ESSAYS IN INTERNATIONAL TRADE, POLITICAL ECONOMY OF PROTECTION AND FIRM HETEROGENEITY by ANDREY STOYANOV B.A., Novosibirsk State University, 2001 M.A., Central European University, 2003 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy in The Faculty of Graduate Studies (Economics) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) July, 2008 ©Andrey Stoyanov, 2008 Abstract This dissertation addresses two issues in international trade: the effect of foreign lobbying on trade policy and the effect of trade on technology adoption by firms. The first two chapters study the effect of foreign lobbies on trade policy of a country which is a member of a Free Trade Agreement (FTA). They rely on a monopolistically competitive political economy model in which the government determines external tariffs endogenously. In the first paper the effect of foreign lobbying under the FTA is examined empirically using Canadian industry-level trade data that allow differentiating of lobby groups by the country of origin. The analysis suggests that the presence of foreign lobbying has a significant effect on the domestic trade policy: the presence of an organized lobbying group in an FTA partner country tends to raise trade barriers while an organized lobbying group of exporters from outside of the FTA is associated with less protection. The second paper analyses political viability of FTAs and their effect on the world trading system in the presence of lobbying by organized foreign interest groups. I show that the FTA in the presence of an organized lobby group in a prospective partner country may cause an increase in the level of protection against imports from third countries and impede trade with non-member countries. I also find that foreign lobby may encourage the local government to enter a welfare-reducing trade-diverting FTA. Finally, I show that the FTA increases the lobbying power of the organized lobby groups of the member countries, which can potentially obstruct the viability of welfare-improving multilateral trade liberalization. The last paper shows that the reason for a higher capital-labor ratio observed for exporting firms is a higher capital intensity of their production technology. Exporters are more productive, more likely to survive and, hence, more likely to repay loans. A higher repayment probability causes creditors to charge lower interest rate and reduces the marginal cost of the firm when a more capital-intensive technology is used. Here, a reduction in international trade costs stimulates exporting firms to use more efficient capital-intensive technologies, while non-exporters switch to less capital-intensive ones. This within-industry change in the composition of technologies reinforces the productivity advantage of exporters and contributes further to industry- wide productivity improvement. The results of model simulations highlight that up to 10% of welfare and productivity gains of trade liberalization come from the adoption of new technologies by existing firms in the industry, thus amplifying the effect of resource reallocation from firms’ entry and exit. ii Table of Contents Abstract……………………………….…………………………………………………………….. ii Table of Contents……………………………….………………………………………………… iii List of Tables……………………………….……………………………………………………… v List of Figures……………………………….……………………………………………………... vi Acknowledgements……………………………….……………………………………………….. vii 1 Introduction……………………………….…………………………………………………….. 1 2 Trade Policy of a Free Trade Agreement in the Presence of Foreign Lobbying…. 3 2.1 Introduction ……………………………….……………………………………………….. 3 2.2 The Role of Foreign Lobbies in Canada….…………………………………………. 6 2.3 The Model………………………………….……………………………………………….. 9 2.4 Data………..……………………………….……………………………………………….. 13 2.4.1 Protection Measures and Market Shares..…………………………………..…….. 14 2.4.2 Political Organization Dummies..………………………………………………….. 15 2.4.3 Elasticities of Substitution……………...…………………………………………... 18 2.4.4 Instrumental Variables…..……………...…………………………………………... 18 2.5 Estimation Procedure………………………….…………………………………………… 19 2.6 Results…… ..………………………………….……………………………………………. 21 2.6.1 Test of a Benchmark GH Model………...…………………………………………. 21 2.6.2 Estimation Results for a Monopolistic Competition Model With Foreign Lobbying.…………………………………………………………………… 22 2.6.3 Robustness Tests…………………...……………………………………………….. 26 2.6.4 Comparison of the Benchmark and Monopolistic Competition GH Models.….. 28 2.7 Conclusion………………………………….……………………………………………….. 29 2.8 Tables and Figures...……….…………….………………………………………………… 31 3 Endogenous Free Trade Agreements and Foreign Lobbying.………………………… 40 3.1 Introduction.……………………………….……………………………………………….. 40 3.2 The Model………………………………….……………………………………………….. 44 3.3 Endogenous FTA..……………..………….……………………………………………….. 51 3.4 Endogenous FTA with Lobbying for Change in a Trade Regime…………………….. 60 3.5 Quantification………………………….…………………………………………………… 64 3.6 Conclusion………………………………….……………………………………………….. 69 3.7 Tables and Figures....…………………….………………………………………………… 71 iii 4 A Model of Trade Liberalization and Technology Adoption with Heterogeneous Firms.……………………………….…………………………………………………………….. 75 4.1 Introduction.……………………………….……………………………………………….. 75 4.2 Empirical Model..………………..…………….…………………………………………… 79 4.2.1 Modeling Strategy..……………………………...…………………………………… 79 4.2.2 Data..………………………………………………………………………………….. 80 4.2.3 Estimation Approach……………...…………………………………………........... 81 4.2.4 Empirical Results…..……………...…………………………………………………. 82 4.2.5 Robustness Tests…..……………...…………………………………………………. 85 4.3 A Model of Technology Adoption and Trade………………………….………………... 88 4.3.1 Demand Side………...……………………………………………………………….. 89 4.3.2 Production Side..…………………………………………………………………….. 90 4.3.3 Capital Market……………….……………...…………….…………………………. 90 4.3.4 The Choice of Technology.…………................................................................... 91 4.3.5 Entry, Exit and Exporting Decisions.………….................................................. 92 4.3.6 Equilibrium Distribution of Productivities.………….......................................... 93 4.3.7 Open Economy Equilibrium.…………................................................................ 94 4.3.8 The Effect of Trade Barriers Reduction.………….............................................. 97 4.4 Quantitative Analysis. ..………………………………….………………………….……. 100 4.4.1 Parameterization………...……………………………………………………………101 4.4.2 Steady State Distribution of Capital Intensities………...……………………….. 102 4.4.3 The Effect of Trade Liberalization………...………………………………………. 103 4.5 Conclusions…………………………………………………………………………………. 106 4.6 Tables and Figures...………………………….…………………………………………… 108 Bibliography....………………………………….………………………………………………… 120 Appendices....………………………………….…………………………………………………… 124 Appendix A: Derivation of the Equilibrium Trade Policy………………………………… 124 Appendix B: Estimation of the Elasticities of Substitution……..………………………… 126 Appendix C: Proofs………………………………………………………..…………………… 130 iv List of Tables Table 2.1 Descriptive statistics for protection measures and market shares, 1997………………………………………….………………… 34 Table 2.2 Descriptive statistics for the number of lobbyists, 1996-97….……………. 34 Table 2.3 Descriptive statistics for the elasticity of substitution and price elasticity……………………………..…………………………………… 34 Table 24 Estimation results for the benchmark GH model with different protection measures…………………………………………………………… 35 Table 2.5 Estimation results for the monopolistically competitive model (10) with foreign lobbying and different protection measures…………………..... 35 Table 2.6 Marginal and average effects of political economy factors on different measures of protection…………………….…………………………………… 36 Table 2.7 Estimation results for the monopolistically competitive model (10) with additional controls……………...………………………………………... 36 Table 2.8 Estimation results for the monopolistically competitive model (10) with additional controls and without elasticity adjustment of the dependent variable.…………………………………………………………….. 37 Table 2.9 Estimation results for a benchmark GH model with import tariff and foreign lobbying being treated as domestic…………………………….. 37 Table 2.10 Estimation results for the monopolistically competitive model (10) with import tariffs and different foreign lobbying structure..……………... 77 Table 2.11 Estimation results for the monopolistically competitive model (10) with import tariffs and foreign lobby being treated as domestic………….. 38 Table 2.12 J-Test for the benchmark and monopolistically competitive GH models… 39 Table 3.1 The effect of an FTA with different lobby structure on the external tariff…………………………………………………………………… 71 Table 4.1 Descriptive Statistics………………………………………………………….. 111 Table 4.2 OLS estimates of the production function for exporting and non-exporting firms……………………………………………………………. 112 Table 4.3 OLS estimates of the translog production function for exporting and non-exporting firms……………………………………………………………. 113 Table 4.4 Robustness of production function estimates……………………………..... 114 Table 4.5 Estimates of the production function for ten largest manufacturing industries……………………………………………………………………….. 115 Table 4.6 Estimation of the production function for current and future exporters… 116 Table 4.7 Estimation of the production function for old and new exporters……….. 117 Table 4.8 Estimation of the CES production function exporters and non-exporters…………………………………………………………………… 118 Table 4.9 The effect of trade barriers reduction, steady state comparison………….. 119 v List of Figures Figure 2.1 Distribution of the number of lobbyists across sectors….…………………. 31 Figure 2.2 Distribution of the number of lobbyists across sectors…………………….. 32 Figure 2.3 Distribution of substitution elasticities………………………………………. 32 Figure 2.4 Comparison of Canadian and US estimates for the elasticity of substitution………………………………………………………. 33 Figure 3.1 Viability of welfare-reducing FTA: a benchmark specification……………... 71 Figure 3.2 Viability of a welfare-reducing FTA with a politically biased government in a partner country (aP = 1)…………………………………... 72 Figure 3.3 Viability of a welfare-reducing FTA with a government’s preference for domestic contributions (bH = 0:5)………………………………………… 72 Figure 3.4 Viability of a welfare-reducing FTA with a more competitive market structure (σ = 5)……………………………………………………………….. 73 Figure 3.5 Viability of a welfare-reducing FTA under cost disadvantage (cH = 1:2)……………………………………………………………………….. 73 Figure 3.6 Viability of a welfare-reducing FTA with domestic and partner country lobbying……………………………………………………… 74 Figure 4.1 The effect of trade openness on technology adoption and productivity…. 108 Figure 4.2 Distribution of capital intensities across French firms (1997-2005)……… 110 vi Acknowledgments I am especially indebted to Matilde Bombardini for continuous support and guidance. For helpful comments and suggestions, I thank Werner Antweiler, Brian Copeland and Vadim Marmer. vii 1 Introduction This dissertation addresses two issues in international trade policy. The rst is to understand how the presence of foreign lobbying a¤ects a countrys trade policy when it forms a free trade agreement (FTA) with another country. Trade policy is traditionally viewed as an outcome of lobbying activ- ities of domestic interest groups, but the recent study by Gawande, Krishna and Robbins (2006) demonstrates that foreign lobbying has a strong potential for trade barrier reductions. However, the behavior of foreign lobbies is considerably altered by FTA membership and is the main focus of the rst two chapters of this dissertation. The second issue surrounds the observation that exporting rms have greater capital intensity than do rms producing solely for the domestic market. The third chapter of this dissertation presents a model of endogenous technology choice that explains this di¤erence and predicts an extra productivity gain from trade liberalization. The rst chapter examines empirically the e¤ect of foreign lobbies on trade policy of a country which is a member of a Free Trade Agreement (FTA). It uses a monopolistically competitive political economy model where external tari¤s are endogenously determined by the government and distinguishes foreign lobbying from within and outside of the FTA. The model predicts that foreign rms from outside of the FTA lobby for trade barriers reduction to facilitate market access, while foreign rms from within FTA lobby for more protectionist trade policy to get competitive advantage over other foreign rms. The model is tested empirically using Canadian industry- level trade data that di¤erentiates lobby groups by the country of origin. The analysis suggests that foreign lobbying has a signi cant e¤ect on the domestic trade policy and its heterogeneity is important: the presence of an organized lobbying group in an FTA partner country tends to raise trade barriers, while an organized lobbying group of exporters from outside of the FTA is associated with less protection. The second chapter provides a detailed theoretical analysis of the model identi ed in the rst chapter, which is further extended by allowing the government to form FTAs endogenously. In this paper I assess the political viability of FTAs and their e¤ect on the world trading system in the presence of lobbying by organized foreign interest groups. I show that the FTA in the presence of an organized lobby group in a prospective partner country may cause an increase in the level of protection against imports from third countries and impede trade with non-member countries. I also nd that the foreign lobby may encourage the local government to enter a welfare- 1 reducing trade-diverting FTA. Finally, I show that the FTA increases the lobbying power of the organized lobby groups of the member countries, which can potentially obstruct the viability of welfare-improving multilateral trade liberalization. The third chapter raises the question of whether di¤erences in capital intensities between rms that export and rms that sell only domestically matter for trade policy outcomes. Using detailed data on French rms, I show that the reason for a higher capital-labor ratio observed for exporting rms is a higher capital intensity of their production technology. Exporters are more productive, more likely to survive and, hence, more likely to repay loans. A higher repayment probability causes creditors to charge lower interest rate and reduces the marginal cost of the rm when a more capital- intensive technology is used. A reduction in international trade costs stimulates exporting rms to use more e¢ cient capital-intensive technologies, while non-exporters switch to less capital-intensive ones. This within-industry change in the composition of technologies reinforces the productivity advantage of exporters and contributes further to industry-wide productivity improvement. The results of model simulations highlight that up to 10% of welfare and productivity gains of trade liberalization come from the adoption of new technologies by existing rms in the industry, thus amplifying the e¤ect of resource reallocation from rmsentry and exit. 2 2 Trade Policy of a Free Trade Agreement in the Presence of Foreign Lobbying 2.1 Introduction In the political economy literature a growing number of studies view trade policy as an endogenous outcome of lobbying activity by special interest groups. Several authors (Goldberg and Maggi, 1999, Gawande and Bandyopadhyay, 2000) have con rmed that lobbying intensity by domestic rms is one of the main determinants of the cross-industry pattern of protection. More recently, Gawande, Krishna and Robbins (2006) also nd that lobbying by foreign rms for trade barriers reduction has a signi cant e¤ect on the structure of tari¤s across industries. However, if a country is a member of a regional free trade agreement (FTA) and foreign rms can a¤ect the governments decision regarding trade policy, it becomes necessary to distinguish foreign lobbying from within and outside of the FTA. Organized foreign interests with preferential market access will lobby for more protection against other foreign rms, and the trade agreement may become more protectionist with a strong lobby group in a prospective FTA partner country. Active foreign lobbying under the preferential trade agreement may not only lead to an increase in trade barriers, but also make welfare-reducing trade agreements politically feasible. In this paper I analyze the e¤ect of foreign lobbying on domestic trade policy when the country is a member of a preferential trade agreement using Canadian post-NAFTA trade data. This analysis reveals two main results. First, the activity of foreign lobbyists in Canada is a signi cant determinant of the Canadian trade policy, and sectors in which foreign rms without preferential market access are politically organized tend to receive less protection. This result supports the previous nding by Gawande, Krishna and Robbins (2006) for the US. Second, NAFTA has an important e¤ect on the structure of foreign lobbies. The data con rm that foreign rms with preferential market access lobby for more protection just as domestic rms do. This result has important implications for the e¤ect of an FTA on a countrys trade policy. It implies that prior to NAFTA, US rms lobbied for Canadian trade barriers reduction like all other foreign rms, but once NAFTA was signed, they switch to lobbying for trade barriers increase. As a result, an FTA with a large and politically strong partner country may raise trade barriers and increase trade 3 distortions, making trade policy of regional trading blocks more protectionist.1 This paper is the rst one that analyzes from theoretical and empirical points of view the e¤ect of foreign lobbying on domestic trade policy in the presence of an FTA. The political economy model presented in this paper incorporates a monopolistically competitive market structure into the Grossman and Helpman (1994) protection for salesetup (henceforth GH) to analyze the role of foreign lobbying in the making of national trade policy. The framework is further extended by allowing for two types of foreign interest groups, namely, lobbying groups formed by rms from an FTA partner country and by rms from countries outside of the FTA, aggregated into the rest of the world (ROW). This di¤erentiation of foreign lobbies by market access is important in the presence of an FTA. When two countries join a regional trade agreement, granting zero import tari¤s to each other, rms from countries with preferential market access will lobby for more protection under the FTA to lock it from competition from the ROW rms, while rms from outside of the FTA will continue to lobby for trade liberalization. This di¤erentiation of foreign lobbying objectives, that follows from the presence of the FTA, implies that in a world where almost every country is a part of at least one regional trade agreement, a complete theory of the e¤ect of foreign lobbing on the national trade system should take this di¤erentiation into account. The model of foreign lobbying in the presence of the FTA is tested using Canadian post-NAFTA trade data. The empirical analysis suggests a strong and statistically signi cant e¤ect of domestic and foreign lobbying on Canadian trade policy, and points to the importance of distinguishing partner country lobbying from ROW lobbying. Using data on lobbying intensity by sector and by country of origin, this paper veri es that the main predictions of the model are consistent with the data. First, the importance of both partner country and ROW foreign lobbying under the FTA is con rmed in the Canadian data: while the presence of the organized domestic lobbying group raises industry import tari¤ by 3-5% relative to the unorganized industry, organized partner country lobbying would raise it by 1-2%, and foreign lobbying from the ROW would lower it by 2-3%. These results are economically meaningful and con rm the main prediction of the model that foreign rms with preferential market access behave just as domestic rms do, which introduces an additional distortion in the policy making process. 