EU 6. Framework Programme This project is funded by the European Commission, Research Directorate General as part of the 6th Framework Programme, Priority 8, "Policy Support and Anticipating Scientific and Technological Needs". Vertical Production Networks: Evidence from France Michel Fouquin, Laurence Nayman, Laurent Wagner Fourth meeting Brussells, 16 March 2006

EU 6. Framework Programme Contents What is it about? The Empirical model Data Estimation Issues Results

What is it about? The rapid development of FDI causes concern over the decline of industrial activities in developed countries and their relocation in emerging economies. Horizontal FDI was viewed as the main channel through which this phenomenon occurred on account of higher trade and wage costs. Nonetheless, vertical FDI expands rapidly and trade in intermediate inputs amounts to 1/3 of world trade. Multinational firms (MNF) are the major actors of trade. A substantial part of exports of US parent companies to their manufacturing affiliates are inputs for further processing against about 40% for French ones.

Share of US exports for further processing in: Source: Borga and Zeile (2004).

The Empirical Model Intra-firm exports of int. inputs would increase with lower trade and wage costs and some policy characteristics of the host country, so argue Hanson, Mataloni & Slaughter (2005). There is then a continuum of production processes from vertical to horizontal FDI, depending on the elasticities of intra-firm exports to trade and wage costs. To test this, we will focus on three flows: - the affiliate imports more or less intermediate inputs and services from the parent company, - the affiliate re-exports the finished product to its parent company, - the affiliate takes advantage of the market size of the host country to sell finished products at home or in the area.

The Empirical Model Objective: Tracing the determinants of intra-firm trade in intermediate inputs for French multinationals, using firm-level data. The estimated equation is: S ipcm = ip + 1 ln Wage c + 3 ln LP c + 3 ln Distance c + 3 ln (1-Corp Tax Rate c )+ 4 ln (1+Tariffs cm ) + X pc + e aipcm The imported inputs in the affiliate total costs are measured by the share of imported inputs for further processing in the affiliates turnover, Sipcm, where i stands for the sector, p the parent company, c the country, and m the imported input for further processing.

The Empirical Model All exogenous variables are in logarithms. Wage stands for the wage per employee in the manufacturing sector of the host country. LP is value added per employee. Distance is a proxy for transportation costs and measures the distance separating France from the host country. X is a vector of miscellaneous control variables This specification also includes a set of parent x industry dummy variables, αip. Equation is estimated by GEE

Data A merge of two surveys: - one from the Ministry of Industry on MNF located in France - one from the Ministry of Economy on affiliates of French firms abroad. Due to missing crossed information, some assumptions had to be made: a single super-affiliate in the host country regrouping several affiliates purchasing different products from their parent company.

Estimation Issues In HMS (2005), it can be read: We assume that U.S parent firms have previously chosen in which countries to locate affiliates. The remaining decision is over which production activities affiliates should perform. We do believe that the analysis of production activities requires, in the first place, to account for unobservable characteristics inherited from location decisions.

Estimation Issues By implementing a cluster analysis, on the whole set of controls used so far, we obtain two main clusters. The opposition between developed and developing through this clustering is unchallenging. We decided to split up our sample in two by considering Spain PPP GDP per capita as the cut-off point between developed and developing countries, since it matches our cluster mapping.

Results Wages : Wages per employee impact positively on the import to turnover ratio for the developed countries sample, and negatively for the developing countries one. Cross-price elasticity of imported demand to wages=1.9 for the developed countries and -0.2 for the developing ones. Wages per employee mirror the average skill intensity of jobs in a country: the higher wages are, the higher the average skill-content of jobs in the country. The less-skilled labour intensive affiliates are located in countries where wage costs are relatively lower, and conversely for the developed countries.

Results Labour productivity : Labour productivity matters a lot in explaining patterns of intra-firm trade. Real unit wages, the difference in the logarithms of wages per employee and labour productivity corroborate the dichotomy between developed and developing economies. It results, for the developed countries, into a negative coefficient on unit wage costs. Conversely, in the developing countries, the coefficient is rather low but positive.

Results Freight Costs : The impact is negative for both developed and developing countries. These results match those of HMS. Transport costs have a negative impact on imported input flows, and this supports the vertical production scheme of MNCs.

Results Tariffs : The strong presence of EU-15 countries does not allow the real effect of tariffs on exports of intermediates to be disentangled in the developed countries sample but at least it is negative, meeting HMS results. For the developing countries sample, only when controlling for the identity of the parent company, is the impact strong and positive (tax rebates granted to these firms?)

Results Corporate tax rate : No real evidence of tax rates effects could be found. Institutional quality: An improvement in economic conditions leads to higher exports of parent companies to their related parties located in developing countries.

Results Partnership trade : Trade performed by the MNCs with non related subcontractors, and intra-firm trade appear to be substitutes. Partnership trade occurs more often with European countries Affiliate operations can be completed by a back-up of production by non-related parties if need be, or else, production can be separately carried out taking advantage of specific assets in each location.

Results Market Potential : The market potential variable turns out to be negative and significant in both samples. This supports the idea that affiliates do not sell their processed production in their own markets to take advantage of the market size.

Results Alternative Sample Splitting : 40% of export flows are matched by imports of parent companies from their affiliates. Hence, parent companies may export intermediate products and import assembled goods. The whole Sample is split twice according to whether products are reshipped to France or not and with respect to the capital to labour intensity of imported inputs. We used capital deepening from the EUKLEMS database for France. Results point to a mix of vertical and horizontal trade.