WPS6330 Policy Research Working Paper 6330 Chinese Trade Reforms, Market Access and Foreign Competition The Patterns of French Exporters Maria Bas Pamela Bombarda The World Bank Development Economics Vice Presidency Partnerships, Capacity Building Unit January 2013 Policy Research Working Paper 6330 Abstract A unilateral trade reform generates two opposite vis Asian liberalization. The findings suggest that lower effects: market access expansion and strengthening of Chinese import tariffs account on average for 7 percent competitive pressures in the liberalized market. Using of the new products exported by French firms, and for 18 detailed trade and firm-level data from France, the percent of additional French export sales. These results authors investigate how French firms’ product scope and are robust when accounting for foreign competition faced export sales changed after Chinese liberalization vis-à- by French firms in the liberalized market. This paper is a product of the Partnerships, Capacity Building Unit, Development Economics Vice Presidency. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at maria.bas@cepii.fr and pamela.bombarda@u-cergy.fr. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Chinese Trade Reforms, Market Access and Foreign Competition: the Patterns of French Exporters Maria Bas and Pamela Bombarda JEL classiï¬?cation: F12, F13, L11. Keywords: unilateral trade liberalization, market access, foreign competition, export margins and ï¬?rm level data. Sector Board: Economic Policy (EPOL) Maria Bas (Corresponding author) is an economist at the Centre d’Etudes Prospectives et d’Informations Internationales e de Cergy-Pontoise and (CEPII); her email is maria.bas@cepii.fr. Pamela Bombarda is an assistant professor at the Universit´ THEMA; her email is pamela.bombarda@u-cergy.fr. The authors thank Agn` es Benassy-Qu´ e, Antoine Berthou, Sebastien er´ Jean, Lionel Fontagn´ e, Elisa Gamberoni, Thierry Mayer, Cristina Mitaritonna, Sandra Poncet and Elisabeth Sadoulet. We also thank two anonymous referees for their valuable comments to the manuscript and their constructive suggestions. 1 INTRODUCTION Unilateral trade liberalization is at the core of economic reform packages implemented in several emerging economies in the last decades.1 Microeconomic effects of trade reform episodes have received a lot of attention recently. Empirical works have concentrated on how trade openness shapes ï¬?rms’ productivity (Pavcnik (2002), Trefler (2004), Amiti and Konings(2007), Lileeva and Trefler (2010) and Bas and Ledezma (2010) among others). Bernard et al.(2010), using ï¬?rm-product level data, also show that trade liberalization affects multi-product ï¬?rms’ export patterns.2 However, there is scarce evidence on the different mechanisms through which unilateral trade liberalization in a country affects the export performance of ï¬?rms located in other countries. This paper sheds new light on intra-ï¬?rm adjustments due to trade integration combining the unilateral trade liberalization process experienced by China at the end of the 90s with ï¬?rm-product-destination data for French ï¬?rms (1999-2005). After a unilateral trade reform involving asymmetric countries, two opposite forces are at stake: market access expansion and strengthening of competitive pressures in the liberalized market. The main contribution of this paper to the literature is to investigate this trade-off. On the one hand, tariff reductions increase foreign demand and induce ï¬?rms to export more products to the liberalized markets. On the other hand, we also expect an intensiï¬?cation of foreign competition in each market. This channel is related to the tougher competition of third countries in the liberalized market that might affect negatively the expansion of French exports to that destination. Addressing the trade-off between market access expansion and tougher competition, we can evaluate French ï¬?rms’ export strategies in the event of a unilateral trade liberalization.3 We give speciï¬?c attention to the France-China trade relationship so to account for the role of China in the global trading system and its trade policy commitment. Using French micro-level data, our empirical strategy consists in investigating how French ï¬?rms’ product scope and export sales changed as a consequence of China’s a-vis Asian trade liberalization. We focus on a sample of Asian largest unilateral tariff liberalization vis-` 2 countries as a group of comparison. This sample represents a homogeneous group of countries in terms of trade integration and geographical proximity.4 To capture the effect of market access expansion, we rely on changes in applied tariffs at HS6 level from TRAINS. These tariff measures are related to the ï¬?rm, country and year level dimension. Since we are interested in unilateral trade liberalization episode, we use the Most Favorite Nation (henceforth MFN) applied tariffs set by each Asian country to the rest of the world. To enter the WTO, each country sets the same tariff cuts with respect to all countries according to a multilateral negotiation. For this reason, it is unlikely that French ï¬?rms have influenced tariff cuts negotiations. Therefore, these tariff measures allow us to exploit an exogenous variation of tariff across ï¬?rm-country pairs. To account for the intensiï¬?cation of competition faced by each ï¬?rm in the liberalized markets, we use three different measures of foreign competition at the ï¬?rm-country level. The ï¬?rst competition measure is a ï¬?rm-level Herï¬?ndahl index that captures competitive pressures at the extensive margin, where the number of foreign competitors is proxied by the number of countries exporting the same product line that the French ï¬?rms towards the Asian markets. The second measure of foreign competition is also captured by a ï¬?rm-level Herï¬?ndahl index, but in this case the number of foreign competitors is proxied by the number of French ï¬?rms exporting the same product line towards the Asian markets. Finally, the third competition measure captures the intensive margin competition faced by French ï¬?rms in each Asian market. Our results present novel insights on product turnover associated to trade reforms. Our ï¬?ndings indicate that: the expansion of French exported products and sales to China after tariff cuts is stronger relative to other Asian destinations. All the speciï¬?cations suggest that the increase in foreign competition in the liberalized market has a negative effect on the number and the value of products exported by French ï¬?rms to the Asian countries. Once accounting for foreign competition in the liberalized market, our estimates indicate that the average Chinese tariff cuts is associated to a larger expansion in the number of exported products (7 percent) and in export sales (8.4 percent) for the average ï¬?rm when compared to 3 the Asian sample. We next investigate which are the exported products that ï¬?rms are expanding the most. We split the sample into intermediate and ï¬?nal good products using the Broad Economic Category (BEC) classiï¬?ca- tion from United Nations. Our ï¬?ndings suggest that Chinese liberalization has almost no effect on the subsample of ï¬?rms exporting ï¬?nal goods. The Chinese tariff cuts have affected mainly the expansion of exports of intermediate goods. This result continues to hold when we control for foreign competition. These ï¬?ndings highlight the relevance of intermediate goods exports to China. This result can further be related to the predominant role of multinational ï¬?rms. To explore this feature, we split the sample into a multinational ï¬?rms subsample. Our point estimates imply that a 7 percentage points decline in Chinese MFN tariff increases more than twice the number of exported products by multinational ï¬?rms located in France relative to other exporting ï¬?rms. Several sensitivity tests were performed. The ï¬?ndings are robust to alternative ï¬?rm level controls such as ï¬?rm size and labor productivity. In order to address the potential reverse causality issue between Asian tariff changes and French ï¬?rms export patterns, we carry out robustness checks using an alternative weighted average tariff measure at the ï¬?rm level using initial weights. Finally, we demonstrate that our ï¬?ndings are not driven by the countries selected in the control group. The results remain unchanged when we restrict the sample to Asian countries with a similar level of economic development by excluding Korea, Japan and Singapore. Our paper contributes to the growing body of literature on micro-economic effects of trade liberal- ization. Recent developments in international trade theory focus on the patterns of multi-product ï¬?rms and the “within-ï¬?rmâ€? adjustments to trade liberalization. Mayer et al. (2009) and Bernard et al. (2010) introduce multi-product ï¬?rms in heterogeneous ï¬?rms’ models based on the pioneering work of Melitz (2003).5 Recent empirical studies using disaggregated data at the ï¬?rm-product level focus on the impact of trade liberalization on export choices of multi-product