1Two considerations should be taken into account when partner country lobbying for more protection is considered. First, the WTO tari¤ binding constrains the lobbying opportunities by the partner country rms; however, they may still play an important role in the future and impede multilateral trade liberalization in those sectors. Second, the WTO precludes countries from raising their tari¤s once the FTA is signed. Yet foreign lobbies may oppose further tari¤ reduction as described above and use anti-dumping and countervailing measures to gain protection. 4 Second, the empirical evidence shows that even politically unorganized sectors receive a positive level of protection from the government of 2-4 percentage points of the ad-valorem tari¤, provid- ing support for the imperfectly competitive structure of the model. Additionally, an important contribution of this work to the political economy literature of trade is its determination of more plausible values for the governments valuation of political contributions. In this paper, the gov- ernment is estimated to value political contributions more than national welfare, which is in sharp contrast with previous tests of the benchmark GH speci cation in which governments were found to have stronger preferences for welfare (Goldberg and Maggi (1999), Gawande and Bandyopadhyay (2000)). This result explains why relatively small political contributions may have strong policy e¤ects. My empirical results provide new perspective on the e¤ect of regional trade blocks on trade policies of member countries. Most of the literature on endogenous trade policy concludes that a country is more likely to reduce its external tari¤when it enters an FTA. Richardson (1993), Bagwell and Staiger (1997), Bohara, Gawande, and Sanguinetti (2004), Bond, Riezman, and Syropoulos (2004) show that a welfare-maximizing government will lower external import tari¤s once an FTA is formed. By doing this, the government restores part of the tari¤ revenue lost due to the shift in import demand from the ROW to the partner country rms. Ornelas (2005a,b) examined the political economy of an FTA without foreign lobbying using an oligopolistic market structure in the GH model. He shows that FTA formation weakens the lobbying power of domestic rms because the elimination of tari¤s between FTA member countries shifts part of the tari¤ rent from domestic rms towards rms from a partner country. The model in this paper, however, allows for cross-border lobbying activity, which lessens the e¤ect introduced by Ornelas (2005a,b) since the reduction in political activity by domestic rms is coupled with an increase in contributions for protection by the FTA partner country rms. As a result, in the presence of foreign lobbying a countrys government may in fact want to raise external tari¤s under the FTA when strong lobbying by an FTA partner country puts extra pressure on the government for higher trade barriers. Activities of foreign interest groups in national policy-making have received a growing attention in the political economy literature. Many scholars have evaluated the intensity of foreign lobbying in the US and argue that it has high potential for policy inuence. Mitchell (1995) found that foreign a¢ liates in the US contributed 5.6% of total corporate political contributions in 1987-88, 5 and 42% of them hired professional lobbyists to promote their interests in Washington. Hansen and Mitchell (2000) claim that, although foreign corporations make lower political contributions than domestic ones due to the existing legal restrictions, they are just as intensive as domestic corporations with respect to lobbying activity and lobbying expenditures. While it has been argued that foreign corporations are quite inuential in US politics, very little research has been done on their e¤ect on trade policy outcomes. In their pioneering work, Gawande, Krishna, and Robbins (2006) demonstrate that foreign agents lobbying expenditure in the US is even greater than political contributions by domestic corporations, and that the elasticity of the US import tari¤ with respect to foreign lobbying is almost as big as with respect to the domestic one. Their paper was the rst to show that foreign lobbying is an important factor in the formation of national trade policy and argue that foreign lobbying may be bene cial to the countrys trade policy as a counter-pressure to domestic interests, helping to reduce distortional trade barriers. In contrast, my paper suggests that the above argument is invalid in the presence of preferential trade agreements, when rms from a partner country prefer to maintain high discriminatory tari¤s for third-country imports. Since most of the countries are members of at least one preferential trade agreement, this paper demonstrates that cross-border lobbying may stimulate more protectionist trade policies of FTAs and disrupt multilateral trade liberalization. The paper is structured as follows. Section 2.2 provides some evidence on the lobbying activities of foreign rms in Canada. Section 2.3 introduces the modi ed GH version of the model that allows imperfectly competitive market structure and two groups of foreign lobbies, and motivates the empirical methodology. Section 2.4 describes the data and Section 2.5 explains the estimation procedure. Results are presented in Section 2.6 and Section 2.7 concludes. 2.2 The role of foreign lobbies in Canada The rst law that regulates the activity of domestic and foreign lobbyists in Canada is the Canada Elections Act, introduced in 1960. This law regulates the amount of political contributions by Canadian nationals to political parties and explicitly bans the use of political contributions from foreigners.2 2No person or party shall accept or use contributions from a person who is not a Canadian citizen . . . , corporation or association that does not carry on business in Canada, foreign political party, or foreign government,(S.217). 6 However, this law su¤ered from a lack of transparency in the lobbying process and on September 30, 1989, a Lobbying Registration Act (LRA) came into force. This piece of legislation introduced a de nition of a lobbyist and a requirement for lobbyists to register with the Lobbyists Registrar. But more importantly, it requires lobbyists to provide information about the name and business address of the organization that has a direct interest in the outcome of the lobbyistsactivities on behalf of the client,all of its subsidiaries and corporate headquarters, if there is one. An amendment to the LRA, introduced in 1996, made this information publicly available. It also introduced a strict disclosure of funds policy applied to political parties. Together with the Canada Elections Act, the LRA made it di¢ cult for foreign rms to lobby their interests in Canada directly. Nevertheless, politically active foreign rms can still inuence trade policy outcomes in at least two ways. First, they can hire Canadian agents and consultants to lobby the executive branch on their behalf and a¤ect policy outcomes in a way that suits the interests of foreign rms. Second, subsidiaries of foreign enterprises can make legal political contributions with their own funds to defeat legislators who are unfriendly to their interests. Since there are no restrictions on the share of foreign capital in the assets of a company that makes political contributions, any local subsidiary of a foreign corporation can make political donations from its own funds if it carries business in Canada.Moreover, almost any big foreign company that exports to Canada has an independent local sales department, which is legally allowed to lobby for a reduction in trade barriers on the products imported by its parent company into Canada. Lobbying e¤orts of such subsidiaries will be counter to the e¤orts of domestic rms and, therefore, pooling all Canadian rms together regardless of their ownership may lead to misleading results and estimation problems. In the trade policy literature, corporate political activity is typically measured by nancial contributions to candidates and political parties, while very little attention has been paid to other means of a¤ecting policy outcomes such as direct lobbying. However, earlier research on the ef- fect of foreign companies on national policy (Hansen and Mitchell, 2000) suggests that foreign corporations prefer direct lobbying to political contributions not only because of legal restrictions on contributions, but also because of informal legitimacy questions for politicians with respect to accepting money from corporate sources with foreign ownership. Hansen and Mitchell found that because lobbying is less visible than contributions, foreigners use it as intensively (and e¤ectively) as domestic rms do. For these reasons, lobbying expenditures seem to be a better measure of for- eign political involvement, especially in countries with legal restrictions on political contributions 7 by foreigners. Unfortunately, data on lobbying expenditures by domestic and foreign corporations are unavail- able for Canada. To measure the foreign lobby intensity in Canada, I collected the information on the number of lobbyists o¢ cially registered with the O¢ ce of the Registrar of Lobbyists, as is required by the LRA.3 Intuitively, the number of lobbyists hired by a company or industry should be highly correlated with the lobbying expenditure and hence can serve as a proxy for political ac- tivity. The lobbyists registration data is publicly available and is discussed in more detail in Section 2.4.2, but the following gures illustrate the relative importance of foreign lobbying in Canada. In 1996-97, there were 1,032 o¢ cially registered lobbyists representing interests of manufacturing rms regarding Canadian trade policy, with 47%, 26% and 27% of them acting on behalf of Canadian, US and ROW rms, respectively. This gures highlight the potential strength of foreign lobbyists in Canada and suggest that the number of contacts of foreign rms with Canadian policymakers was at least not smaller than that of domestic rms. Then, when almost every country in a world trades under preferential agreements, it is important to consider the e¤ect of foreign lobbying under the FTA when a group of foreign rms with preferential market access may lobby for more protection and reinforce trade diversion e¤ects. Although the theoretical models in this and the next sections rely on political contributions by domestic and foreign rms, foreign rms can only use direct lobbying to protect their interests in Canada. Therefore, there is no direct link between foreign lobbying expenditure and the objective function of policymakers, as it is the case in the benchmark GH model. However, there are still several reasons why the government may value lobbying expenditures by domestic and, especially, foreign rms. For example, policymakers may value information embodied in services produced by consultant lobbyists as legislative proposals produced by lobbyists for the government may save politicians time and resources necessary to collect this information. Alternatively, we can think of consultant lobbyists as being a part of the executive branch, and even though they do not participate in the voting process, they are very inuential at the stage of legislation drafting and have become an integral part of the government system. 3The Act de nes a lobbyist as an individual who, for payment, undertakes to lobby on behalf of a client and represents an organization in arranging meetings with public o¢ ce holders, or communicate with a public o¢ ce holder in an attempt to inuence the development of any legislative proposal, ... the making or amendment of any regulation, ... the development or amendment of any program or policy. 8 2.3 The model The theoretical model is based on the Grossman and Helpman (1994) political economy model and presents several modi cations that allow for the presence of foreign lobbying and facilitate econometric estimation. In their original formulation, Grossman and Helpman considered a small open economy, which leaves no room for foreign companies to lobby because pre-tari¤ prices are xed. In this work I develop and build into the GH setup a model of monopolistic competition with di¤erentiated goods to allow foreign rms to gain or lose from import tari¤s. There are N products (industries) in the model and three countries: Canada (Home country), the US (FTA Partner country) and the ROW, denoted by H;P and ROW , respectively. Industries are denoted by index i 2 f1; :::; Ng and countries by j 2 fH;P;ROWg. There are nji rms in country j and industry i. These rms are assumed to be symmetric within the same country and industry, i.e. they share the same cost structure and hence face the same demand functions and charge the same prices. In total, there are nHi + n P i + n ROW i di¤erent varieties of each product i. A representative consumer maximizes a quasilinear utility function with a constant elasticity of substitution index nested into a Cobb-Douglas function: U = X0 + NX i=1 !i lnXi Xi = nHi d H 1 i i x H i 1 i i + n P i d P 1 i i x P i 1 i i + n ROW i d ROW 1 i i x ROW i 1 i i ! i i 1 (1) where Xi is an aggregate consumption index for product i, !i is the share of product i in the total consumers expenditure, xji is the demand for product i produced in country j, i > 1 is the elasticity of substitution between varieties of product i,4 and dji is a country-wide taste (or quality) parameter for product i imported from country j. Maximizing (1) subject to the standard budget constraint, we obtain the demand functions and an aggregate price index for product i: Xi = !i (Pi) 1 ; i 1 (2) xji = !id j i pji pji Pi !1 i (3) 4Substitution elasticity between industries is assumed to be equal to unity. 9 Pi = nHi d H i (p H i ) 1 i + nPi d P i (p P i ) 1 i + nROWi d ROW i (p ROW i ) 1 i 11 i (4) Firms within one country and sector are assumed to have the same constant marginal cost. This al- lows us to consider Canadian market independently from other markets, i.e. prices on the Canadian market depend only on the demand elasticity, the ( xed) number of rms and the xed marginal cost structure. Denoting a speci c import tari¤ set by the home country government on imports of product i from country j as ji , we can write the pro t of a country j rm that produces product i as: ji = (p j i cji ji )qji (5) where qji is the quantity supplied and c j i is the marginal costs. I assume that the number of rms is large enough to ignore the e¤ect of their individual pricing decisions on the industry price index Pi, i.e. each rm takes the price index as given. Knowing product demand functions (3), each rm sets the pro t-maximizing price as a markup over its marginal costs: pHi = i i 1 cHi ; p P i = i i 1 (cPi + P i ); p ROW i = i i 1 (cROWi + ROW i ) (6) For convenience, isolate costs from (6) and write down equilibrium pro ts (5) as: ji = 1 i p j i q j i , 8j (7) The government chooses import tari¤s to maximize a weighted sum of national welfare W and political contributions C: G( j ; Cj) = X i CHi + aW + b X i CPi + c X i CROWi (8) where CHi , C P i and C ROW i are industry-wide political contributions from each country. Coe¢ cient a is a weight that the government assigns to national welfare relative to political contributions. Coe¢ cients b and c reect the governments preferences for the US and ROW contributions, re- spectively, over the contributions by domestic rms. As long as accepting contributions from foreign rms involves risk of reputation loss or law infraction, politicians may prefer domestic contributions to overseas donations thus both coe¢ cients are presumably less than one. Firms in industry i can organize themselves and form a group to lobby the local government for 10 a change in trade policy.5 Firms within the FTA pay no import tari¤s and hence lobby for more protection, while rms from other countries lobby for lower tari¤s for the opposite reason. The lobby representing industry i of country j maximizes its welfare from obtaining protection net of political contribution: (W ji Cji ). As in Grossman and Helpman (1994), the equilibrium trade policy is a solution to a two-stage game. In the rst stage, knowing the governments objective function, each organized lobbying group provides the government with a schedule of political contributions as a function of import tari¤. In the second stage, observing contribution schedules, the government sets trade policy that maximizes its objective function (8). Grossman and Helpman (1994) show that for truthful contribution schedules6 the optimal trade policy is the one that maximizes joint surplus of the government and organized lobbying groups. Let i denote the share of the home country population entitled to the domestic industry i pro ts, and Iji denote an index variable that takes the value of one when industry i in country j is politically organized and zero otherwise. The joint welfare function then takes the form: = NX i=1 IHi W H i + aW + b NX i=1 IPi W P i + c NX i=1 IROWi W ROW i (9) where WHi = n H i H i + i(TR + CS) is welfare of the domestic industry i gross of political con- tributions, TR and CS are total tari¤ revenue and consumer surplus, respectively, W ji = n j i j i ; j 2 fP;ROWg is gross welfare of foreign industries i from exports to the home county market, and W = P i(n H i H i ) + TR + CS is national welfare. Taking the rst order condition of the joint welfare function with respect to the ROW import tari¤ rate and rearranging it, one obtains the expression for the equilibrium trade policy:7 "i ROWi pROWi = 1 i + (i 1) P i pPi sPi + a a+ i 1 i sHi + 1 a+ i 1 i IHi s H i + (10) + bIPi a+ i 1 i sPi + cIROWi a+ i 1 i sROWi 1 where sji = njip j ix j i PiXi denotes the share of country j rms on the Canadian market for product i. On 5With the number of rms in the sector being limited by the endowment of sector-speci c capital, rms in each industry have an incentive to form a lobby group and seek for protection from foreign competition. Here I ignore the free-riding problem within each sector. See Bombardini (2005) for an extensive discussion of rm-level contribution decision. 