and Bailey Klinger, 2006b. South Africa s Export Predicament, Harvard University, CID Working Paper, 129. Levchenko, Andrei A. and Julian di Giovanni, "The Risk Content of Exports," mimeo, IMF, Washington D.C. 8