ï¬?rms. Iacovone and Javorcik (2010) study the patterns of the export boom of Mexican ï¬?rms in 1994-2003. They ï¬?nd a huge product turnover within 4 ï¬?rms, an expansion of the number of traded products and a growth in the volume of pre-existing prod- ucts. Other works focus on how the Canada-US Free Trade Agreement (CUSFTA) has affected US ï¬?rms’ export patterns (Bernard et al., 2010) or Canadian ï¬?rms’ export product scope (Baldwin and Gu, 2009). e (2009) investigate the role of a reduction of Using French ï¬?rm-product level data, Berthou and Fontagn´ trade costs on the product mix of French exporters using the introduction of the euro as a proxy for trade barriers. Dhingra (2009) tests her monopolistic competition model of brand differentiation by examining Thai trade liberalization process (2003-2006).6 The main contribution of this paper is to disentangle the effects of market access expansion vs. foreign competitive pressures after a unilateral trade liberalization episode in a fast growing developing country like China. The rest of the paper is organized as follows. The next section presents the main mechanisms that are at stake after a unilateral trade liberalization episode between asymmetric countries. Section II pro- vides stylized facts about China’s trade liberalization and the patterns of French multi-product exporters. Section III presents the empirical strategy and Section IV depicts the econometric evidence based on ï¬?rm-product-level data for French exporters. Section V presents robustness tests dealing with omitted variable concerns, potential endogeneity issues and country selection issues. Section VI concludes. I. MICROECONOMIC EFFECTS OF UNILATERAL TRADE LIBERALIZATION Before analyzing the relationship between Chinese unilateral trade liberalization and French ï¬?rms’ export patterns, this section provides some insights on the potential effects of a unilateral trade liberal- ization episode on the export performance of ï¬?rms exporting toward the liberalized market. The empirical literature on international trade and heterogeneous ï¬?rms has focused mainly on the effects of bilateral trade liberalization between symmetric countries. Baldwin and Gu (2009), by the means of plant-product level data for Canada, show that the Canada-US Free Trade Agreement (CUS- FTA) has reduced the product diversiï¬?cation and size of non-exporting Canadian plants. Using also the 5 CUSFTA as a case of bilateral trade liberalization process, Bernard et al. (2010) test their model based on ï¬?rm-product level data for the US using a difference-in-difference framework. Their ï¬?ndings show that ï¬?rms concentrate their production in their core competencies (their best selling products) after trade liberalization. They ï¬?nd that ï¬?rms that experienced larger Canadian tariff cuts (above the median) re- duce the number of products they produced for the domestic market relative to ï¬?rms experiencing below median Canadian tariffs reductions. The nature of a unilateral trade liberalization episode is different than the bilateral trade liberaliza- tion. Under a bilateral trade agreement like the CUSFTA, both countries that undertake the reform will experience an increase in market access. Thereby, a bilateral trade reform between symmetric countries, affects mainly the market potential of exporting ï¬?rms belonging to the FTA. Differently, the unilateral trade reform, by affecting all countries in the world, gives rise to two opposite mechanisms. The ï¬?rst mechanism is related to market access expansion. The export patterns of foreign ï¬?rms selling their goods to the liberalizing market are affected by the increase in the foreign demand. The second mechanism is related to the intensiï¬?cation of foreign competition in the liberalizing market. This paper aims at quantifying these two effects associated to a unilateral liberalization episode. More speciï¬?cally, we propose to evaluate how Chinese unilateral trade liberalization affected the pattern of French exporters. Chinese import tariff reductions, by boosting import demand in China, should induce French ï¬?rms to export more products toward the Chinese market. Nevertheless, since Chinese import tariff reductions were addressed to all countries, also ï¬?rms located in other countries should have taken advantage from the new export opportunities in the Chinese market. This process enhances competitive pressures in the Chinese market and thus curbs exports by French ï¬?rms. In the econometric analysis, we disentangle these two channels by using different measures of foreign competition at the ï¬?rm level. In the next section, we present some descriptive evidence on how Chinese tariff cuts might shape French ï¬?rms’ export patterns. 6 II. A FIRST GLANCE AT THE DATA Unilateral trade liberalization took place in several Asian countries in the 1990s. All the Asian countries considered in this study entered the WTO in 1995 but China.7 The main exception is China that joined the WTO at the end of 2001. The Chinese tariff reduction process started well before 2001 though. In fact, to join WTO, China has agreed to undertake a series of important commitments to open and liberalize its regime. This allowed China to be better integrated into the world economy and offer a more predictable environment for trade and foreign investment in accordance with WTO rules. This process started from the mid 1990s, when China gradually eliminated trade barriers and expanded market access to goods from foreign countries.8 Between 1999 and 2005, the average MFN tariff applied by China falls 7 percentage points, while the reduction in the average MFN tariff applied by the other Asian countries in our sample is of the order of 2 percentage points.9 China has important trade relationships with European countries: Europe is China’s largest export market and Europe’s largest source of imports. During the 2000s, EU-China trade increased dramatically, doubling between 1999 and 2005. Despite China being one of the most important challenges for EU trade policy, little is known about the behavior of multi-product ï¬?rms and a-vis China’s liberalization. the importance of product turnover vis-` Figure 1 plots the average growth of the number of French ï¬?rms exporting to China and to other Asian destinations during 1999-2005.10 As one can easily remark, while the number of French exporting ï¬?rms’ to China increases by 60 percent between 1999 and 2005, the number of exporting ï¬?rms to other Asian destinations decreases almost by 10 percent over the same period. To understand the effect of Chinese trade reforms on the exporting behavior of French ï¬?rms, Fig- ures 2 and 3 relate the change in tariffs with the change in export sales and exported products between 2005 and 1999. Figure 2 plots export sales to China (other Asian countries) with respect to the average Chinese applied MFN tariff at ï¬?rm level (average tariff of other Asian countries). Figure 3 presents a 7 Figure 1: Number of exporting ï¬?rms (1999=100) Note: authors calculations based on French customs dataset for 1999-2005, where the base year is 1999. similar relation, but focusing on the number of exported products per ï¬?rm to China (other Asian coun- tries). These ï¬?gures provide a preliminary evidence of French exports expansion towards the Chinese liberalizing market over the period under analysis. However, French ï¬?rms exporting to China might also suffer from tougher competition, due to the speciï¬?city of Chinese liberalization. To account for tougher competition, we use a concentration Herï¬?nd- ahl index at the ï¬?rm level. In this measure, the number of foreign competitors is proxied by the number of countries exporting the same product line towards each of the Asian markets. Figure 4 shows the evolution of this measure over time with respect to China and to the other Asian destinations.11 The Herï¬?ndhal index decreases more for China relative to other Asian destinations. Notice that a lower Herï¬?ndhal index implies less concentration and more competition (in terms of number of countries). This descriptive evidence indicates an increase in competitive pressures faced by French ï¬?rms in the Chinese market after the unilateral trade liberalization. To sum up, this preliminary evidence highlights that French ï¬?rms expanded their exports to China after Chinese unilateral trade reforms. Nevertheless, they also faced stronger competitive pressures in 8 Figure 2: Evolution of Intensive margin with respect to tariff cuts Note: authors calculations based on French customs dataset for 1999 and 2005. Chinese MFN tariffs are from TRAINS dataset. Figure 3: Evolution of Extensive margin with respect to tariff cuts Note: authors calculations based on French customs dataset for 1999 and 2005. Chinese MFN tariffs are from TRAINS. 