6Contribution is de ned as truthful if it reects the true preferences of lobbying group for any possible policy outcome. 7Details on derivation of equation (10) are provided in Appendix A. 11 the left-hand side of (10), ROWi =p ROW i is the ad-valorem tari¤ on the ROW imports, which is multiplied by the price elasticity of demand for the ROW imports "i. Therefore, as in the benchmark GH model, trade protection is inversely related to the import demand elasticity.8 The rst term on the right-hand side is negative: the model predicts that with more di¤erentiated varieties will receive import subsidy. This result is a direct consequence of monopolistic competition model with speci c import tari¤.9 The second element on the right-hand side shows the positive relation between the FTA external and internal tari¤s and reects a tari¤ complementarity e¤ect: if the tari¤ rate for the partner country is high, it is optimal for the government to raise the external tari¤ as well.10 Intuitively, an increase in the within-FTA tari¤ rate causes a decline in imports from the ROW, and tari¤ revenue collected on the ROW imports is higher for higher ROW . This, in turn, rises imports from the partner country, that generates more tari¤ revenue for higher partner country tari¤ rate. The tari¤ complementarity e¤ect is proportional to the market share of the partner country rms sPi and is stronger if the partner country and the ROW exports are close substitutes. In contrast to the benchmark case, even for unorganized industries protection may still be positive due to the imperfectly competitive market structure, as emphasized by the third term, since the coe¢ cient aa+ is positive. Because the share of domestic rms on the market reects their ability to capture protection bene ts, the tari¤ level is proportional to sHi and increasing with . The fourth term is similar to the benchmark GH model: a politically organized domestic industry receives more protection from the government. Moreover, the level of protection is higher if domestic and imported varieties are close substitutes and if the domestic sector is relatively large, as the domestic lobby has more to gain from protection in this case. The fth term reects the e¤ect of political activity by partner country rms on the national trade policy. Ipi enters the equation positively, making protection more likely in those sectors where partner country exporters are organized into lobbying groups and where product varieties are closer substitutes. Similarly to the domestic lobby, the e¤ect of partner country rms lobbying on the 8 It should be noted that without the MFN rule, a set of equilibrium tari¤s for all importers would be determined by a system of simultaneous equations with the number of equations being equal to the number of importing countries. With the MFN and the FTA, the number of equations goes down to two. However, under complete trade liberalization agreement, a within-FTA tari¤ is exogenously set to zero and the second term on the right-hand side of (10) vanishes. 9 In the model of monopolistic competition, the e¤ect of a speci c tari¤ on price is ampli ed by producers markup. Therefore, for low the price elasticity with respect to tari¤ is high and the gain in consumer surplus from a subsidy outweighs the increase in governments expenditure. 10This result is consistent with other studies, e.g. Bagwell and Staiger (1997), Ornelas (2005a,b). 12 import tari¤ is proportional to their market share. The last term is negative and reects the e¤ect of lobbying e¤orts by the ROW rms to reduce protection. As before, the scaling factor i 1i reects higher motivation by the ROW rms to lobby for trade liberalization when the degree of substitution between varieties within a given industry is high, but unlike domestic and partner country lobbying, the ROW lobbying intensity declines with the market share. The intuition behind this result is an increased damage from protection for small ROW industries, and as a consequence these industries will resist tari¤ increase more intensively.11 As in the GH model, domestic and partner countrys lobbying results in overprotection and welfare reduction relative to the rst-best outcome. The presence of an organized foreign lobby from the ROW may help to (partially) restore the optimal level of import tari¤s and raise national welfare. However, the presence of the ROW lobbying alone causes underprotection and is thus welfare-reducing. Therefore, the overall net e¤ect from the presence of the partner country lobbying is likely to lead to welfare reduction, whereas the overall welfare e¤ect of the ROW lobbying activity is unambiguous: the e¤ect is positive if ROW rms counter-lobby against the e¤ort of domestic and partner country rms to raise protection and negative if ROW rms form a single organized lobbying group in the sector. 2.4 The data The empirical section of this paper estimates the e¤ects of domestic, partner country and ROW lobbying activity on the Canadian post-NAFTA trade policy. Primarily, I use two measures for trade barriers: import tari¤s and the share of imports that is subject to tari¤ or non-tari¤ trade restrictions. The US was treated as a Canadian FTA partner country, while all other countries that have no preferential trade agreements with Canada were aggregated into ROW. The estimation of equation (10) requires the following data: the measure for trade protection, imports by the country of origin and by sector, domestic output by sectors, substitution and price elasticities, political organization dummies, and three sets of instruments for market shares. This study is conducted for 249 Canadian 6-digit NAICS manufacturing sectors (NAICS 31-33) for the period of 1996-97. 11This result follows from the Cobb-Douglas utility function. Fixed product expenditure shares imply that the import tari¤ imposed on one variety will raise consumers expenditure on all varieties through aggregate price index proportionally to their market shares. Therefore, the higher is the ROW market share (and the lower is the share of other varieties), the less harmful is the import tari¤ for the ROW exporters. 13 2.4.1 Protection measures and market shares Domestic manufacturing shipments data for 249 NAICS-6 industries are provided by Industry Canada. The values of Canadian imports, as well as customs duties collected, were obtained from Statistics Canada at the HTS-10 level and aggregated to NAICS-6 using the concordances tables from the International Trade Division of Statistics Canada. In the original formulation, the GH model was meant to analyze the political economy of import tari¤ formation. However, tari¤ rates are often argued to be an imperfect measure of trade protection for the analysis of endogenous trade policy formation in the presence of WTO tari¤ regulation. With limitations on the magnitude of tari¤s imposed by the WTO, organized interests would seek non-tari¤ protection from import competition. The advantage of non-tari¤ barriers (NTBs) as a measure of trade distortion is their unilateral adoption by di¤erent countries, as opposed to tari¤s that are set cooperatively in WTO negotiations. Unless organized industries can protect their interests during WTO negotiations, NTBs are a preferred measure for trade protection. However, measurement di¢ culties and the qualitative nature of NTBs are important issues, which can be a serious problem in a small sample estimation. In light of this, I used tari¤, NTBs and protection coverage share as a measure of protection. Ad- valorem tari¤ rates were obtained as the ratio of aggregated duty collected by customs over the value of imports.12 NTBs for Canadian imports were obtained from the TRAINS database maintained by UNCTAD, which shows the proportion of imports that is covered by one or more qualitative restrictions. These data were available at the HS-6 level and were aggregated into NAICS-6 groups. In addition, the protection share variable was constructed as the share of Canadian imports that is subject either to the positive import tari¤ or NTBs. Descriptive statistics for protection measures and market shares are presented in Table 2.1. In 1997 the average tari¤ rate, NTBs and protection coverage ratios for the ROW imports were 4.8%, 18.2% and 77.5%, respectively. Tari¤s and NTBs are highly correlated both within and outside of the FTA, which implies that di¤erent measures of protection are still highly complementary. 12Therefore, tari¤ measure controls for some non-tari¤ distortions as well, such as antidumping or countervailing duties. 14 2.4.2 Political organization dummies Previous studies that have tested the GH model empirically used rm-level political contributions to assign the value for the political organization dummy variable.13 Although these data are available for Canada for 1997 and afterwards, this paper uses a di¤erent approach. As was previously men- tioned, foreign corporations prefer direct lobbying to political contributions because transparency of political contributions may raise concerns about foreign interference into political processes. Fur- thermore, since di¤erent means of political involvement are highly correlated (Hansen and Mitchell, 2000), direct lobbying seems to be an appropriate measure for domestic political activity as well. In this work, the degree of political activity in an industry is measured by the number of lobbyists representing the corporate interests of that industry. The Lobbyists Registration Act (LRA) requires every individual to register at the Lobby Registrar Canada if the person seeks a meeting or a phone call to any public o¢ ce holder regarding the development, modi cation or cancellation of legislative proposals, regulations, public policies and programs. The assumption that political contributions will be ine¤ective for the determination of trade policy without such contact seems to be reasonable and, therefore, political contributions should be followed up by a personal contact with a policymaker. For this reason, the number of registered lobbyists is used to measure rm-level lobbying intensity within an industry. The main advantage of this data set is the large amount of detailed information lobbyists are required to submit. This includes information on the business address of a corporation that bene ts from lobbying, its subsidiaries and headquarters, and the objective of the meeting with a public o¢ ce holder. This information is very helpful in determining the nationality and industrial a¢ liation of lobbyists representing interests of multi-product multinational corporations. Another advantage of this data set is that it gives a very narrow de nition of a lobbyist. Any person representing his or her own interests, and who is not being paid for arranging the meeting with the public o¢ ce holder, is not obliged to register. This removes information on the very small rms. Large rms, which have high lobbying power and can e¤ectively inuence the decisions of policymakers, typically use the service of professional consultants or corporate lobbyists, who are required to register. 13See, for example, Goldberg and Maggi (1999), Gawande and Bandyopadhyay (2000), Bombardini (2005), Gawande, Krishna, and Robbins (2006). 15 Firms were assigned a NAICS-6 industry code using the Canadian Company Capability data- base maintained by Industry Canada. Assigning an industry code to multi-product rms involves some degree of discretion. For example, some rms in the automobile sector operate in more than ten NAICS-6 industries. Since the number of such rms is relatively small, I assigned primary, sec- ondary and tertiary NAICS codes to such rms using di¤erent information sources: the Canadian Company Capability database, the Federal Corporations Registry and the North America Compu- stat database. The databases listed above allow assigning industry codes to US and ROW rms. The LRA also requires lobbyists to declare a subject-matter in respect to which an individual undertakes to communicate with a public o¢ ce holder.In many cases, information on the purpose of lobbying activity reported in the lobbyist registration form allowed me to attribute a rm to a single (or a small number of) primary NAICS code. National a¢ liation of each rm that a particular lobbyist is representing was determined from two sources. First, the lobbyist registration form requires registrants to provide the name and business address of the parent corporation and those subsidiaries which directly bene t from the lobbying. Sometimes, lobbyists provide incomplete information and in this case it was comple- mented by the information from other databases mentioned previously. Again, quite often the nationality of the rm was determined by the objective section of the lobbyist registration form.14 There are two more advantages of using lobbyistsregistry data over using political contribution data. First, the necessity of reporting the objective of the contact with the public o¢ ce holders allows one to isolate e¤ectively those rms that lobby particularly for a change in trade policy. This is especially a problem for domestic lobbyists: on average, only one out of eight lobbyists, representing the domestic manufacturing sector, is concerned with trade policy. Therefore, pool- ing political contributions by all domestic rms may cause serious measurement problems for the political organization variable. Second, the main channel used by foreign rms to lobby their interest in Canada is through local subsidiaries, which distribute the imported goods within Canada. Formally, these rms should be assigned to a trade sector code (NAICS 41-45) and dropped from the sample, but this would substantially underestimate lobbying e¤orts by foreign rms. For each trade rm concerned with 14For instance, a lobbyist of the Japanese Automobile Manufacturers Association of Canada registered in Ontario, Canada was attributed to the ROW on the basis of the meeting purpose to secure international trade for electric equipment. 16 international trade policy issues I used the company pro le and lobbying objectives information to assign an appropriate manufacturing industry code and country. The amendment to the LRA announced in 1995 introduced several important re nements that made it more desirable to use post-1995 lobbying data. First, for the purpose of transparency, the lobbyistsconnections to the public o¢ ce holders were opened to public scrutiny, and the lobbyists registry database became available for research purposes. Second, this amendment extended the amount of information that must be reported. Most importantly, it made disclosed information more complete and reliable. For the rst time lobbyists were obliged to provide all the information, and e¤ective enforcement devices were introduced to encourage better compliance. It extended the power of the Lobbyists Registrar, which was authorized to seek clari cation of information submit- ted. The registrar was allowed to conduct an audit of provided information and, when necessary, investigate the provided information. Sanctions for violating the LRA were also increased: every individual who contravenes any part of the Act . . . is liable to a ne not exceeding twenty- ve thousand dollars. On proceedings by way of indictment, an individual is liable to a ne of up to one hundred thousand dollars or to imprisonment for a term not exceeding two years, or to both (LRA, Section 14). For these reasons, the data for political activity by rms were collected for the 1997 election cycle and complemented with the 1996 lobbying data to take into account any possible small lag in trade policy response to lobbying e¤orts. Similarly with regard to other studies, several thresholds for the number of lobbyists in an industry will be set to determine the values of political organization dummies. As a nal note on the lobbyistsregistry data, I will briey discuss the reliability of this database. In 2001 an independent study of compliance to the LRA was conducted by KPMG Consulting Inc. (2001). The already-registered lobbyists were asked if they were aware of any non-compliance behavior. Reported results indicate that 50% of consultant lobbyists and 15% of corporate lobbyists were aware of non-registered lobbying, while they evaluated the aggregate compliance rate at 70% and 100%, respectively. In general, compliance was perceived to be high, although non-compliance behavior is still an important issue. Descriptive statistics for the number of lobbyists is provided in Table 2.2 and the distribution of lobbyists across sectors is shown in Figure 2.1. 