9 Figure 4: Extensive margin of foreign competition (N countries) faced by French ï¬?rms 1999 2000 2001 1999 2000 2001 .5 .4 .4 Herfindhal index Other Asian (N° countries) .3 Herfindhal index China (N° countries) .3 .2 .2 .1 .1 0 0 2002 2003 2004 2002 2003 2004 .5 .4 .4 .3 .3 .2 .2 .1 .1 0 0 Note: authors calculations based on BACI dataset and French customs dataset for 1999-2005. China. To evaluate the net effect of Chinese unilateral liberalization on French ï¬?rms’ export patterns, we should take into account explicitly the role of foreign competition in the liberalized market. III. EMPIRICAL STRATEGY Data description We use individual export data on manufacturing goods for France, collected by French Customs.12 These data contain the value of exports by product, ï¬?rm and destination over the time period 1999-2005. This database classiï¬?es export flows at the ï¬?rm level within 8 digit product categories. Our analysis is restrict to manufactured products. We use MFN applied tariffs at the HS6 product level collected from TRAINS.13 From this database, we consider the tariffs applied over the period 1999 to 2005 by our set of Asian countries, i.e. China, India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. We exclude Hong Kong and Taiwan from our analysis since they are ï¬?nancial and trade centers, where the wholesale activity is very important. The next section explains how we construct the ï¬?rm level tariff. We also control for 10 country size using GDP from the Penn World Tables and the real exchange rate. We use the bilateral real exchange rate between France and other countries using producer prices of France and importer countries from the International Financial Statistics (IFS) of the IMF. Finally, to compute the proxy for foreign competition at the extensive and intensive margin, we use the BACI dataset at the HS6 product level by country of origin.14 In order to keep a constant sample throughout the paper and to establish the stability of the point es- timates, we only consider ï¬?rms that have information on all control variables. In the main speciï¬?cations, this leaves us with around 4,900 ï¬?rm-country pairs for the 9 Asian countries in the period 1999-2005, a total of 34,327 observations. To perform the robustness check exercises, we use two additional ï¬?rm level datasets. First, to identify French multinational ï¬?rms, we match our main dataset with ï¬?rm level dataset on multinational groups located in France from the Enquete Echanges Internationaux Intra-Groupe produced by the French Of- ï¬?ce of Industrial Studies and Statistics (SESSI). These data provide a good representation of the activity of international groups located in France. They account for around 82 percent of total trade flows by multinationals, and for 55 and 61 percent of total French imports and exports respectively. Second, to add information on ï¬?rms’ characteristics, we merge our main dataset with the Annual French Business Surveys (EAE), available from INSEE. The EAE survey is conducted every year and provides detailed ï¬?rm-level information for all French ï¬?rms with more than 20 employees whose main activity is in manu- facturing. This survey allows us to have information on ï¬?rms’ employment and labor productivity (value added per worker). Market Access: ï¬?rm level tariff measures To identify the impact of Asian’s trade liberalization on ï¬?rms’ export patterns, we use tariffs at the HS6 product level to construct a ï¬?rm level measure of tariff which varies by year and country of origin. 11 For each ï¬?rm f and country of destination j , we generate a simple average tariff over all HS6 products exported by that ï¬?rm to that country in year t. To avoid possible endogeneity issues, we use the simple average of the tariffs faced by each ï¬?rm in a particular country, instead of using the weighted average. In the last section we present a robustness test with an alternative tariff measure constructed using weighted average and ï¬?xing the weights at the beginning of the period. Tariffs for each ï¬?rm-country-year, Ï„f jt , are g ∈G Ï„gjt computed in the following way: Ï„f jt = Nf jt , where G is the set of products exported by ï¬?rm f and Nf jt is the number of products that ï¬?rm f exports to country j in year t. Previous empirical works focus on tariff variations across industries where the ï¬?rm produces. One of the few exceptions is the work of Teshima (2009), who uses plant-level tariffs to study import competition and R&D in Mexico. The changes in ï¬?rm-level tariff will then capture market access differences across ï¬?rms. These differences depend on the type of HS6 products that each ï¬?rm produces and exports. These tariff changes are most likely exogenous from the perspective of French ï¬?rms. In fact, Asian’s tariff changes were part of the negotiation process aimed at allowing entrance into WTO. We will thus exploit this exogenous variation in tariffs, to identify how changes in market access affect French ï¬?rms’ export patterns. Foreign competition measures: at the ï¬?rm-country level Our market access measure varies at the ï¬?rm-country level. Thus, the foreign competition measure should also vary at the ï¬?rm-country level. To control for the foreign competition at this level, we build three different measures. These measures, which capture different types of competition, are called: (i) extensive margin competition (Nâ—¦ countries), (ii) extensive margin competition (Nâ—¦ ï¬?rms) and (iii) 15 intensive margin competition. The ï¬?rst measure is a product-speciï¬?c concentration measure, based on the number of countries. We compute this measure using import flows at the HS6 product level and country of origin from BACI for 198 developed and developing countries.This measure captures, product by product, the geographical 12 concentration of imports in each of the Asian countries. Said this differently, it measures from how many countries an HS6 product is imported by each of our possible destination countries. Using notations, this concentration measure indicates the number of exporting countries, k , from which a product g is imported in each of our possible destination countries, j , thus: ng (sjk 2 g ) − 1/ng k=1 Hgj = (1) 1 − 1/ng where time subscripts are omitted for simplicity. In this expression, sjk jk jk g = mg /mg , and mg is the import value of product g imported by country j from country k and ng is the total number of countries 16 exporting the product g . To obtain a competition measure at the ï¬?rm-country level, the next step is to match the above product- speciï¬?c index with information on French exports at the ï¬?rm-product-country of destination level over the period 1999-2005. This allows us to compute the average of the Herï¬?ndahl index at the ï¬?rm level. To avoid compositional effects and potential endogeneity concerns, we take the average of this ï¬?rm level index keeping constant the HS6 product range of French exporters at 1999, while allowing the concentration index to vary over time.17 In this ï¬?rst measure, the number of foreign competitors is proxied by the number of countries ex- porting the same product line towards the Asian markets. To account for the competitive pressures that could emerge from other ï¬?rms selling the same product line to the same destination, a second measure of competition is needed. To proxy foreign competition suffered from other ï¬?rms, we introduce a second measure that captures the degree of competition due to the mass of French ï¬?rms selling the same HS6 product in the liberalized market.18 Using French Customs export data at the ï¬?rm-product-country of destination level, we construct an Herï¬?ndahl index that this time captures the concentration of export shares of other French ï¬?rms exporting same product to same country. The inverse of this index captures the degree of competitive pressures generated by French ï¬?rms selling in that country. Similar to the pre- 13 vious measure, we compute an average at the ï¬?rm level of this index keeping constant the HS6 product range exported by each ï¬?rm in 1999. After a liberalization event, it might also happen that the number of countries or ï¬?rms exporting a certain HS6 variety to destination j stays constant, while the volumes exported might change. This in turns would certainly have an impact on the export decisions of French ï¬?rms. To capture this type of competition, we construct a measure of intensive margin competition faced by French ï¬?rms in each Asian markets. To do so, we combine import data at HS6 product and country of destination level from BACI with the French customs dataset. Similarly to the other measures, we generate for each ï¬?rm f and country of destination j the average of import volumes faced by each HS6 product g over all products exported by that ï¬?rm to that country in year t. As usual, we keep constant the HS6 product range of each ï¬?rm at year 1999. Baseline Speciï¬?cation In this section we exploit the exogenous variation in tariffs across ï¬?rm-country pairs to analyze the role of Chinese unilateral trade liberalization on French ï¬?rms’ export behavior relative to tariff changes in other Asian destinations. To capture these effects, we will then estimate the following equation: lnXf jt = ατf j,t−1 + β (Ï„f j,t−1 × Chinaj ) + γZjt + µf + κj + Ï…t + f jt (2) where the dependent variable, Xf