17 2.4.3 Elasticities of Substitution To my knowledge, there are no studies to date that estimate substitution elasticities for Canadian NAICS-6 industries, especially within a framework of monopolistic competition. In this study, substitution elasticities were estimated using the approach by Feenstra (1994), recently applied by Broda and Weinstein (2006) to a large set of US imported commodities. This approach identi es supply and demand elasticities using no instrumental variables, and relies only on the assumption of independent supply and demand disturbances. The complete derivation of the estimator is presented in Appendix B. It also includes a discussion of a possible estimation bias due to a small sample size and a proposed solution to that problem. The summary statistics and a histogram of the elasticity estimates are presented in Table 2.3 and Figures 2.2 and 2.3. Price elasticities of the ROW import demand were calculated using (9a). As a robustness check, I estimated the substitution elasticities for NAICS industries at various level of aggregation and veri ed that more aggregated commodities are more di¤erentiated: the average value of decreased from 5.85 to 5.34 and 4.56 while moving respectively from six to ve and four digits NAICS. As another robustness check, I estimated US elasticity of substitution using the same procedure for the same year and industry classi cation. Presumably, Canada and the US should have similar tastes, and varieties that are close substitutes in Canada should be close substitutes in the US as well. This suggestion is supported by Figure 2.4, which compares the percentage deviation from the mean for Canadian and US estimates and shows a very close relationship in the perception of product characteristics in Canada and US. 2.4.4 Instrumental variables In equation (10), market shares are likely to be determined simultaneously with the tari¤ rates and should be properly instrumented. Trezer (1993) proposed to instrument the import penetration ratio with industry factor endowments as the measure of comparative advantage independent of the level of protection. Following this approach, a list of instruments for the Canadian market share includes: the share of production to non-production workers, the share of capital stock in machinery and construction in the output, and the share of fuel and electricity in total costs. All of these data are provided by Statistics Canada. The same list of instruments was constructed for 18 the US market share in Canada using the US Census data.15 To instrument the ROW share in the Canadian market, the gravity-type distance measure between Canada and the average exporter was constructed. For every product, the pair-wise log- distance between Canada and the exporting country was weighted by the share of this country in the global export of the product.16 The total exports by country and by sector were constructed using the UNCTAD database. The data on geographic distance, weighted by population density and economic activity within each country, were taken from the Centre dEtudes Prospectives et dInformations Internationales. The rationale for using this distancevariable is the following: if main producers of a particular good are located far from Canada, transportation costs are high and the ROW share in the Canadian market is likely to be small regardless of Canadian trade policy. Political organization dummies are likely to be measured with error and are potentially endoge- nous. To instrument the US and Canadian political organization dummies, I use the information industrial concentration, such as shares of large and medium rms and the CR-4 concentration ratio. Industrial concentration is mostly technologically determined and at the same time it is easier for rms in more concentrated industries to overcome free-riding problem and form a lobby group. The ROW lobbying intensity is instrumented similarly to Gawande, Krishna, and Robbins (2006) with the ratio of exports by ROW rms to Canada relative to their worldwide exports in an industry.17 2.5 Estimation procedure Equation (10) motivates the following form of the estimation equation: Yi = 0 + 1s h i + 2I h i s h i + 3I p i s p i + 4I f i [1 sfi ] (11) Yi = i i 1 "i fi pfi + 1 i ! 15 In Chapter 4 of this dissertation I argue that industry capital intensity and, hence, factor proportions are a¤ected by trade policy. However, the relationship that I identify in Chapter 4 holds only in a long-run and factor proportions are still mainly technologically determined. 16Since the exporters share on the Canadian market is endogenous, I use the share on the global market, assuming it is una¤ected by Canadian tari¤ rate. 17Since this share may be correlated with import tari¤s and NTBs, it may not be valid instrument. In future I will use the concentration measure for the ROW exporters. Although I treat all foreign countries from outside of the FTA as a single agent, in practice it is easier for foreign rms to organize into lobby group if production is concentrated in a small number of countries rather than spread across large group of countries. 19 1 = a a+ ; 2 = 1 a+ ; 3 = b a+ ; 4 = c a+ (12) In equation (10) the inverse elasticity was taken on the left-hand side and both sides were multiplied by ii 1 because substitution elasticity is likely to be measured with error. Using (12), the four coe¢ cient estimates of the reduced form (11) can be used to derive four structural parameters of the model. There are several issues regarding the estimation of (11) that should be addressed. First, the right-hand side variables of (11) include non-linear combinations of endogenous variables. To deal with this problem the set of instrumental variables includes the ones described in the Section 2.4.4 above, their square terms and interactions.18 Second, using many instruments relative to the sample size and to the number of instrumented variables leads to a bias of 2SLS towards OLS results (Hansen, Hausman and Newey, 2006). The third problem with a small ratio of endogenous variables to the number of instruments is a problem of weak instruments. When the correlation between the endogenous variable and most of the instruments is weak, conventional asymptotics may provide poor approximations for the reduced form estimates and test statistics, and inferences on the signi cance of coe¢ cients may be very misleading. Finally, Goldberg and Maggi (1999) point to the problem of heteroskedasticity of the error term in (11), which may be correlated with the imperfectly measured elasticity of substitution. Therefore, equation (11) requires an estimator that will be robust to the problems of many instruments, many weak instruments and heteroskedasticity. Several estimation procedures that address these problems were proposed recently in the econometrics literature.19 I used LIML with Bekker (1994) standard error correction. Hansen, Hausman, and Newey (2006) demonstrated that this approach has better small sample properties than 2SLS and is asymptotically correct in the presence of many instruments and many weak instruments. It results in tests with the correct size and the tests can be made robust to heteroskedasticity through the use of conventional robust covariance matrix estimators. 18This IV estimator was proposed by Kelejian (1971). 19Wright, and Yogo (2002) and Andrews and Stock (2005) provide an excellent survey of this literature. 20 2.6 Results 2.6.1 Test of a benchmark GH model As a starting point, I will present the results on a GH version of the model with homogeneous goods to test how well the new data on Canada can t the benchmark model and compare its performance with the results of previous empirical studies. Since in the benchmark model markets are perfectly competitive and import supply is in nitely elastic, there is no reason for foreign rms to participate in lobbying, and in the benchmark case I will consider only the e¤ect of domestic lobbying groups on the home country trade policy. Following Grossman and Helpman (1994), the optimal import tari¤ for a small open economy with politically organized sector-speci c factors of production takes the following form: "fi fi pfi = a+ Xhi Mi + 1 a+ Xhi Mi Ii (13) where Xhi is a domestic value of shipments and Mi is a total value of imports. The model predicts that the inverse import penetration ratio enters the equation negatively, while the coe¢ cient on its interaction with political organization dummy is positive. Table 2.4 represents estimation results for equation (13) using tari¤s, NTBs and protection share data as a measure of Canadian trade barriers. As in Goldberg and Maggi (1999), several thresholds were used for the number of lobbyists in the construction of political organization dummies to verify that the results are not driven by the way these dummies are assigned. In the rst column of Table 2.4 an industry is considered to be politically organized if it is represented by at least one lobbyist. For the second and third columns the threshold is two and three lobbyists, respectively. First, consider the estimation with tari¤s as a protection measure. The estimates of the regres- sion model (13) are of correct signs across all speci cations: politically organized sectors receive more protection, while protection in unorganized sectors is negative and increases with import pen- etration. The latter result is statistically signi cant at a 5% con dence level. In organized sectors protection declines with the import penetration, but this result is not statistically signi cant. The model estimates are very robust to the way political organization dummies are constructed. When 21 trade barriers are measured with NTBs and protection share, results correspond closely to those obtained for the tari¤ equation: the parameter estimates preserve correct signs, but are estimated with less precision when trade barriers are measured with protection share. Overall, the estimates of the structural model (13) using Canadian data are generally in the line with the results of the studies by Goldberg and Maggi (1999) and Gawande and Bandyopadhyay (2000). The estimates of the structural parameters of the model vary considerably across di¤erent measures of protection. However, the variances of these parameters are very high and one cannot reject hypotheses that both and a are the same across all speci cations considered. The fraction of the population represented by a lobby, , is estimated to be around 0:6 for the speci cations with tari¤s, 0:2 for speci cations with protection shares, and greater than one for speci cations with NTBs. In general, 95% con dence interval for includes the whole [0; 1] interval and the model does not allow one to obtain a precise measure for . Nevertheless, the obtained results do not contradict the previous estimates of by Goldberg and Maggi (1999) and Gawande and Bandyopadhyay (2000), who estimated the share of the population represented by interest groups to be around 0:85 and unity, respectively. The estimates of the governments political bias vary from 10 in the speci cation with shares to 100 in speci cation with tari¤s. The values of the parameter a greater then ten imply that the government assigns approximately equal weights to political contributions and to a national welfare net of political contributions, which supports the results of studies that use US data.20 Overall, the results of the GH model with Canadian data are broadly consistent with those obtained by Goldberg and Maggi (1999) and Gawande and Bandyopadhyay (2000) and other studies for the US. These results will serve as a benchmark against which the results of the monopolistic competition model with foreign lobbying and FTA participation will be compared in the next section. 2.6.2 Estimation results for the monopolistic competition model with foreign lobby In this section I present the estimation results for the political economy model of trade with monopolistic competition, FTA membership and two groups of foreign lobbies. The results from 20The governments objective function C + aW is equivalent to a1C + a2 (W C), in which a = a2a1 a2 , a1 is the weight on political contributions and a2 is the weight on a welfare net of political contributions. Therefore when a is much greater than one, a1 = a+1 a a2 a2. 22 the equation (11) appear in Table 2.5, in which several measures are used to measure trade barriers. Columns with di¤erent numbers denote di¤erent threshold levels for construction of the political organization variable. Columns (1) and (2) report the results when an industry is assumed to be politically organized if it has at least one and three lobbyists, respectively. In column (3) a countrys j industry is organized if it is represented by at least three lobbyists and accounts for strictly more than one third of a total number of lobbyists in that industry. This measure was constructed to exclude sectors with a positive but small number of lobbyists relative to the whole industry. First, note that for any measure of protection and political organization, the coe¢ cient 1 is positive and almost always statistically signi cant, implying that domestic industries receive a positive level of protection regardless of political economy factors. This is consistent with the prediction of the model that the welfare-maximizing government always nds it worthwhile to protect home country producers against competing importers when domestic and foreign products are close substitutes and markets are imperfectly competitive. The positive level of protection for unorganized sectors is in contrast to the benchmark GH model and nds strong support in the data. As the theory predicts, among politically unorganized sectors protection increases with the share of domestic rms on the market. When protection is measured with tari¤s, the point estimate for 1 in the most preferred speci cation in terms of the log-likelihood function (column (3)) is 0:21, which converts to the welfare-maximizing ad-valorem import tari¤ of 2:2% for an average Canadian industry, given the average Canadian market share, price and substitution elasticities of 0:66, 5:32 and 5:83, respectively. In terms of the optimal level of NTBs and protection share, the welfare- maximizing NTB coverage for the average industry is estimated to be 8:9% of total imports from outside of the FTA, while the welfare-maximizing share of imports subject to any trade restriction is 17:9%. The e¤ect of a politically organized domestic lobbying (coe¢ cient 2) is always estimated to be positive and very signi cant, independently of the construction of the political organization dummy and the measure of trade distortion. Everything else being equal, active domestic lobbying in the industry leads to a higher level of protection and this e¤ect is signi cant and robust across all speci cations. The presence of a politically organized domestic lobby tends to increase import tari¤s by 5:4% for the average industry, the NTB coverage ratio by 29:7%, and the protection share by 31:7%. 23 The novel results of this section are the estimates of coe¢ cients 3 and 4. The coe¢ cient 3 measures the e¤ect of the FTA partner countrys lobbying and is always estimated to be positive, although in speci cations with tari¤s it is only marginally signi cant. The e¤ect of the ROW lobbying (coe¢ cient 4) is always negative and signi cant at 5% except for two speci cations with NTBs. According to these results, the presence of politically organized US industries in Canada leads to higher Canadian import barriers and the ROW lobbying e¤ort is negatively correlated with the Canadian protection measures. Thus, it seems safe to conclude that, while our results do not allow researchers to get a precise estimate of 3, they strongly support the hypothesis of foreign lobby di¤erentiation with respect to market access implied by the FTA and predicted by the theoretical model of Section 2.3. The point estimates of coe¢ cient 3 in speci cations with the best data t are 0:345 (column (3) for tari¤s), 3:033 (column (3) for NTBs) and 5:041 (columns (2) for protection share). These estimates imply that a politically organized industry in the FTA partner country tends to increase a countrys average import tari¤ by 1:2%, average NTB coverage by 10:5% and average protection coverage by 17:4%. Similarly, the presence of a politically organized group of exporters from outside of the FTA tends to decrease the average Canadian import tari¤ by 3:4%, the average NTB by 9:8% and the average protection share by 34:9%. These numbers are economically plausible and provide additional support to the estimates of the model. Estimates of equation (11) allow one to derive the values of the structural parameters of the model. As expected, absolute magnitudes of coe¢ cient 4 are always less than 2, which implies that ROW lobbying is relatively less important for Canadian policymakers than lobbying by do- mestic rms (c < 1). This result is not surprising given the legislative restrictions on contributions by foreign rms in Canada. In spite of that, policymakersvaluation of contributions from abroad is still positive and signi cantly di¤erent from zero. The estimates of the governments preferences to- wards political contributions from the FTA partner country (parameter b) depend on the protection measure that US rms are trying to a¤ect. In lobbying for tari¤s, partner country contributions are estimated to be 35 80% less e¤ective than domestic contributions, while in lobbying for NTBs the e¤ectiveness of domestic and partner country lobbying is approximately equal. Finally, when trade distortions are measured by protection share, the governments valuation of political contributions from the FTA partner country relative to domestic ones rises to 1:63, but the hypothesis that b 1 can never be rejected. At the same time, US contributions seem to be more important than contributions from the 24 ROW. Depending on the speci cation, the governments valuation of US contributions is estimated to be two times the valuation of contributions from other countries. Among possible reasons are greater proximity of the US and Canadian nancial systems and a more sophisticated mixture of asset structure that makes it more di¢ cult to distinguish between Canadian and US rms (relative to the distinction between Canadian and ROW rms). The estimates of the population share organized into lobbying vary substantially across di¤erent speci cations. In tari¤ equations, the parameter is always greater than one, but the standard error is very large and 95% con dence intervals almost always overlap with the [0; 1] interval. In equations with NTBs and protection shares, point estimates for are also very imprecise and in all speci cations one cannot rule out the possibility that is negative.21 As for the relative importance of national welfare for the government, the parameter a is esti- mated to be 0.3-0.5 with reasonable degree of precision as compared to the benchmark GH model. It implies that when organized interests lobby for a change in tari¤ policy, the governments valu- ation of political contributions is three to ve times higher than the valuation of national welfare net of contribution. The result that policymakers are driven mostly by political contributions they receive from di¤erent lobby groups sharply contrasts with results obtained previously in a perfectly competitive setup: Gawande and Bandyopadhyay (2000) estimated a to be over 3,000; in Goldberg and Maggi (1999) its value is around 70. My own estimates for Canada fall in the range 50-65 (see Section 2.6.1). However, Gawande and Bandyopadhyay (2000) recognized that high estimates of a contradict the empirical evidence that welfare loss from protection is always greater than the amount of political contributions policymakers receive in exchange for protection. Intuitively, a < 1 is what we should expect to get in the empirical model, because an increase in pro t from a 1% increase in tari¤ (i.e., the maximum contribution that the government could get for protection) is always lower than the corresponding decrease in welfare. Therefore, in order to observe any positive tari¤ above the welfare maximizing level in the equilibrium, the government should put a stronger emphasis on contributions relative to welfare and thus a < 1 is a desirable feature of the empirical model. Estimates of the reduced form (11) shown in Table 2.5 have a meaningful economic interpretation for an average Canadian industry. Given the average market shares of the three groups of rms, 21 It should be pointed out that measuring is a standard problem in the literature and its estimation proved to be di¢ cult. 25 price and substitution elasticities, one can back up the welfare maximizing level of protection and the average e¤ect of each lobby group on Canadian trade policy. Table 2.6 presents the average and marginal e¤ect of each lobby group on di¤erent measures of protection. Marginal e¤ects are calculated as the e¤ect of a particular lobbying group on an industry with average characteristics, and are discussed earlier in this section. Average e¤ects are calculated as a simple average of the tted values of the dependent variable implied by the model estimates. As a result, the average Canadian industry receives 1:33% import tari¤s, 7:31% NTB coverage and 7:81% protection share from lobbying by domestic rms; 0:21% tari¤s, 1:87% NTBs and 4:26% protection share from lobbying by partner country rms; 0:48% tari¤s, 1:38% NTBs and 6:24% protection share from lobbying by ROW rms. These results are economically reasonable and statistically signi cant. 2.6.3 Robustness tests Up to now, we tested structural relationship between trade policy outcomes and lobbying activity by domestic and foreign interest groups. However, the main message of the previous section still holds in a more general setup. Table 2.7 presents extended regressions that include other explanatory variables of trade policy predicted by earlier literature. We expect that more concentrated industries are more easily get organized and are more e¤ective in lobbying for protection; labor-intensive industries receive more protection from the government because in a capital-abundant country these are more likely to be import-competing industries; and industries with greater numbers of unskilled low-paid workers are more protected by the government for social justice reasons. Table 2.7 indicates that all additional variables have expected signs and mostly signi cant. But more importantly, all coe¢ cients of the central concerns, 2, 3 and 4, preserve their signs and signi cance. Table 2.8 reports results for speci cation (11) with additional control variables and without adjustment of the dependent variable by elasticity terms. Again, US lobbying activity in Canada is always found to lead to more protection, while ROW lobbying is associated with lower levels of protection. The remaining of this section explores the robustness of the previous results to the treatment of foreign lobbies operating in Canada. Table 2.9 represents a sensitivity analysis of the benchmark model results when one cannot di¤erentiate between domestic and foreign lobbying activity and attributes all or part of politically active foreign rms to domestic industry lobbying. Table 2.9 26 only shows the estimates of the equation in tari¤s. The rst three columns replicate the results for the benchmark GH model (with a Canadian lobby only) from Table 2.4 to provide a standard for comparison. The next three columns report speci cations in which US rms that are politically active in Canada are treated as Canadian. Interestingly, the results remain qualitatively the same as in the benchmark case, with only a small change in parameter estimates and a small increase in the likelihood function. In the last three columns of Table 2.7 all foreign rms that are politically active in Canada are treated as Canadian, and they show that the results are not robust to the treatment of the ROW lobbying. Coe¢ cients on X h i Mi Ii are biased towards zero and measured with a larger error, reecting the preferences of organized foreign lobbies for protection removal. This result is preserved when other protection measures are used and should not be surprising given conicting interests of domestic and ROW lobbying groups. Table 2.10 shows how robust the results of the monopolistically competitive model are to the exclusion of one or two foreign lobby factors. As in the previous case, I only consider the tari¤ equation here for two reasons. First, results are qualitatively the same when the NTB or protection share are used to measure trade distortions. Second, equations in tari¤s give higher precision and t the data better. The rst three columns replicate the results from Table 2.5 for comparison. The estimation results in Table 2.10 suggest that including the e¤ect of the US and ROW signi cantly improves the t of the model. However, to a large extent, this is due to the inclusion of the ROW lobby: the likelihood ratio test always rejects the model without ROW lobbying. At the same time, omitting the lobbying activity by US rms reduces the log-likelihood, but the likelihood ratio test fails to reject the model without US lobbying. Another important implication of Table 2.10 is that the estimates of 1 and 2 and thus the structural estimates of the model are not a¤ected seriously when foreign lobbying is excluded. The second set of robustness checks repeats the same thought experiment used for the bench- mark model, by assuming that one cannot distinguish domestic from foreign lobbying activity. In panels A and B of Table 2.11 the US rms that are politically active in Canada are treated as Canadian. Furthermore, in column A the lobbying activity by ROW rms is ignored. These two cases represent a test of the model when Canadian and US rms lobby jointly for protection. In panel C the researcher is assumed to overlook the e¤ect of an FTA on the structure of foreign lobbying and treats US rms lobbying in Canada as being from the ROW. Finally, panel D rep- resents a case when one does not consider the e¤ect of foreign ownership on the lobbying process 27 and treats all lobbyists that are active in Canada as representing Canadian interests. Results of Table 2.11 suggest that the inability to distinguish Canadian lobbying from US lobbying tends to overestimate the e¤ect of domestic lobbying and its average e¤ect on import tari¤, but does not make a big e¤ect on the models t, unless ROW lobbying is excluded. Panel C of Table 2.11 shows that the e¤ect of foreign lobbying on trade policy is very sensitive to the presence of a preferential trade agreement. When the e¤ect of NAFTA on the structure of foreign lobby is ignored, not only the coe¢ cient on the ROW political activity becomes insigni cant, but also it gets biased towards zero. This result is not particularly surprising given the fact that the lobbying objectives of ROW rms are opposite to those of US rms. Similarly, the estimates of Table 2.5 are also not robust to the treatment of the foreign lobby in general: in panel D when all lobbyists that are active in Canada are treated as domestic, the estimates of the coe¢ cient on domestic lobbying intensity, 2, are biased downwards and estimated with larger variance. Therefore, the concern raised at the beginning of this paper that the inability to separate the e¤ect of foreign lobbying from domestic may lead to misleading results even when a proper set of instruments is used nds strong support in the data. 2.6.4 Comparison of the benchmark and monopolistic competition GH models To test two competing versions of the GH model estimated by instrumental variables, I used a J-Test for non-nested hypothesis testing (Davidson and MacKinnon (1993), p. 388). The testing procedure requires the same dependent variable, so for the purpose of this section both sides of (11) were multiplied by ii 1 , and 1 i was moved to the right-hand side. The alternative benchmark speci cation (13) remains unchanged. As in Davidson and MacKinnon (1993), let f(sji ; I j i ; i) be a LIML regression model for the monopolistic competition speci cation (11), and let g(shi ; I h i ) be a LIML regression model for the benchmark speci cation (13). The validity of the benchmark GH model can be tested using an augmented LIML regression eg = g(shi ; Ihi ) + 1 bf , where bf is constructed from tted values of the LIML estimation of (11). The benchmark GH model is rejected in favor of the monopolistic competition model if the test for 1 = 0 is rejected. In a similar way, the monopolistic competition model is rejected in favor of the benchmark GH if in the regression model ef = f(si; Iji ; i) + 2bg the hypothesis 2 = 0 is rejected. 28 The J-test results for three measures of protection are presented in Table 2.12. They suggest that neither model can be rejected as a correct one when protection is measured with tari¤s. How- ever, the benchmark GH model is always weakly rejected in favor of the model with monopolistic competition as the monopolistically competitive speci cation of the model adds more explanatory power. The second part of each section of the table displays J-test results for the benchmark model and monopolistic competition with only domestic lobbying, using the same set of instruments. Again, neither model can be rejected, but these results suggest that simply taking into account the imperfectly competitive structure of the market improves the explanatory power of the Protection for salemodel. When protection is measured with NTBs, J-test results indicate that the benchmark model is rejected in favor of the monopolistic competition model with foreign lobbies and weakly rejected in favor of the model without foreign lobbies, while the monopolistic competition version is never rejected. This implies that the monopolistically competitive model developed in this paper captures more information than the model with perfect competition. 2.7 Conclusion The main objective of this paper is to study the e¤ect of foreign lobbies and FTA formation on a countrys trade policy. This paper accomplishes this goal by modifying the original GH model through the introduction of a monopolistically competitive market structure to allow for the presence of two types of foreign lobbies: lobbies from the FTA partner country and from the ROW. The paper shows that signing a free trade agreement with a big country leads to a welfare-reducing increase in import tari¤s in the presence of foreign lobbying, in particular lobbying by an FTA partner country. The empirical results suggest that the GH model with monopolistic competition has more explanatory power than its original version. For Canada, the e¤ect of foreign lobbying is statistically signi cant and is in line with theoretical predictions of the model: presence of an organized lobbying group in the FTA partner country tends to raise import tari¤s, while an organized lobbying group of exporters from other countries is associated with lower import tari¤s. A major bene t of the approach used in this paper is a more plausible set of values obtained for the government valuation of political contributions. Contrary to previous studies, the result that the government values 29 political contributions more than national welfare is very persistent and explains why relatively small political contributions have signi cant e¤ect on trade policy through lobbying. 30 2.8 Tables and Figures Figure 2.1: Distribution of the number of lobbyists across sectors. 31 Figure 2.2: Distribution of the number of lobbyists across sectors. Figure 2.3: Distribution of substitution elasticities. 32 Figure 2.4: Comparison of Canadian and US estimates for the elasticity of substitution. 33 Table 2.1: Descriptive statistics for protection measures and market shares, 1997. US tariff ROW tariff US NTB coverage ROW NTB coverage US protection share ROW protection share Canadian market share US market share ROW market share Mean 0.013 0.048 0.172 0.182 0.563 0.775 0.661 0.223 0.116 Median 0.003 0.034 0 0 0.646 0.986 0.683 0.205 0.075 Standard Deviation 0.076 0.053 0.314 0.307 0.419 0.328 0.222 0.162 0.134 Minimum 0 0 0 0 0 0 0 0 0 Maximum 1.17 0.54 1 1 1 1 1 0.70 0.64 Corr. with tariff 1 1 0.46 0.68 0.14 0.45 Corr. with NTB 0.46 0.68 1 1 0.45 0.35 Corr. with protection share 0.14 0.45 0.45 0.35 1 1 No. of observations 248 248 248 248 248 248 248 248 248 Table 2.2: Descriptive statistics for the number of lobbyists, 1996-97. Canada US ROW Average number of lobbyists per sector 1.94 1.08 1.15 Median number of lobbyists 1 0 0 Standard Deviation 3.30 2.06 2.32 Minimum 0 0 0 Maximum 18 12 14 Total number of lobbyists 481 267 284 % of sectors with at least one lobbyist 0.50 0.32 0.35 % of sectors with at least two lobbyists 0.38 0.26 0.26 % of sectors with at least three lobbyists 0.23 0.19 0.15 Table 2.3: Descriptive statistics for the elasticity of substitution and price elasticity Elasticity of substitution Elasticity of import demand Mean 5.83 5.32 Standard Deviation 3.06 2.97 Median 4.98 4.52 Minimum 1.43 0.17 Maximum 21.47 21.36 No. of observations 248 248 34 Table 2.4: Estimation results for the benchmark GH model with di¤erent protection measures Dependent variable: Tariffs NTBs Protection share (1) (2) (3) (1) (2) (3) (1) (2) (3) -0.0059** -0.0057** -0.006** -0.0349** -0.0348** -0.0354** -0.0163 -0.0164 -0.0177? B1 (0.0024) (0.0024) (0.0024) (0.0139) (0.0136) (0.0142) (0.0200) (0.0198) (0.0204) 0.0097 0.0111 0.0102 0.0271 0.0313 0.0249 0.0826 0.0864 0.0831? B2 (Canadian lobbying) (0.0081) (0.0087) (0.0088) (0.0448) (0.0486) (0.0484) (0.0709) (0.0784) (0.0768) Structural parameters: 0.61 0.51 0.59 1.29 1.11 1.42 0.20 0.19 0.21 a (1.279) (0.946) (1.221) (3.708) (2.897) (4.629) (0.928) (0.833) (0.928) 102.5 89.3 97.8 35.6 30.9 38.8 11.9 11.4 11.8a (168.7) (125.3) (161.9) (95.05) (73.88) (119.4) (25.12) (23.00) (25.88) N 248 248 248 248 248 248 248 248 248 Log-Likelihood -191.6 -192.0 -194.8 -637.8 -635.9 -640.7 -688.2 -693.1 -689.4 AIC 1.569 1.573 1.596 5.168 5.152 5.191 5.574 5.614 5.584 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In columns (1), (2) and (3) industry is considered as politically organized if it is represented by 1, 2 and 3 lobbyists, respectively. Table 2.5: Estimation results for the monopolistically competitive model (10) with foreign lobbying and di¤erent protection measures. Dependent variable: Tariffs NTBs Protection share (1) (2) (3) (1) (2) (3) (1) (2) (3) 0.072 0.179** 0.210** 0.828 0.939* 0.868* 0.915 1.757** 1.741**?B1B (0.093) (0.085) (0.084) (0.636) (0.570) (0.524) (0.848) (0.802) (0.792) 0.472*** 0.557*** 0.528*** 2.320*** 3.512*** 2.895*** 2.966*** 3.091*** 3.086***?B2B (Canadian lobbying) (0.059) (0.070) (0.069) (0.403) (0.468) (0.429) (0.537) (0.655) (0.648) 0.150 0.156 0.345* 3.047*** 2.152* 3.033*** 4.064*** 5.041*** 5.515***?B3B (US lobbying) (0.158) (0.167) (0.174) (1.079) (1.119) (1.082) (1.438) (1.568) (1.634) -0.109** -0.264*** -0.246*** 0.005 -0.914** -0.715 -1.910*** -2.542*** -2.491***?B4B (ROW lobbying) (0.051) (0.067) (0.073) (0.345) (0.447) (0.453) (0.460) (0.626) (0.684) Structural parameters: 1.97 1.47 1.50 0.07 0.02 0.05 0.03 -0.25 -0.24a (0.270) (0.228) (0.244) (0.313) (0.188) (0.208) (0.330) (0.317) (0.310) 0.15 0.32 0.40 0.36 0.27 0.30 0.31 0.57 0.56a (0.238) (0.191) (0.201) (0.351) (0.200) (0.222) (0.360) (0.350) (0.340) 0.32 0.28 0.65 1.31 0.61 1.05 1.37 1.63 1.79b (0.356) (0.314) (0.355) (0.591) (0.352) (0.428) (0.623) (0.691) (0.698) 0.23 0.47 0.47 0.00 0.26 0.25 0.64 0.82 0.81c (0.112) (0.131) (0.156) (0.153) (0.131) (0.164) (0.193) (0.261) (0.290) N 248 248 248 248 248 248 248 248 248 Log-Likelihood -74.2 -70.1 -68.9 -550.7 -542.1 -522.1 -622.0 -625.7 -624.3 AIC 0.64 0.61 0.60 4.48 4.41 4.25 5.06 5.09 5.08 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In columns (1) and (2) an industry is politically organized if it is represented by 1 and 3 lobbyists, respectively. In column (3) an industry is politically organized if it is represented by at least 3 lobbyists and accounts for strictly more than one third of the total number of lobbyists in that industry. 