jt , is the number of products exported by ï¬?rm f to country j in year t. In an alternative speciï¬?cation, we also explore how the intensive margin of trade is affected by Chinese tariff reductions using as a dependent variable ï¬?rms’ export sales. Ï„f j,t−1 is the average tariff faced by ï¬?rm f when exporting to country j in year t − 1.19 Chinaj is a dummy variable equal to one if the country of destination, j , is China. Ï„f j,t−1 × Chinaj is an interaction term between the average lagged tariff faced by each ï¬?rm and the dummy variable for China. Zjt are controls at the country level that 14 vary over time. µf , κj and Ï…t are respectively a full set of ï¬?rm, country and year ï¬?xed effects. Finally, f jt is the random error term. The coefï¬?cient of the interaction term in equation (2), β , captures how China’s trade liberalization affected French exports relative to the average effect of liberalization in other Asian destinations. The total impact of Chinese import tariff cut on French ï¬?rms’ product scope and export sales is captured by α + β . Our subsample of Asian countries includes: India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. This sample represents a homogeneous group in terms of trade integration and geographical proximity. Nevertheless, to ensure that the selection of countries is not affecting the results, several robustness tests are carried out. Sub-section titled Alternative country samples in Section V presents two sensitivity tests. The ï¬?rst test restricts the sample to least developed (henceforth LDC) Asian countries, excluding high income countries like Korea, Japan and Singapore. The other test consists in excluding one by one each of the Asian LDC countries in the sample. The identiï¬?cation strategy in equation (2) disentangles the variation in the extensive (intensive) mar- gin of exports due to changes in China’s trade barriers. We expect a negative and signiï¬?cant α coefï¬?cient: tariff reduction in destination j increases the number of products exported (extensive margin of trade) and export sales (intensive margin of trade) towards each Asian destination. For what concern, our other coefï¬?cient of interest, β , we also expect it to be negative and signiï¬?cant. This result will indicate that Chinese tariff reduction induces French ï¬?rms to expand by a larger amount their exports towards China relative to the comparison group. Since our variable of interest (tariff measure) varies at the ï¬?rm-country level, in all the estimations, disturbances are corrected for clustering across ï¬?rm-country pairs. We control for macroeconomic shocks, ï¬?rm and destination unobservable characteristics that might affect French exports, by using year, ï¬?rm and destination ï¬?xed effects. It is important to stress that the outstanding role of China in the world trading system is not only related to trade liberalization, but also to its remarkable economic growth which has taken place during the same period. Thus, failing to control 15 for country observable characteristics that might evolve over time can generate misleading results. To deal with this issue, we include the GDP and real exchange rates (RER) for each destination country. We control for variations in real GDP across countries over time using the logarithm of lagged real GDP to capture differences across countries in terms of economic development. Finally, we take into account the effect of bilateral variations in RER. To do this, we compute the bilateral real exchange rate between France and each Asian destination country using producer prices in France and in the importing country. Controlling for Foreign Competition To account for third country competition effect faced by each ï¬?rm in the destination country, the reduced form equation in (2), becomes: lnXf jt = ατf j,t−1 + β (Ï„f j,t−1 × Chinaj ) + γZjt + Ï?F Cf jt + µf + κj + Ï…t + f jt (3) where F Cf jt is a vector containing the three foreign competition measures at ï¬?rm-country. As de- scribed in sub-section Foreign competition measures of Section III, these measures control for different types of competitive pressures that each French ï¬?rm is facing in each destination market. IV. RESULTS Baseline results: Market Access and Export Patterns In this section we present the results of the baseline estimations which explores how Asian tariff changes affect ï¬?rms’ export performance. Estimation results of equation (2) are reported in Tables 1 and 2. The former reports the results using the number of exported products as a dependent variable, 16 while the latter focus on export sales per ï¬?rm. In every table presented we control for unobserved ï¬?rm, destination and year ï¬?xed effects. We start with the extensive margin of exports. Column (1) in Table 1 shows that tariff reductions increase the number of French exported products across all destinations. In columns (2) and (3) we include country level controls. Column (2) shows that our results are also robust to cross countries price variations, proxied by the real exchange rate at the country level (RER). The coefï¬?cient of RER is negative and signiï¬?cant as expected. Column (3) introduces real GDP to capture differences across countries in terms of economic growth, development and market size. As expected the coefï¬?cient of real GDP is positive and signiï¬?cant. The coefï¬?cient on tariffs remains stable and robust to the inclusion of country observable characteristics varying over time. This ï¬?nding implies that tariff changes are not picking up effects of market size, economic growth or price variations across countries. The point estimate in column (3) indicates that the 3.5 percentage points decline in the average MFN tariff across Asian countries in our sample is associated to almost 1.6 percent expansion in the average ï¬?rm export product scope. Column (4) includes our main variable of interest: the interaction term between the average tariff at the ï¬?rm level and the dummy equal to one when the importer country is China. The coefï¬?cient is negative and statistically signiï¬?cant, at the 1% conï¬?dence level. This result indicates that China’s tariff cuts have, as expected, a positive effect on the amount of exported products towards this destination. The coefï¬?cient of the interaction term shows the impact of Chinese tariff cuts on French exports relative to the average effect of liberalization in other Asian destinations. The point estimate implies that a 10 percentage points fall in Chinese tariff results in almost 16 percent expansion more when a ï¬?rm exports to China relative to the Asian sample. During our period of analysis, from 1999 to 2005, Chinese tariffs declined on average 7 percentage points. Thus, according to our results, this would imply that the average ï¬?rm experiences an additional expansion of almost 11 percent in the number of products exported when exporting to China relative to other Asian destinations. The net effect of Chinese unilateral liberalization on French ï¬?rms’ product scope is captured by the sum of the coefï¬?cient on tariff and the interaction term 17 (α and β in equation (3)). In this case the quantiï¬?cation of the results is very similar. Table 1: The impact of Chinese unilateral trade liberalization on the extensive margin of exports Dependent variable Log Exported products of ï¬?rm f in country j in year t (1) (2) (3) (4) Tarifff j,t−1 -0.492*** -0.497*** -0.453*** -0.167 (0.136) (0.136) (0.137) (0.150) Tarifff j,t−1 × Chinaj -1.586*** (0.312) RERj,t−1 -0.311* -0.045 -0.141 (0.164) (0.170) (0.171) GDPj,t−1 0.316*** 0.235** (0.094) (0.096) Firm ï¬?xed effects Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Observations 34,327 34,327 34,327 34,327 R2 0.542 0.542 0.543 0.543 Notes: The regressions are OLS estimations of equation (2) for the period 1999-2005. The dependent variable is the logarithm of the number of products exported to country j in year t by ï¬?rm f . Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. The destinations are China and other Asian countries that already integrate WTO such as India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. Tarifff jt is the tariff measure faced by ï¬?rm f when exporting to country j in year t. Tarifff jt × Chinaj is an interaction term between the ï¬?rm level tariff measure and a dummy variable equal to one when the country of destination of exports is China and zero otherwise, Chinaj . GDPjt is the natural log of the GDP of country j from the Penn World Tables. RERjt is the bilateral real exchange rate between France and China and countries in the control group using producer prices of France and importer countries from the International Financial Statistics (IFS). Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Table 2 reports similar results for the intensive margin of exports. After controlling for observable country level characteristics in columns (2) and (3), the average reduction of Asian tariffs leads to almost 3 percent increase in French ï¬?rms’ export sales. The coefï¬?cient of the interaction term between the tariff measure with the China dummy variable in column (4) shows a larger effect due to Chinese liberalization relative to the average Asian tariff cuts. This allow us to conclude that China’s tariff reductions increase by a larger amount export sales of French ï¬?rms towards this destination relative to the others: Chinese tariff cuts boost by almost 19 percent (7 times β ) French export sales. While the net effect of Chinese liberalization on French ï¬?rms export sales is 21 percent (7 times (α + β )). 18 Table 2: The impact of Chinese unilateral trade liberalization on the intensive margin of exports Dependent variable Log Export sales of ï¬?rm f in country j in year t (1) (2) (3) (4) Tarifff j,t−1 -0.868*** -0.876*** -0.834*** -0.350* (0.182) (0.182) (0.184) (0.195) Tarifff j,t−1 × Chinaj -2.678*** (0.416) RERj,t−1 -0.536** -0.279 -0.441** (0.212) (0.221) (0.221) GDPj,t−1 0.305*** 0.169 (0.117) (0.118) Firm ï¬?xed effects Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Observations 34,327 34,327 34,327 34,327 R2 0.574 0.574 0.574 0.576 Notes: The regressions are OLS estimations of equation (2) for the period 1999-2005. The dependent variable is the logarithm of ï¬?rm f ’s export sales to country j in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. The destinations are China and other Asian countries that already integrate WTO such as India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. Tarifff jt is the tariff measure faced by ï¬?rm f when exporting to country j in year t. Tarifff jt × Chinaj is an interaction term between the ï¬?rm level tariff measure and a dummy variable equal to one when the country of destination of exports is China and zero otherwise, Chinaj . GDPjt is the natural log of the GDP of country j from the Penn World Tables. RERjt is the bilateral real exchange rate between France and China and countries in the control group using producer prices of France and importer countries from the International Financial Statistics (IFS). Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Unilateral Trade Liberalization and Foreign competition The ï¬?ndings presented in the previous section might suffer from an important omitted variable bias related to foreign competition faced by each ï¬?rm in the destination market. To control for this issue, the baseline speciï¬?cation in (2) is extended to include our three ï¬?rm-level measures of foreign competition.20 A positive and signiï¬?cant coefï¬?cient for both Herï¬?ndahl indexes implies that an increase in foreign competition in destination j faced by each ï¬?rm (i.e. a reduction in the ï¬?rm level Herï¬?ndahl concentra- tion index) reduces the number of products exported and export sales to that destination. On the other hand, for the intensive margin competition measure, we expect a negative and signiï¬?cant coefï¬?cient. This measure captures the variation on the volume of imports by incumbent competitors towards the lib- eralized market. Chinese unilateral liberalization will increase the volume of imports of ï¬?rms/countries 19 already selling in China. This in turn might reduce French ï¬?rms exports towards that destination due to tougher competition at the intensive margin. Tables 3 and 4 report the estimation results of equation (3) for the number of exported products and for export sales respectively. Columns (1) to (3) introduce the different measures of foreign competition. As expected the coefï¬?cients on both Herï¬?ndahl indexes (extensive margin competition proxies) are pos- itive and signiï¬?cant in all speciï¬?cations, suggesting that the higher the competitive pressures faced by the average French ï¬?rm in a destination market, the lower will be the number of products exported and export sales to that destination. While the intensive margin competition proxy is negative and signiï¬?- cant implying that the greater increase in the volume of imports of incumbent competitors of the same products to the same destination after the liberalization reduces French ï¬?rms’ export propensity. As a benchmark, column (4) and (6) report the baseline estimation results presented in column (3) and (4) of Table 1. Once we take into account the foreign competition pressures induced by the unilateral trade lib- eralization faced by French ï¬?rms, the coefï¬?cient of the average Asian tariff cuts is negative but no longer signiï¬?cant (column 5) and the coefï¬?cient of interest is still negative and signiï¬?cant but the magnitude is reduced by 35 percent (column 7). The results presented in Table 4 for the intensive margin of exports are in the same line. Findings in Tables 3 and 4 indicate that once we address the possible omitted variable issue, by controlling explicitly for foreign competition at the ï¬?rm-country level, the 7 percentage points reduction of Chinese tariff cuts result in an additional expansion of almost 7 percent (instead of 11) in the number of products exported and 12 percent in export sales (instead of 19) by the average ï¬?rm when exporting to China relative to other Asian destinations. 20 Table 3: Third country effects. Foreign competition measures at the ï¬?rm-country level Dependent variable Log number of exported products (1) (2) (3) (4) (5) (6) (7) Extensive margin competitionf jt (Nâ—¦ countries) 1.080*** 0.672*** 0.889*** 0.758*** 0.752*** (0.093) (0.091) (0.093) (0.089) (0.089) â—¦ Extensive margin competitionf jt (N ï¬?rms) 0.900*** 1.024*** 0.806*** 0.807*** (0.058) (0.059) (0.054) (0.054) Intensive margin competitionf jt -0.090*** -0.088*** -0.088*** (0.006) (0.006) (0.006) Tarifff j,t−1 -0.492*** -0.185 -0.167 0.002 (0.136) (0.129) (0.150) (0.142) 21 Tarifff j,t−1 × Chinaj -1.586*** -1.041*** (0.312) (0.293) Country level controls No No No Yes Yes Yes Yes Firm ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Observations 34,327 34,327 34,327 34,327 34,327 34,327 34,327 R2 0.547 0.555 0.560 0.542 0.582 0.543 0.582 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005. The dependent variable is the logarithm of the number of exported products to country j by ï¬?rm f in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. The destinations are China and other Asian countries that already integrate WTO such as India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. Tarifff jt is the tariff measure faced by ï¬?rm f when exporting to country j in year t. Tarifff jt × Chinaj is an interaction term between the ï¬?rm level tariff measure and a dummy variable equal to one when the country of destination of exports is China and zero otherwise, Chinaj . The extensive margin competitionf jt (Nâ—¦ countries) variable is the ï¬?rm-level Herï¬?ndahl index, where the number of foreign competitors is proxied by the number of countries exporting the same product line that the French ï¬?rms towards the Asian markets. The extensive margin competitionf jt (Nâ—¦ ï¬?rms) variable is the ï¬?rm-level Herï¬?ndahl index, where the number of foreign competitors is proxied by the number of other French ï¬?rms exporting the same product line towards the Asian markets. Finally, the intensive margin competitionf jt variable is the average import volumes at the product level faced by each ï¬?rm in the destination markets. GDPjt is the natural log of the GDP of country j from the Penn World Tables. RERjt is the bilateral real exchange rate between France and China and countries in the control group using producer prices of France and importer countries from the International Financial Statistics (IFS). Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Table 4: Third country effects. Foreign competition measures at the ï¬?rm-country level Dependent variable Log export sales (1) (2) (3) (4) (5) (6) (7) Extensive margin competitionf jt (Nâ—¦ countries) 2.039*** 1.324*** 1.441*** 1.431*** 0.594*** (0.121) (0.114) (0.117) (0.117) (0.095) Extensive margin competitionf jt (Nâ—¦ ï¬?rms) 1.579*** 1.647*** 1.647*** 1.026*** (0.079) (0.080) (0.080) (0.064) Intensive margin competitionf jt -0.049*** -0.049*** -0.010 (0.008) (0.008) (0.007) Tarifff j,t−1 -0.868*** -0.687*** -0.350* -0.601*** 22 (0.182) (0.178) (0.195) (0.159) Tarifff j,t−1 × Chinaj -2.678*** -1.824*** (0.416) (0.331) Country level controls No No No Yes Yes Yes Yes Firm ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Yes Yes Yes Observations 34,327 34,327 34,327 34,327 34,327 34,327 34,327 R2 0.585 0.599 0.600 0.574 0.601 0.576 0.593 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005. The dependent variable is the logarithm of the export sales in country j by ï¬?rm f in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. The destinations are China and other Asian countries that already integrate WTO such as India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. Tarifff jt is the tariff measure faced by ï¬?rm f when exporting to country j in year t. Tarifff jt × Chinaj is an interaction term between the ï¬?rm level tariff measure and a dummy variable equal to one when the country of destination of exports is China and zero otherwise, Chinaj . The extensive margin competitionf jt (Nâ—¦ countries) variable is the ï¬?rm-level Herï¬?ndahl index, where the number of foreign competitors is proxied by the number of countries exporting the same product line that the French ï¬?rms towards the Asian markets. The extensive margin competitionf jt (Nâ—¦ ï¬?rms) variable is the ï¬?rm-level Herï¬?ndahl index, where the number of foreign competitors is proxied by the number of other French ï¬?rms exporting the same product line towards the Asian markets. Finally, the intensive margin competitionf jt variable is the average import volumes at the product level faced by each ï¬?rm in the destination markets. GDPjt is the natural log of the GDP of country j from the Penn World Tables. RERjt is the bilateral real exchange rate between France and China and countries in the control group using producer prices of France and importer countries from the International Financial Statistics (IFS). Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Disentangling Input and Output trade liberalization A further question worthwhile exploring is related to the types of goods: which are the French exported products that are affected the most by Asian liberalization? To test the relationship between market access and types of traded products, we estimate equation (3) by splitting the sample into ï¬?rms producing intermediate and ï¬?nal goods. To classify HS6 products into intermediate and ï¬?nal products, we use BEC (Broad Economic Cat- egories) classiï¬?cation from United Nations.21 In line with the ï¬?rm-level tariff built in section III, we construct construct a similar measure for both intermediate and ï¬?nal goods using information at the HS6 to classify products. Table 5 reports the estimate results for our two subsamples of ï¬?rms: ï¬?rms exporting intermediate and ï¬?nal products respectively. When controlling for observable country characteristics and foreign compet- itive pressures, we ï¬?nd that Chinese tariff cuts are mainly signiï¬?cant for French exports of intermediate products (columns (1) and (3)). Our point estimate implies that a 10 percentage points fall in Chinese ap- plied tariff, increases by almost 17 percent the number of intermediate products exported and by almost 28 percent French export sales to China relative to the other Asian destinations. When we restrict the sample to ï¬?rms exporting only ï¬?nal goods, interestingly we do not ï¬?nd any signiï¬?cant effect of Asian or Chinese liberalization on French extensive margin (column (2)), while the effect on the intensive margin is only signiï¬?cant at the 10 percent level (column (4)). These ï¬?ndings conï¬?rm the results found by a number of recent works on the increasing role of intermediate inputs in international trade and the effects of input liberalization on ï¬?rm performance. Amiti and Konings (2007), using ï¬?rm-level data for Indonesia, show that the effect of reductions on input tariffs on ï¬?rm total factor productivity improvements is much important than the effect of reductions on ï¬?nal good tariffs. Goldberg et al. (2010), using ï¬?rm level data for India, demonstrate that input tariff liberalization allows ï¬?rms to expand their product scope for the domestic market. Bas and Strauss-Khan 23 (2011), using ï¬?rm level data for France, show that using imported intermediate goods improves French ï¬?rms export scope, the number of export destination countries and export sales. Along the same line, Bas (2012), using ï¬?rm level data for Argentina, ï¬?nds that input tariff cuts have a positive effect on ï¬?rms’ export decision. 24 Table 5: Robustness checks: alternative subsamples Dependent variable Log number of exported products Log export sales Intermediate products Final products Intermediate products Final products (1) (2) (3) (4) Tarifff j,t−1 -0.005 0.403 -0.499 0.041 (0.300) (0.246) (0.421) (0.318) Tarifff j,t−1 × Chinaj -1.684** -0.664 -2.764*** -1.195* (0.725) (0.541) (0.981) (0.654) Extensive margin competitionf jt (Nâ—¦ countries) 0.909*** 0.898*** 1.651*** 1.086*** (0.164) (0.167) (0.198) (0.194) Extensive margin competitionf jt (Nâ—¦ ï¬?rms) 0.892*** 0.936*** 1.501*** 1.406*** (0.083) (0.148) (0.118) (0.180) Intensive margin competitionf jt -0.056*** -0.067*** -0.023* -0.037*** (0.010) (0.012) (0.013) (0.014) Observations 11,375 8,973 11,375 8,973 R2 0.601 0.656 0.639 0.688 MNF Non-MNF MNF Non-MNF 25 Tarifff j,t−1 -0.487* 0.023 -0.603* -0.025 (0.268) (0.181) (0.335) (0.227) Tarifff j,t−1 × Chinaj -2.229*** -1.131*** -3.141*** -2.213*** (0.673) (0.336) (0.861) (0.455) Extensive margin competitionf jt (Nâ—¦ countries) 0.714*** 0.984*** 1.650*** 1.287*** (0.148) (0.118) (0.201) (0.137) â—¦ Extensive margin competitionf jt (N ï¬?rms) 1.009*** 1.018*** 1.897*** 1.443*** (0.098) (0.070) (0.139) (0.090) Intensive margin competitionf jt -0.119*** -0.076*** -0.081*** -0.030*** (0.012) (0.007) (0.015) (0.009) Observations 9,230 25,097 9,230 25,097 R2 0.549 0.556 0.537 0.585 Country level controls Yes Yes Yes Yes Firm ï¬?xed effects Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005 for the subsample of ï¬?rms exporting intermediate and ï¬?nal goods (in the bottom part of the table). In columns (1) to (3) the dependent variable is the logarithm of the number of exported products to country j by ï¬?rm f in year t, and in columns (4) to (6) the dependent variable is the logarithm of the export sales in country j by ï¬?rm f in year t. Control variables deï¬?nitions are reported in table 4. Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. The Role of Multinational Firms The evidence presented in the previous section emphasizes the importance of French intermediate goods exports during Chinese liberalization. This result can be explained by considering the predominant role of multinational ï¬?rms and intra-ï¬?rm trade in the world economy. To account for this important dimension, in this section we split our sample into multinational ï¬?rms located in France which export to Asian countries on the one side, and the remaining exporting ï¬?rms on the other side.22 Columns (1) and (3) in the bottom part of Table 5 show the results for the subsample of multinational ï¬?rms located in France, and columns (2) to (4) for the subsample of ordinary exporters. Comparing the results of these two subsamples of ï¬?rms, it is straightforward to notice the larger effect of both Asian and Chinese liberalization on multinational ï¬?rms’ extensive margin: the effect of liberalization is two times larger for MNFs. Our point estimate suggests that a 10 percentage points fall in Chinese tariff results in almost 22 percent expansion for products exported to China by multinational ï¬?rms relative to other Asian countries (column (1)). While the amount of products exported by non-multinational exporters increases only by 11 percent more for the average exporting ï¬?rm to China relative to Asian destinations (column (2)). Columns (3) and (4) present the results for the intensive margin of trade. In line with our previous ï¬?ndings, the effect of Chinese tariff reductions on export sales is more pronounced for multinational ï¬?rms than for non-multinational exporting ï¬?rms. V. ROBUSTNESS CHECKS Firm level controls In this section we explicitly deal with potential omitted variable concerns that might affect the pre- vious results. To test that our main variable of interest, ï¬?rm-level tariff, is not picking up the effects of 26 observable ï¬?rm characteristics which varies over time, we carry on an additional robustness check. In speciï¬?cation (3) we include two additional control variables: ï¬?rms’ size and labor productivity.23 Table 6 reports the results where we account for ï¬?rms’ size (employment) and ï¬?rms’ labor produc- tivity. Despite the reduction in the number of observations, our coefï¬?cient of interest remains negative and signiï¬?cant implying that Chinese tariff reductions increase both the number of exported products (column (2)) as well as export sales (column (4)) to China relative to other Asian destinations. These ï¬?ndings conï¬?rm that the previous results do not suffer from omitted variable concerns. Potential endogeneity issues and alternative tariff measures Tariff measure used in the previous estimations is constructed as a simple average of HS6 tariffs over all products exported by a ï¬?rm to destination. In the estimation procedure, to avoid the potential endogeneity bias between tariff measure and the number of exported products at each point in time, we used tariffs lagged by one period. In this section we propose an alternative way of dealing with this endogeneity issue. The alternative tariff measure is a weighted average ï¬?xing the weights in the initial year, 1999. So this tariff measure is based on the basket of all HS6 products exported by each ï¬?rm in 1999, to each speciï¬?c destination. This strategy should avoid that the tariff changes as a result of the increase in export products. Table 7 reports the estimation results.24 Our results remain robust when using this alternative tariff measure. Once taking into account foreign competitive pressures, Asian and Chinese unilateral trade liberalization has a positive effect on both the number of products exported and