35 Table 2.6: Marginal and average e¤ects of political economy factors on di¤erent measures of pro- tection. Dependent variable (%): Tariffs NTB Protection share (1) (2) (3) (1) (2) (3) (1) (2) (3) Observed average 4.81 4.81 4.81 18.19 18.19 18.19 77.53 77.53 77.53 Active domestic lobbying 4.85 5.72 5.42 23.83 36.07 29.73 30.46 31.74 33.49 Active partner country lobbying 0.52 0.54 1.19 10.54 7.44 10.49 14.06 17.44 25.65 Active ROW lobbying -1.50 -3.62 -3.38 -0.68 -12.55 -9.81 -26.22 -34.90 -42.77 Average effect of domestic lobbying 2.65 1.41 1.33 13.04 8.88 7.31 16.67 7.81 7.80 Average effect of partner country lobbying 0.22 0.13 0.21 4.53 1.82 1.87 6.04 4.26 3.40 Average effect of ROW lobbying -0.06 -0.65 -0.48 0.00 -2.24 -1.38 -10.51 -6.24 -4.81 Table 2.7: Estimation results for the monopolistically competitive model (10) with additional controls. Dependent variable Tariffs NTB Protection share (1) (2) (3) (1) (2) (3) (1) (2) (3) 0.210** 0.307*** 0.210*** 1.503*** 1.555*** 1.247* 1.999*** 2.727*** 1.941***?B1 B (0.099) (0.089) (0.091) (0.686) (0.610) (0.658) (0.909) (0.857) (0.856) 0.412*** 0.506*** 0.393*** 2.024*** 3.302*** 2.165*** 2.506*** 2.630*** 2.498***?B2 B (Canadian lobbying) (0.058) (0.067) (0.056) (0.406) (0.460) (0.405) (0.538) (0.646) (0.527) 0.194 0.224 0.307** 3.254*** 2.636*** 3.432*** 4.401*** 5.165*** 5.727***?B3 B (US lobbying) (0.159) (0.163) (0.156) (1.104) (1.116) (1.127) (1.463) (1.568) (1.467) -0.100*** -0.260*** -0.202*** 0.044 -0.878** -0.458 -1.849*** -2.563*** -1.989***?B4 B (ROW lobbying) (0.051) (0.065) (0.055) (0.355) (0.449) (0.397) (0.470) (0.631) (0.517) 0.029*** 0.038*** 0.032*** 0.017*** 0.022*** 0.021*** 0.014* 0.020** 0.009*CR4 (0.011) (0.011) (0.010) (0.007) (0.007) (0.007) (0.008) (0.010) (0.005) -0.075* -0.067 -0.075* 0.013 0.198 -0.051 -0.018*** -0.017*** -0.018***K/L (0.045) (0.043) (0.042) (0.310) (0.296) (0.306) (0.004) (0.004) (0.004) -0.094*** -0.106*** -0.087*** -0.060*** -0.068*** -0.054*** -0.263*** -0.397*** -0.202***Wage (0.0061) (0.0059) (0.0058) (0.0042) (0.0040) (0.0042) (0.056) (0.057) (0.054) N 248 248 248 248 248 248 248 248 248 Log-Likelihood -61.4 -53.0 -48.9 -542.3 -530.3 -538.7 -612.1 -614.7 -603.9 AIC 0.56 0.49 0.46 4.44 4.34 4.41 5.00 5.02 4.93 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In columns (1) and (2) an industry is politically organized if it is represented by 1 and 3 lobbyists, respectively. In column (3) an industry is politically organized if it is represented by at least 3 lobbyists and accounts for strictly more than one third of a total number of lobbyists in that industry. 36 Table 2.8: Estimation results for the monopolistically competitive model (10) with additional controls and without elasticity adjustment of the dependent variable. Dependent variable Tariffs NTB Protection share (1) (2) (3) (1) (2) (3) (1) (2) (3) -0.009 0.006 -0.008 0.065 0.070 0.031 -0.355*** -0.239*** -0.357***?B1 B (0.014) (0.012) (0.013) (0.084) (0.075) (0.081) (0.081) (0.079) (0.076) 0.057*** 0.067*** 0.051*** 0.261*** 0.408*** 0.267*** 0.301*** 0.233*** 0.277***?B2 B (Canadian lobbying) (0.008) (0.009) (0.008) (0.050) (0.057) (0.050) (0.048) (0.060) (0.047) 0.002 0.003 0.006 0.336*** 0.243* 0.327*** 0.231* 0.280* 0.275**?B3 B (US lobbying) (0.022) (0.023) (0.022) (0.136) (0.137) (0.139) (0.131) (0.145) (0.130) -0.012* -0.039*** -0.033*** 0.036 -0.138*** -0.075 -0.208*** -0.292*** -0.252***?B4 B (ROW lobbying) (0.007) (0.009) (0.008) (0.044) (0.055) (0.049) (0.042) (0.058) (0.046) 0.031** 0.044*** 0.037*** 0.019*** 0.028*** 0.026*** -0.052 0.037 -0.079CR4 (0.015) (0.015) (0.014) (0.009) (0.009) (0.009) (0.089) (0.093) (0.086) -0.003 -0.007 0.000 0.002 0.004 0.001 -0.024*** -0.025*** -0.023***K/L (0.006) (0.0609) (0.059) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) -0.225*** -0.237*** -0.216*** -0.107*** -0.114*** -0.098*** -0.324*** -0.428*** -0.272***Wage (0.085) (0.082) (0.080) (0.005) (0.005) (0.005) (0.050) (0.052) (0.048) N 248 248 248 248 248 248 248 248 248 Log-Likelihood 428.7 436.8 441.7 -22.2 -10.4 -19.0 -13.3 -24.0 -2.5 AIC -3.39 -3.46 -3.50 0.24 0.15 0.22 0.17 0.26 0.08 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In columns (1) and (2) an industry is politically organized if it is represented by 1 and 3 lobbyists, respectively. In column (3) an industry is politically organized if it is represented by at least 3 lobbyists and accounts for strictly more than one third of a total number of lobbyists in that industry. Table 2.9: Estimation results for a benchmark GH model with import tari¤ and foreign lobbying being treated as domestic. Dependent variable: Tariffs Only Canadian lobbyists Canadian and US lobbyists Canadian, US and ROW lobbyists Threshold number of lobbyists: 1 2 3 1 3 5 1 3 5 -0.0059** -0.0057** -0.006** -0.0059** -0.0059** -0.0058** -0.0059** -0.0059** -0.0065***?B1 (0.0024) (0.0024) (0.0024) (0.0024) (0.0023) (0.0023) (0.0025) (0.0023) (0.0024) 0.0097 0.0111 0.0102 0.0099 0.0090 0.0106 0.0102 0.0075 0.0038?B2 (Canadian lobbying) (0.0081) (0.0087) (0.0088) (0.0082) (0.0106) (0.0120) (0.0083) (0.0104) (0.0103) Structural parameters: 0.61 0.51 0.59 0.59 0.66 0.55 0.58 0.79 1.68a (1.279) (0.946) (1.221) (1.283) (1.804) (1.495) (1.314) (2.630) (5.986) 102.5 89.3 97.8 100.8 110.8 94.0 97.0 131.7 258.4a (168.7) (125.3) (161.9) (169.0) (240.0) (197.4) (173.4) (358.1) (844.5) N 248 248 248 248 248 248 248 248 248 Log-Likelihood -191.6 -192.0 -194.8 -191.0 -191.5 -192.2 -194.0 -192.5 -214.9 AIC 1.569 1.573 1.596 1.565 1.568 1.574 1.589 1.577 1.758 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In the first six columns partner country lobbying is treated as domestic. In the last three columns all foreign lobby is treated as domestic. 37 Table 2.10: Estimation results for the monopolistically competitive model (10) with import tari¤s and di¤erent foreign lobbying structure. Dependent variable: Tariffs (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) 0.072 0.179** 0.210** 0.056 0.199** 0.195** 0.034 0.153* 0.162* 0.103 0.224** 0.248***?B1 B (0.093) (0.085) (0.084) (0.089) (0.085) (0.085) (0.089) (0.083) (0.083) (0.093) (0.087) (0.086) 0.472*** 0.557*** 0.528*** 0.480*** 0.552*** 0.566*** 0.488*** 0.573*** 0.547*** 0.460*** 0.537*** 0.543***?B2 B (Canadian lobbying) (0.059) (0.070) (0.069) (0.058) (0.071) (0.071) (0.058) (0.069) (0.069) (0.059) (0.072) (0.070) 0.150 0.156 0.345* 0.192 0.151 0.395**?B3 B (US lobbying) (0.158) (0.167) (0.174) (0.158) (0.172) (0.177) -0.109** -0.264*** - 0.246*** -0.115** -0.264*** -0.259***?B4 B (ROW lobbying) (0.051) (0.067) (0.073) (0.050) (0.067) (0.073) Structural parameters: 1.97 1.47 1.50 1.97 1.45 1.42 1.98 1.48 1.53 1.95 1.45 1.39a (0.270) (0.228) (0.244) (0.266) (0.234) (0.225) (0.261) (0.222) (0.239) (0.277) (0.241) (0.231) 0.15 0.32 0.40 0.12 0.36 0.34 0.07 0.27 0.30 0.22 0.42 0.46a (0.238) (0.191) (0.201) (0.215) (0.189) (0.183) (0.209) (0.174) (0.183) (0.249) (0.209) (0.202) 0.32 0.28 0.65 0.42 0.28 0.73b (0.356) (0.314) (0.355) (0.370) (0.335) (0.356) 0.23 0.47 0.47 0.24 0.46 0.47c (0.112) (0.131) (0.156) (0.108) (0.126) (0.151) N 248 248 248 248 248 248 248 248 248 248 248 248 Log-Likelihood -74.2 -70.1 -68.9 -77.2 -78.2 -76.9 -74.7 -70.6 -70.8 -76.5 -77.8 -74.5 AIC 0.64 0.61 0.60 0.65 0.65 0.64 0.63 0.60 0.60 0.65 0.66 0.63 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In columns (1) and (2) an industry is politically organized if it is represented by 1 and 3 lobbyists, respectively. In column (3) an industry is politically organized if it is represented by at least 3 lobbyists and accounts for strictly more than one third of the total number of lobbyists in that industry. Table 2.11: Estimation results for the monopolistically competitive model (10) with import tari¤s and foreign lobby being treated as domestic. Dependent variable: Tariffs A B C D Threshold number of lobbyists: 1 3 5 1 3 5 1 3 5 1 3 5 0.082 0.211* 0.218** 0.065 0.169** 0.1529* -0.039 0.125 0.124 0.129 0.249*** 0.276***?B1 B (0.088) (0.084) (0.082) (0.088) (0.082) (0.080) (0.106) (0.098) (0.091) (0.096) (0.090) (0.088) 0.497*** 0.542*** 0.629*** 0.500*** 0.555*** 0.652*** 0.495*** 0.573*** 0.585*** 0.357*** 0.351*** 0.449***?B2 B (Canadian lobbying) (0.058) (0.063) (0.070) (0.058) (0.062) (0.068) (0.058) (0.072) (0.071) (0.065) (0.065) (0.071) -0.102** -0.252*** -0.367*** -0.134 -0.119 -0.158?B4 B (ROW lobbying) (0.050) (0.065) (0.086) (0.099) (0.100) (0.104) Structural parameters: 1.85 1.46 1.24 1.87 1.5 1.3 2.1 1.53 1.5 2.44 2.14 1.61a (0.252) (0.227) (0.190) (0.251) (0.219) (0.182) (0.290) (0.241) (0.229) (0.463) (0.463) (0.327) 0.17 0.39 0.35 0.13 0.3 0.23 -0.08 0.22 0.21 0.36 0.71 0.61a (0.206) (0.187) (0.158) (0.202) (0.175) (0.146) (0.254) (0.215) (0.192) (0.333) (0.336) (0.249) 0.2 0.45 0.56 0.27 0.21 0.27c (0.105) (0.127) (0.141) (0.217) (0.185) (0.183) N 248 248 248 248 248 248 248 248 248 248 248 248 Log-Likelihood -75.3 -73.7 -70.5 -73.3 -66.5 -61.6 -76.4 -77.6 -75.9 -92.6 -91.9 -87.2 AIC 0.63 0.62 0.59 0.62 0.57 0.53 0.65 0.66 0.64 0.77 0.77 0.73 Notes: *, ** and *** denote statistical significance at 10%, 5% and 1%, respectively. Standard errors in parenthesis. In the first six columns partner country lobbying is treated as domestic. In the last three columns all foreign lobby is treated as domestic. In columns A and B partner country lobbying is treated as domestic. In column C US lobby is treated as representing ROW interests. In column D all foreign lobby is treated as domestic. 38 Table 2.12: J-Test for the benchmark and monopolistically competitive GH models. Dependent variable: Tariffs P-value H0: H1: 1 2 3 MC GH with foreign lobby is true Benchmark GH is true 0.50 0.55 0.59 Benchmark GH is true MC GH with foreign lobby is true 0.25 0.26 0.11 MC GH w/o foreign lobby is true Benchmark GH is true 0.42 0.45 0.43 Benchmark GH is true MC GH w/o foreign lobby is true 0.37 0.34 0.19 Dependent variable: NTBs P-value H0: H1: 1 2 3 MC GH with foreign lobby is true Benchmark GH is true 0.70 0.43 0.52 Benchmark GH is true MC GH with foreign lobby is true 0.22 0.07 0.04 MC GH w/o foreign lobby is true Benchmark GH is true 0.79 0.47 0.61 Benchmark GH is true MC GH w/o foreign lobby is true 0.49 0.15 0.52 Dependent variable: Protection share P-value H0: H1: 1 2 3 MC GH with foreign lobby is true Benchmark GH is true 0.01 0.15 0.00 Benchmark GH is true MC GH with foreign lobby is true 0.18 0.18 0.00 MC GH w/o foreign lobby is true Benchmark GH is true 0.01 0.21 0.00 Benchmark GH is true MC GH w/o foreign lobby is true 0.12 0.30 0.17 Notes: P-values denote the confidence level for rejecting H0. MC stays for “Monopolistic competition” version of GH model. In columns (1) and (2) an industry is politically organized if it is represented by 1 and 3 lobbyists, respectively. In column (3) an industry is politically organized if it is represented by at least 3 lobbyists and accounts for strictly more than one third of the total number of lobbyists in that industry. 39 3 Endogenous Free Trade Agreements and Foreign Lobbying 3.1 Introduction With proliferation of regional trade agreements, among the most important questions about free trade agreements (FTAs) in the recent literature are their e¤ects on excluded countries and the governments incentives to form regional trade agreements, as opposed to multilateral agreements. Formation of a regional trade agreement is a¤ected by both economic and political factors, and special interest groups continuously lobby for or against free trade in all countries of the proposed agreement. Since many of these agreements serve as a mean for smaller countries to secure preferen- tial access to large markets, smaller countries use di¤erent political measures to overcome resistance to the agreement in a prospective FTA partner country and, therefore, many trade agreements are formed in the presence of domestic and foreign political pressure. The main objective of this paper is to analyze which trade agreements are more likely to be signed in the presence of cross-country political pressure and what would be the optimum trade policy of these agreements. A large body of literature analyzes the FTA trade policy to determine its welfare e¤ects for the member and non-member countries. In this paper I develop a political economy model where an additional e¤ect of an FTA on the external import tari¤ comes from the presence of politically organized interest groups in a prospective partner country. I nd that depending on the lobbying structure within the FTA, the e¤ect of the agreement on the multilateral tari¤ system is ambiguous: in the presence of an organized interest group in a prospective partner country the external tari¤ may increase, and always decrease in its absence. Active lobby in a partner country can also push the local government to endorse a welfare-reducing FTA and block a welfare-improving multilateral trade liberalization. Thus, this paper identi es an additional potential adverse e¤ect of regionalism on the multilateral trade system. Most of the literature on the endogenous trade policy conclude that the welfare-maximizing government always nds it optimal to lower the external tari¤ under an FTA to minimize the tari¤ revenue loss.22 In the political economy context recent works by Ornelas (2005a,b) and Maggi and Rodriguez-Clare (2005) show that politically motivated governments tend to reduce the external tari¤ deeper since the FTA lowers incentives of speci c capital owners to lobby for protection. As 22See, for example, Richardson (1993), Bagwell and Staiger (1997), Bond, Riezman, and Syropoulos (2004). 40 for the welfare e¤ects of FTAs on the global trading systems, the literature on the endogenous trade policy emphasizes that the reduction of the post-FTA external tari¤, caused either by the lobby weakening motive or tari¤ revenue considerations, results in the overall trade creation and global welfare enhancement. On the other hand, the models with exogenously xed import tari¤s (Grossman and Helpman, 1995, and Krishna, 1998) show that countries are more likely to form trade diverting agreements because it allows rms from member countries to divert more trade from the rest of the worlds (ROW) rms. This paper shows that in the presence of foreign lobbying an FTA may either accelerate or slow down the multilateral trade liberalization of the member countries depending on the lobbying structure by country. While the volume of trade between FTA partners will increase as a result of tari¤ cuts, the external tari¤ (and trade volume with other countries) may either increase or decrease. In particular, the presence of an organized lobby in a partner country leads to increased political contributions for trade protection, and without counter-lobbying by the ROW rms the external tari¤ will increase once the trade agreement is signed. This result follows from the change in lobbying objectives of the organized groups in a prospective partner country: without the FTA, they lobby for the reduction of a common MFN tari¤ rate, while with preferential access to the home country market they switch to lobbying for more protection. Therefore, there is a potential for trade diversion e¤ect of the FTA with tari¤s being endogenously chosen by a politically biased government . While most of the previous literature asserts that FTAs are welfare-improving when tari¤s are endogenous, this work shows that foreign lobbying is a potential channel for trade agreement to be welfare-reducing. These ndings provide an additional support to the literature that emphasizes a potential harm of foreign interference into national politics. The results rationalize a recent trend in the policies of many countries to restricts foreign intervention in the political processes.23 Although the main reason for restricting foreign contributions is a political sovereignty concern, regulations of contributions by foreign corporations aim to limit the e¤ect of foreign agents on domestic policy 23During the last two decades many countries introduced legislation that restricts foreign intervention in the political processes. Among them are Canada (The Canada Elections Act, s.217, 1993), UK (The Political Parties, Elections and Referendums Act, 2000), Singapore (The Political Donations Act, 2001), Russian Federation (Federal Law No175, 2002), India (The Foreign Contribution (Management and Control) Bill, 2005), France (Code electoral, Article L52-8, 2005). Germany (The Law on Political Parties (Party Law), 2002) and Spain (Ley organica, de regimen electoral general, 2003) prohibit non-EU political contributions. In April 2003, the Committee of Ministers of the Council of Europe adopted Recommendation that states should speci cally limit, prohibit or otherwise regulate donations from foreign donors. 41 making, including the e¤ect on trade policy. Recently, several authors evaluated the e¤ect of foreign lobby on the US trade policy. Gawande, Krishna, and Robbins (2006) in their pioneering work demonstrated that foreign agents lobbying expenditure in the US is even greater than political contributions by US rms, and the e¤ect of foreign lobbies on the US trade policy is about as strong as the e¤ect of domestic organized interest groups. Using the same data, Kee, Olarreaga, and Silva (2003) showed that lobbying expenditures by Latin American countries is an important determinant of tari¤ preferences these countries get from the US. Gawande, Krishna, and Robbins (2006) argue that foreign lobbying may be bene ciary to the countrys trade policy as it helps to reduce distortional trade barriers. This paper suggests that the above argument is invalid under the presence of preferential trade agreement, when rms from a partner country lobby the government to maintain high discriminatory tari¤s for third countries imports. Since most of the countries are members of at least one preferential trade agreement, cross- border lobbying may stimulate more protectionist trade policies of FTAs and disrupt a multilateral trade liberalization. In the second part of this work I endogenize a governments decision to enter an FTA when the government takes into account the e¤ect of an agreement on the amount of political contributions from domestic and foreign rms. If foreign agents can nd the way to inuence the formation of the domestic trade policy, it is natural to assume that foreign interest groups can also lobby the local government for a given trade regime.24 The main result of this section is that depending on the lobbying structure, the government may either block a welfare-improving FTA, as in Ornelas (2005a), or choose to enter a welfare-reducing FTA. I identify two e¤ects of the FTA on the amount of political contributions from organized industries.25 First, the FTA changes the number of competing rms a¤ected by the import tari¤ and alters the ability of trade policy to shift market shares and pro ts toward organized groups, i.e. the FTA alters the amount of political rent created by the lobbying process. Second, under the FTA each organized group from one of the member countries cannot utilize the full bene t of its lobbying activity since part of the protection is 24There is a number of examples when foreign rms in partnership with domestic exporters formed a coalition that support implementation of bilateral trade agreement and launch lobbying campaign for its approval by policymakers. For example, US-Chile Free Trade Coalition, made up of approximately 400 US and Chilean companies and business organizations, initiated extensive campaigns to secure congressional approval of this trade agreement at Washington. Similar coalitions were established to lobby the US-Australia FTA (American-Australian Free Trade Agreement Coalition) Canada-Israel FTA (Canada-Israel Committee) and many others. 25Ornelas (2005a,b) call these e¤ects rent destruction. Here I changed the name because their e¤ect on contri- butions from foreign agents may be positive. 42 captured by other rms with preferential access to the market that are not necessarily participating in lobbying. In other words, the FTA changes the allocation of protection rents across groups that bene t from the policy change. I call these e¤ects pro t-shiftingand free-riding, respectively. Pro t-shifting and free-riding e¤ects of the FTA weaken the domestic lobbying power through reallocation of the market share and part of protection bene ts to partner country rms. Decreased contributions from domestic rms make the government better o¤ under the FTA only if welfare increase is high enough to compensate this loss. The size of the pro t-shifting and free-riding e¤ects of the FTA on contributions from the partner country rms depends on the market structure: when the share of the ROW rms on the home country market is relatively large, political tari¤has higher potential to reallocate pro ts toward rms with preferential access and partner country rms will contribute more under the FTA. On the other hand, they will contribute less under the FTA when the share of the ROW rms on the home country market is small for the opposite reason. When the FTA causes an increase in political contributions, the government may tolerate a small reduction in welfare and endorse an ine¢ cient FTA. With all three lobbying groups in place and assuming they are equally strong (equal valuation of foreign and domestic contributions by the government), political contributions under the FTA remain unchanged and the decision to form the agreement is based on welfare-maximization criteria. Thus, depending on the lobby structure by country, the government may choose not to enter a welfare-improving FTA or to enter a welfare-reducing FTA for political economy reasons. It is important to note that weakening of the domestic lobby by the FTA reduces the role of politics in the governments decision making and is a source of extra welfare gain from the FTA. At the same time, presence of the foreign lobby (both in the partner country and in the ROW) reinforces the politically motivated distortion and represents a threat for the national welfare and multilateral trade liberalization. This result is in contrast with Gawande, Krishna, and Robbins (2006) who argue that foreign lobby tend to reduce a distortional tari¤ and raise welfare. However, the result of Gawande, Krishna and Robbins turns to opposite in the presence of heterogeneous foreign lobby caused by the discriminatory FTA preferences. The rest of the paper is organized as follows. Section 3.2 presents a model of an FTA where the government chooses tari¤ rate endogenously and is subject to domestic and foreign political pressure. This section discusses the e¤ect of the FTA on the intensity of each lobbing group and analyzes trade policy of a county when it forms an FTA in the presence of lobbying by foreign 43 rms. Section 3.3 introduces a model of trade policy of the FTA with foreign lobbying when decision to form the agreement is endogenously determined by potilically-motivated government. This model is extended in Section 3.4 where domestic and foreign rms are allowed to lobby for or against the FTA directly. Section 3.5 presents quantitative version of the model to illustrate viability of welfare-reducing FTAs in the presence of lobbying by rms from the FTA trade partner. Concluding remarks are in Section 3.6. 3.2 The Model In this section the monopolisticly competitive market structure was incorporated into the GH protection for sale model. For simplicity it was assumed that there are only two industries: one numeraire industry with homogeneous output X0 and one monopolisticly competitive with a consumption index X. There are three countries in the model indexed by H; P and ROW . Country H (home) has an opportunity to form a free trade agreement with country P (prospective FTA partner country) and grant each other a preferential access to their markets relative to the ROW (rest of the world) exporters. Production of one unit of the homogeneous product requires one unit of labor, and since price of good 0 is normalized to one, the equilibrium wage rate must equal to one whenever the homogeneous good is produced in the equilibrium. In each country j 2 fH;P;ROWg there are nj symmetric rms that produce a di¤erentiated product. They use labor and country-speci c capital in a constant returns to scale production function, which implies constant marginal costs cj . Therefore, there are (nh + np + nf ) di¤erent varieties of the product available in each country j. The analysis is conducted from the perspective of a Home countryH. A representative consumer in the Home country maximizes quasilinear utility function: U = X0 + lnX X = nHxH 1 + nPxP 1 + nROWxROW 1 1 where is the elasticity of substitution between varieties of a di¤erentiated product, xj is a con- sumption of a product of a rm that operates in country j, and X is the consumption index for 44 di¤erentiated product. Maximizing the utility function subject to the standard budget constraint, the demand functions and an aggregate price index for the di¤erentiated product can be derived: X = P 1 xj = P 1 pj P = nH(pH)1 + nP (pP )1 + nROW (pROW )1 1 1 Firms in country j 2 fP;ROWg can export to the home country subject to the import tari¤. I denote a speci c import tari¤ set by the home country on imports from country j as j . Then a pro t function of a countrys j rm that sells on the domestic market is: j = (pj cj j)qj where qj is the quantity supplied and cj is the marginal costs. Faced with isoelastic demand structure, each rm sets an equilibrium price as a markup 1 over its marginal costs and applied import tari¤. In the equilibrium, the total return to the speci c factor owners in country j from selling on the home country market equals to j = 1njpjqj . The governments preferences are de ned as in Grossman and Helpman (1994). The government has the objective function G that increases in national welfare and in total political contributions received from organized industries. The economy welfare W aggregates consumerssurplus, pro- ducers surplus and tari¤ revenue. The government also receives political contributions C from the organized industries that attempt to inuence the choice of a trade policy. Therefore, the Home country government sets the trade policy to maximize its linear objective function that is a weighted sum of the national welfare and political contributions: G =W + aCH + abCP + acCROW The weight a shows how much one dollar of political contributions is worth to the government in terms of the national welfare and indicates the governments political bias. This speci cation of the objective function allows politicians to regard foreign contributions di¤erently from the domestic ones to reect a possible negative public attitude to foreign political interventions and the political risk involved with accepting contributions from foreign rms. It also allows di¤erent valuations 45 of contributions from the perspective partner country and other foreign rms since the initiation of an FTA may signal about closer political relations between countries H and P . Parameters b and c reect the governments preferences for contributions from the domestic rms relative to the partner countrys and ROW contributions, respectively. Presumably, these parameters satisfy the following condition: 0 < c b < 1. In this model government has a control over two trade policy instruments. Besides setting import tari¤s, it also chooses a trade regime, i.e. it can either form an FTA with a prospective partner country or not. In this and the next section the focus is made on the political processes within the Home country only. This implies that the Home country government receives an o¤er from a prospective partner country to form a trade agreement and would adopt the agreement only if it raises the governments objective function. In Section 3.4 I allow the domestic and partner country governments to set trade agreement cooperatively. Having described the main features of the model, we can proceed to the analysis of how the FTA a¤ects a countrys trade policy in the presence of organized foreign interest groups. Let yF and y0 denote the value of variable y with and without the FTA, respectively, and F y to be the change in variable y caused by the FTA. Maximization of a joint objective function of the government and organized interest groups yields an equilibrium tari¤ rate of an FTA member for the ROW imports:26 "F tF = 1 + ( 1)tPF sPF + 1 + aIH 1 + 1 sHF + abIP 1 + 1 sPF + acIROW 1 + 1 (sROWF 1) (14) where tF is an ad-valorem Most Favored Nation (MFN) tari¤ for the ROW imports, tPF is an ad- valorem tari¤ within an FTA, "F is the ROW import demand elasticity, sj = njpjqj XP is the share of country j producers in the Home countrys market, and Ij is an exogenously given dummy variable that takes the value of one when speci c capital owners in country j are politically organized and participate in the political game in the Home country. Since the FTA eliminates all trade barriers between member countries, the tari¤ rate for the partner countrys imports is restricted to be equal to zero under the FTA (tPF = 0) and the second term on the right-hand side of (14) disappears. Without the FTA, the external tari¤ would have a similar expression. Initially, in line with the 26A complete derivation of the equilibrium trade policy of the government in the presence of an FTA, foreign lobbying and monopolistically competitive market structure can be found in Appenxi A. 46 MFN clause of the GATT, the Home country imposes the same tari¤ rate on all imports irrespective of its origin. The condition tROW0 = t P 0 implies certain restriction on the structure of the political game in the absence of the FTA: without the agreement, both foreign lobbies (P and ROW ) share the same interests and lobby for the lower MFN import tari¤ set by the home country government. The MFN import tari¤ precludes competition among foreign lobbies for preferential access to the market and allows treating all countries other than H as a single rest of the world. Thus, in the equilibrium without trade agreement, lobbying e¤ort by country P rms has a negative e¤ect on the import tari¤ rate: "0t0 = 1 + 1 + aIH 1 + 1 sH0 + abIP 1 + 1 (sP0 kP0 ) + acIROW 1 + 1 (sROW0 kROW0 ) (15) where "0 = ( 1)(sP0 + sROW0 ) "F is the elasticity of the total import demand, kP0 is the share of a partner country in total imports of county H, and kROW0 is the share of the ROW in total imports (kP0 + k ROW 0 = 1). Propositions (1-4) show how the FTA a¤ects the trade policy of a member country in the current political game setup and di¤erent lobby structure. All proofs are presented in the Appendix C. Proposition 1 . Without political economy factors, the external tari¤ will always fall as a result of an FTA formation. The external tari¤ will always decrease by less then a decrease in the internal tari¤. Proposition 1 postulates the standard optimal tari¤ result of an FTA in a setting when welfare- maximizing government sets tari¤ endogenously: when two countries get preferential access to each others market, they both nd it optimal to lower the external tari¤ rate. This reects a complementarity e¤ect between tari¤s on imports from di¤erent countries in (14). High tari¤ rates on one source of imports raises imports from another country, making it optimal to tax it more and increase the tari¤ revenue collected. Therefore, tari¤ revenue considerations force the government to reduce tari¤ for one source of imports in response to lower tari¤ for competing import source. The second e¤ect of the FTA on the external tari¤ follows from a reduced share of domestic rms on the Home countrys market, which weakens the terms of trade motive for the government to maintain positive external tari¤. The result of the negative e¤ect of the FTA on the external tari¤ holds for the imperfectly competitive market structure, for any combination of country sizes, for any 47 elasticity of substitution greater than one, and for any relative marginal costs between domestic, partner countrys and ROW rms. Now, consider the e¤ect of the FTA on the external tari¤s when political economy considerations are taken into account. Proposition 2 shows that with di¤erent lobby structure the negative e¤ect of an FTA on the external tari¤ can be either enhanced or reversed. I identify two channels through which the FTA can alter the e¤ectiveness of the lobbying activity. The rst one, that I call pro t- shifting e¤ect, reects the change in the number of rms that the import tari¤ a¤ects. Once partner country rms get free access to the home country market, import tari¤s does not provide domestic producers a competitive advantage over this group of rms anymore, and the ability of import tari¤ to shift pro ts from foreign to domestics rms gets weaker. From the perspective of the ROW rms, the FTA allows them to use lobbying for the lower import tari¤ to gain a competitive advantage over the partner country rms in addition to domestic rms. The ability of the import tari¤ reduction to shift more pro t from competitors motivates ROW rms to lobby more intensively under the FTA. For the partner country rms the e¤ect is ambiguous and depends on the relative market shares of domestic and ROW rms. The second e¤ect, a free-riding e¤ect, relates to the ability of a lobbying group to capture the full bene t of protection or its removal. For the domestic rms, the FTA forces them to share protection bene ts with partner country rms. The latter free-ride on the lobbying activity of the domestic producers and enjoy the bene ts of higher external tari¤ without paying for it. For the ROW rms, the FTA allows them to enjoy the full bene t of lobbying for lower tari¤ without sharing bene ts of protection removal with competing exporters. As with the pro t-shifting e¤ect, the overall free-riding e¤ect on the partner country rms is mixed. Proposition 2 . The presence of an organized domestic and/or ROW lobby tends to reduce the FTA members external tari¤s. With an organized domestic lobby, the external tari¤ will fall by 1 a 1+ h sH0 "0 sHF"F i , and with an organized ROW lobbying the external tari¤ will fall by 1 ca 1+ h (1 sROWF ) "F kROW0 sROW0"0 i . This Proposition says that if there is an organized domestic or ROW lobbying (or both, but no partner countrys lobbying) the e¤ect of Proposition 1 is magni ed by free-riding and pro t-shifting e¤ects. 48 First, preferential access of a partner country rms to the local market weakens the lobbying power of the home industry through negative pro t-shifting e¤ect: under the FTA political contri- butions do not provide the competitive advantage over partner countrys rms any more. Before the FTA came into force, increased protection as a result of domestic lobbying a¤ected foreign rms that controlled (sP + sROW ) share of the total market, and shifted part of their pro ts to the domestic rms. Under the FTA the same amount of the political contribution and the following rise in the external tari¤ would shift part of the pro t of foreign rms that control only sROW share of the total market. Thus, preferential access reduces the range of rms a¤ected by import tari¤s and the e¤ectiveness of the domestic lobbying for protection. The free-riding factor has also the negative e¤ect on the domestic lobbying for higher tari¤s. Free access of partner country rms to the local market lowers the ability of domestic rms to capture the full bene t of protection. Intuitively, the FTA not only reduces the number of rms a¤ected by the external tari¤, but also weakens the ability of domestic rms to capture market share of the a¤ected rms by means of the import tari¤. With the agreement, higher tari¤ reduces the market share of the ROW rms and redistribute it between domestic and partner country rms, i.e. partner country rms free-ride on the rent-seeking lobbying expenditure of the domestic rms. For the ROW lobbyists pro t-shifting and free-riding e¤ects of an FTA tari¤ reductions increase their lobbying power against protection. The free-riding e¤ect is positive and reects the ability of the ROW lobbyists to utilize the whole bene t of their political activity. Prior to the FTA, part of the protection removal bene t followed from political contributions passed to the prospective partner countrys rms that faced the same MFN external tari¤. The trade agreement removes free- riding e¤ect of the partner country rms and motivates ROW industry to lobby more for protection removal. The positive pro t-shifting e¤ect follows from the fact that the lower import tari¤provides third country importers with a competitive advantage over the larger set of producers. Without the agreement import tari¤ reduction shifts a part of the domestic rmsmarket share towards ROW and partner country rms, while tari¤ reduction in the presence of the FTA allows ROW rms to capture part of the partner countrys rms market share as well. Hence, discrimination of the ROW rms under the FTA strengthens their lobbying power for less protection if they are politically organized. Proposition 2 says that the ROW lobbying power will increase once the FTA is signed, causing the FTA external tari¤ to fall. The next result shows how the trade policy responds to the change in a trade regime when rms in a partner country are organized into lobbying group. 