export sales by French ï¬?rms. 27 Table 6: Robustness checks: ï¬?rm level controls Dependent variable Log number of exported products Log export sales (1) (2) (3) (4) Labor productivityf,t−1 0.061* 0.051 0.144*** 0.128*** (0.034) (0.034) (0.034) (0.034) Sizef,t−1 0.115** 0.111** 0.249*** 0.242*** (0.054) (0.053) (0.065) (0.063) Tarifff j,t−1 -0.077 -0.664** (0.222) (0.284) Tarifff j,t−1 × Chinaj -2.482*** -3.980*** (0.548) (0.900) â—¦ Extensive margin competitionf jt (N countries) 1.006*** 1.441*** (0.131) (0.159) â—¦ 28 Extensive margin competitionf jt (N ï¬?rms) 1.069*** 1.799*** (0.068) (0.095) Intensive margin competitionf jt -0.094*** -0.060*** (0.008) (0.011) (0.220) (0.293) Country level controls Yes Yes Yes Yes Firm ï¬?xed effects Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Observations 17,241 17,241 17,241 17,241 R2 0.511 0.536 0.529 0.565 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005. In columns (1) and (2) the dependent variable is the logarithm of the number of exported products to country j by ï¬?rm f in year t, and in columns (3) and (4) the dependent variable is the logarithm of the export sales in country j by ï¬?rm f in year t. Country level control variables deï¬?nitions are reported in table 4. Firm size f t represents the logarithm of employment at the ï¬?rm level and ï¬?rm labor productivityf t is measured by value added per worker. Heteroskedasticity-robust standard errors clustered by ï¬?rm-country pairs are reported in parentheses. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Table 7: Robustness checks: alternative tariff measures Dependent variable Log number of exported products Log export sales (1) (2) Tariff 99f j,t−1 -0.172*** -0.552*** (0.061) (0.126) Tariff 99f j,t−1 × Chinaj -0.301** -0.806*** (0.150) (0.301) â—¦ Extensive margin competitionf jt (N countries) 0.388*** 0.513*** (0.038) (0.078) â—¦ Extensive margin competitionf jt (N ï¬?rms) 0.445*** 0.970*** (0.023) (0.050) Intensive margin competitionf jt -0.054*** -0.019*** (0.003) (0.006) Country level controls Yes Yes Firm ï¬?xed effects Yes Yes Country ï¬?xed effects Yes Yes Year ï¬?xed effects Yes Yes Observations 24,394 24,394 R2 0.596 0.613 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005. In columns (1) to (4) the dependent variable is the logarithm of the number of exported products to country j by ï¬?rm f in year t, and in columns (5) to (8) the dependent variable is the logarithm of the export sales in country j by ï¬?rm f in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. Tarifff jt is the tariff measure faced by ï¬?rm f when exporting to country j in year t. Using HS6 product level tariff data from TRAINS, we construct the ï¬?rm level tariff by taking the average of tariff over the basket of all the HS6 products exported by ï¬?rm f in the initial year (1999) to country j . This basket is kept ï¬?xed over the period. Country level control variables deï¬?nitions are reported in table 4. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Alternative country samples In the previous regressions, the sample of Asian countries selected as a group of comparison included countries that are homogeneous in terms of trade integration and geographical proximity, but they differ in terms of economic development. Since the export patterns of French ï¬?rms might differ according to the income level of the destination country, in this section we carry out two additional sensitivity tests. We ï¬?rst restrict the sample to LDC Asian economies and exclude the high-income Asian countries like Korea, Japan and Singapore. Table 8 reports these results. The previous ï¬?ndings are robust to the countries selected as a group of comparison. When we compare the results in columns (2) and (3), on the one hand, and columns (5) and (6), on the other, the effect of Chinese tariff liberalization relative to 29 other Asian countries liberalization on export scope and export sales of French ï¬?rms is almost two times larger for the multinational ï¬?rms relative to ordinary exporters. Finally, we carry out a last sensitivity test to see weather our results are not driven by the inclusion of any particular country in the comparison group. Based on the subsample of least developed Asian countries, we start excluding one country at a time of the sample. Table 9 presents the results. The name of the country that is excluded from the estimation is presented in the heading of each column. The ï¬?ndings are robust to these changes in the country coverage, which affect neither the sign of the coefï¬?cients nor their signiï¬?cance. 30 Table 8: Subsample of French ï¬?rms exporting to LDC Asian countries Dependent variable Log Nâ—¦ exported products Log export sales (1) (2) (3) (4) (5) (6) Full sample MNF Non-MNF Full sample MNF Non-MNF Tarifff j,t−1 -0.216 -0.340 -0.048 -0.228 -0.694** 0.155 (0.134) (0.239) (0.163) (0.191) (0.321) (0.236) Tarifff j,t−1 × Chinaj -0.782*** -1.372** -0.762*** -1.767*** -2.607*** -1.761*** (0.262) (0.543) (0.293) (0.304) (0.569) (0.357) â—¦ Extensive margin competitionf jt (N countries) 0.459*** 0.364*** 0.550*** 0.943*** 1.070*** 0.851*** (0.092) (0.130) (0.127) (0.123) (0.185) (0.160) 31 Extensive margin competitionf jt (Nâ—¦ ï¬?rms) 0.945*** 0.920*** 0.927*** 1.415*** 1.592*** 1.232*** (0.067) (0.098) (0.093) (0.088) (0.138) (0.113) Intensive margin competitionf jt -0.065*** -0.072*** -0.059*** -0.018* -0.016 -0.017 (0.008) (0.014) (0.010) (0.010) (0.018) (0.012) Firm ï¬?xed effects No No Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Yes Yes Observations 13,224 4,394 8,830 13,224 4,394 8,830 R2 0.590 0.568 0.587 0.628 0.566 0.622 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005 for the subsample of least developed Asian countries: China, India, Indonesia, Pakistan, Philippines and Thailand. The dependent variable is the logarithm of the number of exported products to country j by ï¬?rm f in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. Country level control variables deï¬?nitions are reported in table 4. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. Table 9: Sensitivity test to the selected group of LDC Asian countries Dependent variable Log number of exported products by ï¬?rm f to country j in year t (1) (2) (3) (4) (5) India Indonesia Pakistan Philippines Thailand Tarifff j,t−1 -0.230 -0.074 -0.086 -0.044 -0.698*** (0.145) (0.167) (0.140) (0.150) (0.177) Tarifff j,t−1 × Chinaj -0.766*** -1.334*** -0.918*** -0.503** -0.579** (0.270) (0.294) (0.265) (0.255) (0.285) R2 0.611 0.621 0.596 0.605 0.589 Dependent variable Log export sales by ï¬?rm f to country j in year t 32 Tarifff j,t−1 -0.318 -0.282 -0.114 -0.062 -0.153 (0.202) (0.230) (0.198) (0.211) (0.256) Tarifff j,t−1 × Chinaj -1.855*** -2.264*** -1.851*** -1.672*** -1.362*** (0.314) (0.330) (0.308) (0.323) (0.357) R2 0.652 0.670 0.634 0.646 0.636 Foreign competition measures Yes Yes Yes Yes Yes Country level controls Yes Yes Yes Yes Yes Firm ï¬?xed effects Yes Yes Yes Yes Yes Country ï¬?xed effects Yes Yes Yes Yes Yes Year ï¬?xed effects Yes Yes Yes Yes Yes Observations 11,630 9,880 12,730 11,871 11,740 Notes: The regressions are OLS estimations of equation (3) for the period 1999-2005 for the subsample of least developed Asian countries: China, India, Indonesia, Pakistan, Philippines and Thailand. The name of the excluded country is presented in the heading of each column. The dependent variable is the logarithm of the number of exported products (export sales) to country j by ï¬?rm f in year t. Fixed effects by ï¬?rm, country and year and a constant are included in all speciï¬?cations. Country level control variables deï¬?nitions are reported in table 4. ∗ ∗ ∗, ∗∗, and ∗ indicate signiï¬?cance at the 1, 5 and 10 percent levels respectively. VI. CONCLUDING REMARKS This paper contributes to the literature on the microeconomic effects of trade reform by disentangling and identifying two channels through which a unilateral trade liberalization episode affects ï¬?rms’ export performance: the expansion of market access and the intensiï¬?cation of foreign competitive pressures. This paper quantiï¬?es the different effect of Chinese versus Asian tariff cuts on French exporters. Our ï¬?ndings can be summarized as follows. First, we ï¬?nd a positive effect of Chinese unilateral liberalization on both the extensive and intensive margins of French exporting ï¬?rms. Although this effect is reduced when controlling for speciï¬?c foreign competitive pressures at the ï¬?rm level in the destination market, there is still a net positive effect of Chinese tariff cuts on French ï¬?rms exports relative to other Asian countries. Indeed, the number of exported products and export sales by French ï¬?rms towards China increased by a larger amount when compared to other Asian destinations. Second, our ï¬?ndings suggest that the effect of Chinese tariff reductions is more important for ï¬?rms exporting intermediate goods. Finally, in line with the previous ï¬?nding, we show that multinational ï¬?rms play a predominant role in explaining the positive effect of Chinese liberalization on the expansion of French ï¬?rms’ product scope. 33 Notes 1 Baldwin (2011) proposes a political economy study which disentangles the theoretical mechanisms through which a unilateral liberalization affects developing countries. 2 e (2009) also explore how ï¬?rms adjust their product mix and exported value as a consequence of a Berthou and Fontagn´ reduction of trade costs. 3 Our sample includes: China, India, Indonesia, Korea, Japan, Pakistan, Philippines, Singapore and Thailand. 4 Several robustness tests demonstrate that our results are not driven by the countries included in the comparison group. 5 Multi-product ï¬?rms’ models include Allanson and Montagna (2005), Baldwin and Gu (2009), Feenstra and Ma (2008), Eckel and Neary (2009), Nocke and Yeaple (2008), Bernard et al. (2009), Mayer et al. (2009), Arkolakis and Mundler (2008) and Dhingra (2009). 6 Her ï¬?ndings point out that Thai tariffs cut has a negative effect on process and product innovation among exporters, while it has a positive effect on product innovation of less export-oriented ï¬?rms. 7 India, Indonesia, Japan, Korea, Philippines, Pakistan, Thailand and Singapore. 8 For industrial goods the average bound tariff level will go down to 8.9 percent with a range from 0 to 47 percent. 9 There is a lot of heterogeneity across other Asian destinations. For example, the average MFN tariff applied by Korea, Japan and Singapore is not changing over time, while Indonesia’s and Philippines’s average tariff falls by 3 p.p. and Pakistan and Thailand by around 1 p.p. and India’s MFN tariff falls more than 10 p.p. 10 Other Asian destinations includes: India, Indonesia, Korea, Japan, Philippines, Singapore, Pakistan and Thailand. 11 This ï¬?gure is based on BACI dataset and French customs dataset (1999-2005). To make the ï¬?gure clearer the outliers observations are excluded. Similar ï¬?gures for the alternative competition measures are available upon request. 12 Export information is collected in the following way: exports outside EU are reported if ï¬?rms’ annual trade value exceeds 1,000 Euros or a weight of a ton. 13 The source of MFN applied tariffs is TRAINS: http://unctad-trains.org. 14 The BACI dataset is provided by the CEPII and constructed based on COMTRADE dataset from the UN. This dataset pro- vides bilateral trade flows at the 6-digit product level (Gaulier and Zignago, 2010). From BACI dataset we take information on import flows in each of our destination country. BACI is downloadable from http://www.cepii.fr/anglaisgraph/bdd/baci.htm. 15 We thank two anonymous referees for suggesting these two alternative measures of foreign competition (ii) and (iii). 16 The assumption behind this measure of extensive margin competition is that equal HS6 varieties imported from different countries are perfect substitutes. This idea has been recently discussed by the trade literature on quality (e.g. Khandelwal, 34 2010). 17 We thank an anonymous referee for suggesting this way of computing this measure. 18 Ideally, we need a measure that captures the competitive pressures faced by French ï¬?rms in each Asian market by other local and foreign ï¬?rms. To compute such measure we will need ï¬?rm-product level information on the HS6 products sold by local and foreign producers in each domestic market. To the best of our knowledge this disaggregated information at the ï¬?rm-HS6 product level is not available for China and the other Asian countries in our sample. 19 To further address the potential endogeneity issue between import tariff and export patterns, we use lagged tariff mea- sures. 20 Section III describes in detail how these measures are constructed. 21 Serious missing information problems prevented us from considering a more disaggregated product level dimension. 22 To identify multinational ï¬?rms, we combine our main dataset with the Enquete Echanges Internationaux Intra-Groupe produced by the French Ofï¬?ce of Industrial Studies and Statistics (SESSI). This latter dataset is based on a ï¬?rm-level survey of manufacturing ï¬?rms belonging to groups with at least one afï¬?liate in a foreign country and with international transactions totaling at least one million euros for the year 1999. 23 To add information on ï¬?rms’ characteristics, we match our main customs dataset with the Annual French Business Sur- veys (EAE), available from INSEE. This survey allows us to have information on ï¬?rms’ employment and labor productivity (value added per worker). Since this implies restricting the sample to exporters with more than 20 employees which have manufacturing as their main activity, the number of observations is reduced by a half. We have no ï¬?rm-level information for the Food and Beverages industry (corresponding to ISIC 15). This restricted sample covers around 14,000 observations, while the main sample has almost 28,000 observations. 24 Since the basket of products is kept constant over the period, the number of observations is reduced to almost 24,394. 35 References [1] Amiti M. and J. Konings. 2007. Trade liberalization, intermediate inputs, and productivity: Evi- dence from Indonesia. American Economic Review; 97 (5); 1611-1638. [2] Arkolakis C. and M.-A. Muendler. 2009. The Extensive Margin of Exporting Goods: A Firm-level Analysis. University of California, San Diego, manuscript. [3] Baldwin, R.E., 2005. Heterogeneous Firms and Trade: Testable and Untestable Properties of the Melitz Model. NBER Working Papers 11471. [4] Baldwin, R.E., 2011. Unilateral Tariff Liberalization. CTEI Working Paper number 2011-04. [5] Baldwin J. R. and W. Gu. 2009. The Impact of Trade on Plant Scale, Production-Run Length and Diversiï¬?cation. In T. Dunne, J.B. Jensen and M.J. Roberts (eds.), Producer Dynamics: New Evidence from Micro Data, University of Chicago Press, Chicago. [6] Bas M. and I. Ledezma. 2010. Trade Integration and Within-Plant productivity Evolution in Chile. The Review of World Economics / Weltwirtschaftliches Archiv , Vol. 146, Nâ—¦ 1, pp. 113-146. [7] Bas M. 2012. Input trade liberalization and ï¬?rm export decisions: Evidence from Argentina. Journal of Development Economics 97 (2012) 481 493. [8] Bas M. and V. Strauss Khan. 2011. Does importing more inputs raise exports? Firm level evidence from France. CEPII Working paper. [9] Bernard A., Redding S., and P.K. Schott. 2010a. Multi-product ï¬?rms and trade liberalization. NBER Working Papers n. 12782. Quarterly Journal of Economics Forthcoming. [10] Bernard A., Redding S., and P.K. Schott. 2010b. Multi-Product Firms and Product Switching. American Economic Review, Vol. 100, No. 2, May, 444-448 [11] Berthou A. and L. Fontagne. 2009. The Euro and the Intensive and Extensive Margins of Trade: Evidence from French Firm-Level Data. CEPII Working paper (WP2008-06) version. 2009. [12] Dhingra S. 2010. Trading away wide brands for cheap brands. Working paper University of Wisconsin-Madison. [13] Eckel C. and Neary J.P. 2010. Multi Product Firms and Flexible Manufacturing in the Global Econ- omy. Review of Economic Studies, 77:1, 188-217. [14] Feenstra R., and H. Ma. 2008. Optimal Choice of Product Scope for Multiproduct Firms under Monopolistic Competition. In Helpman, Elhanan, Dalia Marin and Thierry Verdier (eds.) The Or- ganization of Firms in a Global Economy, Harvard University Press, Cambridge MA. [15] Gaulier G., S. Zignago (2010), “BACI: International Trade Database at the Product-level. The 1994-2007 Versionâ€?, CEPII Working Paper 23, 2010. [16] Goldberg, P., Khandelwal, A., Pavcnik, N. and P. Topalova, 2010. Imported intermediate inputs and domestic product growth: evidence from India. Quarterly Journal of Economics 125 (4), 1727- 1767. [17] Levinsohn J, Petrin A. Estimating Production Functions Using Inputs to Control for Unobservables. Review of Economic Studies 2003; 70(2); 317-341. 36 [18] Iacovone L. and B. Javorcik. 2010. Multi-Product Exporters: Product Churning, Uncertainty and Export Discoveries. Economic Journal, Royal Economic Society, vol. 120(544), pages 481-499, 05. [19] Lileeva A. and D. Trefler. 2010. Improved Access to Foreign Markets Raises Plant-Level Produc- tivity for Some Plants. Quarterly Journal of Economics 125(4). [20] Mayer, T., Melitz, M. and G. Ottaviano. 2011. Market Size, Competition, and the Product Mix of Exporters. CEPR Discussion Paper No. 8349. [21] Melitz M. 2003. The impact of Trade on Intra-Industry Reallocations and Agregate Industry Pro- ductivity. Econometrica 2003; 71; 1695-1725. [22] Nocke, V., and S. Yeaple. 2005. Endogenizing Firm Scope: Multiproduct Firms in International Trade. University of Pennsylvania, Mimeo. [23] Pavcnik N. 2002. Trade Liberalization, Exit and Productivity Improvement: Evidence from Chilean Plants. Review of Economic Studies; 69; 245-276. [24] Schott, P.K. 2008. The Relative Sophistication of Chinese Exports. Economic Policy 53:5-49, Jan- uary. [25] Teshima, K. 2009. Import Competition and Innovation at the Plant Level: Evidence from Mexico. Mimeo. [26] Trefler D. 2004. The Long and Short of the Canada-U. S. Free Trade Agreement. American Eco- nomic Review, American Economic Association, vol. 94(4), pages 870-895, September. 37