49 Proposition 3 . An FTA has a positive e¤ect on a countrys external tari¤ in the presence of an organized foreign lobby in a prospective partner country. With organized lobbying by rms from a partner country, the external tari¤ will increase by 1 ba 1+ h sPF "F + kP0 sP0 "0 i . Proposition 3 implies an increase in the external tari¤ of an FTA member country if only partner country rms are organized into lobbying groups. This result is very intuitive: without the FTA, the partner countrys rms lobby against protection, hence, once the FTA is signed, lobbying for protection removal by the partner country rms will fall by k P 0 sP0 "0 . At the same time, under the FTA, they start lobbying for higher external tari¤s ( s P F "F ). Together, these two e¤ects imply that the government will be more motivated to raise the external tari¤ under the pressure of organized interest groups in the partner country. The result that the FTA external tari¤ may raise is in contrast with most of the previous regionalism literature and it points to a potentially welfare-reducing e¤ect of the FTA as a result of trade diversion. Proposition 3 is one of the key results of this paper. It demonstrates that foreign lobby can represent a threat for the national economy when FTAs are allowed and contrasts the results of Gawande, Krishna, and Robbins (2006), who claim that lobbying by foreign agents lead to the reduction of trade distortions and welfare-improvement. Proposition 3 states that their argument should be revised in the environment when almost every country in a world is a member of at least one regional trade agreement. Now it is possible to discuss the tari¤ implications of an FTA in the presence of a more general lobbying structure. Proposition 4 . If industries in all three countries are politically organized and the government values contribution from each country equally, the external tari¤ of an FTA member country will decrease. The tari¤ will fall by the same amount as in Proposition 1. When all three groups of producers are organized into lobbying, the overall e¤ect depends on the government valuation of political contributions from di¤erent countries. Proposition 4 claims that if producers in all three countries are politically organized and b = c = 1 holds, the political economy factors will cancel each other out and the change in the external tari¤ will be the same as in the case of the welfare-maximizing government. The situation where all three lobby groups are 50 politically organized corresponds to the special case of the benchmark GH model with all voters being represented by lobby. In this case the bid of one interest group is matched by counter bids of competing lobbies, so that the equilibrium policy outcome is identical to the outcome without any lobbying. However, each group still provides the government with a positive contribution to support this outcome rather than the one that favours its competitors. The main message of Propositions 2-4 is the dependence of F t on the lobby structure and governments attitude towards foreign contributions. With all three lobbies being active, political bias towards some group of rms can lead to an increase (b > 1 or c < 1) or decrease (b < 1 or c > 1) in the external tari¤ once the FTA is signed. In the special case when the government prefers contributions from the domestic agents and does not di¤erentiate between foreign contributions of di¤erent origin (b = c < 1), the overall result is dominated by the weakened domestic lobbying and import tari¤ will fall as a result of the FTA. However, in a more general case when 0 < c b < 1, the FTA e¤ect on tari¤ can be either positive or negative. The e¤ect of di¤erent lobby structure on the FTA external tari¤ is summarized in Table 3.1. It should be noted that in the presence of any two active lobbying groups, the FTA would imply a non-zero change in the tari¤ rate, unless a certain restrictions on parameters b and c holds. In the absence of either domestic or ROW lobby, the FTA leads to the higher external tari¤, while absence of the partner country lobbying causes a decline in the import tari¤ rate. Hence, this section identi es three important consequences of the FTA on the role played by lobbying for change in a trade policy. First, the FTA restricts the power of the domestic lobby, reducing politically driven distortions in tari¤ rates. Second, the FTA strengthens the non-member countries lobby, which may encourage politically biased government to set ine¢ ciently low import tari¤s. Finally, active lobbying by a partner country rms is reinforced by the FTA and the discrepancy between welfare-maximizing and political tari¤s may rise. 3.3 Endogenous FTA In the previous section it was assumed that organized interest groups lobby for tari¤s under a given trade regime and the FTA was exogenously given. I now turn to the analysis of the conditions when the government chooses to enter the agreement only if the government nds it worthwhile. I 51 will look at the e¤ect of a foreign lobbying on a governments objective function and the decision to form a trade agreement to analyze the e¤ect of endogenous FTAs on the external tari¤s. For now, I assume that the interest groups do not lobby for or against the FTA directly. Instead, in its decision to form the FTA, the government only considers how the FTA a¤ects the national welfare and the ow of political contributions. In the following section I will consider the case when lobbyists can inuence the governments decision regarding the adoption of the FTA. Assumption 1. Organized industries do not lobby for or against an FTA. Throughout this section I maintain this assumption that interest groups lobby only for the change in the trade policy but take an existing trade regime as given. Assumption 1 is a convenient starting point because it allows to consider the domestic market independently from other markets. Under the Assumption 1, rms lobbying for protection is always country-speci c as opposed to lobbying for a trade agreement directly. In the latter case, rms also need to allocate political contributions in support of or against the FTA tari¤ elimination between policymakers of two member countries. Yet, assuming away the direct lobbying for FTAs is quite reasonable as a rst step, especially if it is more costly for producers to overcome free-riding problems at the economy- wide level than at the industry level. Without trade agreement, the governments payo¤ equals to the sum of the domestic welfare (W0), that includes pro ts of domestic rms abroad, and political contributions: G0 =W0 + aI HCH0 + abI PCP0 + acI ROWCROW0 Under the trade agreement, the equilibrium payo¤ becomes: GF =WF + aI HCHF + abI PCPF + acI ROWCROWF As long as the FTA does not a¤ect the ROW trade policy, pro ts of domestic rms in third countries remain una¤ected, and the change in the governments payo¤ as a result of a trade agreement becomes: FG = FW + aI HFC H + abIPFC P + acIROWFC ROW = FW + aFC (16) 52 where FC is the total change in preference-weighted amount of political contributions: FC = I HFC H + bIPFC P + cIROWFC ROW At the same time, the FTA is welfare-improving when FW > 0. Thus, the e¤ect of the FTA tari¤ reduction on three groups of contributions completely determines the governments incentive to deviate from the rst-best outcome. In the remainder of this section I will investigate the relationship between the FTA and the amount of political contributions from the organized interest groups to analyze how the FTA a¤ects the governments incentives to change the trade regime. Following Grossman and Helpman (1994), I will continue to assume that the contribution sched- ules are truthful and locally di¤erentiable around the equilibrium. In this case contribution of country j lobby takes the following form: Cj = W j Bj , where W j is a gross welfare of group j, and Bj is a control scalar (net welfare of group j). For the Home country lobby W j includes pro ts, consumer surplus and redistributed tari¤ revenue of the population fraction involved into lobbying, while for foreign lobbies it is simply a pro t of a group j from importing because local tari¤s does not a¤ect consumer surplus and tari¤ revenue abroad. Let te denote the equilibrium external tari¤ and t j the value of the external tari¤ that would appear if politically organized group j were not participating in lobbying. Then, Grossman and Helpman (1994) showed that the equilibrium contributions by group j should compensate the government for the reduction in welfare and (preference-weighted) contributions of other lobbying groups from providing protection to group j: Cj = a X i6=j viIifW i(t j) W i(te)g+ fW (t j) W (te)g, where vH = 1; vP = b; vROW = c (17) In order to predict the change in political contributions that domestic government receives, we need to analyze the change in pro ts of three groups of rms that operate on the Home countrys market. Proposition 5 . An FTA has the following e¤ects on the absolute size of political contributions by each lobbying group: 53 A) If domestic rms are organized into a lobbying group, the amount of political contributions from the domestic lobby will decrease as a result of an FTA. B) If ROW rms are organized into a lobbying group, the amount of political contributions from the ROW lobby will increase as a result of an FTA. C) If rms in a prospective partner country are organized into a lobbying group, the e¤ect of an FTA on the amount of political contributions from the partner country lobby is ambiguous: there exists a threshold level for the ROW market share esROW0 (sH0 ) as a function of sH0 such that for sROW0 < esROW0 (sH0 ) contributions will decrease and for sROW0 > esROW0 (sH0 ) contributions will increase. In general, when country joins an FTA, pro ts of each group of rms may either increase or decrease depending on the lobbying structure. Proposition 5 says that, regardless of the lobbying structure, the FTA membership has an unambiguous e¤ect on the amount of political contribu- tions from domestic and ROW rms, while the change in contributions from a partner country is uncertain. The change in the amount of political contributions from domestic rms FCH caused by an FTA membership is always negative. As in the case with lobbying for the external import tari¤, there are two factors: pro t-shifting and free-riding e¤ects reduce the e¢ ciency of political con- tributions, weakening the incentive to lobby for protection. First, a smaller fraction of imports is subject to the discriminatory tari¤ under the FTA, hence, the agreements deteriorates the e¤ec- tiveness of the trade policy to shift pro ts from foreign to domestic rms. Second, the e¤ectiveness of lobbying will fall since with an FTA a part of a gain from tari¤ increase will pass to the partner country rms. For the ROW lobbyists there are two positive e¤ects of an FTA on the amount of political contribution FCROW . The free-riding e¤ect reects the stronger lobbying power under an FTA since without an agreement part of a gain from tari¤ reduction passes to the (prospective) partner countrys rms, while with the agreement ROW rms enjoy the whole bene t of their lobbying activity. The pro t-shifting e¤ect raises the e¢ ciency of the ROW contributions even further: without the FTA, the external tari¤ reduction reallocates part of the domestic rms market share toward the partner country and ROW rms, while under the agreement the ROW rms would also 54 be able to capture part of the partner country rmsmarket share with the lower external tari¤. Thus, one unit of political contributions by importers from third countries attracts more consumer expenditure under the FTA and the amount of political contribution FCROW will increase. As Proposition 5 says, the overall FTA e¤ect on the amount of political contributions by the partner country rms is ambiguous. Depending on the exact market structure, pro t-shifting and free-riding e¤ects may be either positive or negative for the partner country. First, with or without the FTA, partner country lobby su¤ers from a free-riding problem. When all foreign rms are faced with the same MFN tari¤ rate and only partner country rms lobby for protection removal, the ROW rms free ride on the lobbying e¤orts of a partner country industry and capture part of the protection removal bene ts. Under the FTA, partner country rms lobby for more protection and unorganized domestic sector free ride on the rent-seeking expenditure of the organized partner country lobby. In either case, the necessity to share bene ts of lobbying activity with either domestic or ROW competitors demotivates from lobbying, and the overall free-riding e¤ect of the FTA on contributions from the partner country depends on the relative size of domestic and ROW sectors. When the ROW sector is larger than the domestic industry, it pulls over a greater share of the lobbying bene ts and the FTA will loosen the free-riding problem. Similarly, when domestic sector is big relative to the ROW, the FTA will reinforce the free-riding problem and a¤ect negatively the amount of contributions from partner country. The pro t-shifting e¤ect of an FTA on partner countrys political contributions also depends on the relative size of domestic and ROW sectors, sH and sROW respectively. Without an FTA, the purpose of the lobbying activity by the prospective partner country rms is to capture a part of the domestic rmsmarket share and this part will be larger the greater is sH . With an FTA, the lobbying activity by the prospective partner country rms is targeted to capture a part of the ROW rmsmarket share and this part is increasing with sROW . Therefore, similarly to the free-riding e¤ect, the FTA provides more incentive for the partner country rms to make monetary contributions to the government when the ROW sector is big relative to the domestic one. Proposition 5 proves that for any sH0 and s ROW , there always exists a threshold level for the size of the ROW exporters esROW0 (sH0 ) such that for sROW < esROW0 (sH0 ) the amount of political contributions from a partner country would decrease, and for sROW > esROW0 (sH0 ) contributions would increase. To determine the e¤ect of the FTA on the total amount of political contributions in a more 55 general setting with a more diverse lobby structure, one needs to consider the interaction of three e¤ects identi ed in Proposition 5. The next proposition shows this impact in the case with three active lobby and allows to make inferences about FC for any combination of two lobbies. Proposition 6 . If rms in all three countries are politically organized and the government values contributions from each country equally, the total amount of political contributions remain una¤ected by an FTA. Proposition 6 indicates than when all three industries are politically organized, the FTA impact on the total amount of political contributions that government receives (FC) depends on the parameters b and c. A special case with three active lobbyists is when the government values contributions from di¤erent countries equally (b = c = 1): only in this case the amount of political contributions will not be a¤ected by a trade agreement. Intuitively, this result follows from the monopolisticly competitive structure of the model. Isoelastic demand curves imply that total pro ts of all rms that operate on the market does not depend on tari¤s,27 and Import tari¤s simply redistribute market shares from foreign to domestic rms. In this setup, bene ts of one lobby group from the FTA exactly match the loss of another. Accordingly, if all three industries are politically organized, a bid for change in a trade regime by one group is o¤set by a counter-bid of other lobbies,28 and political factors play no role in decision-making process. In deriving the above result, we have assumed that the government values all foreign contribu- tions equally. In a more general case with three lobbies, the e¤ect of the FTA on FC depends on the sign of FCP . If sROW < esROW0 (sH0 ) and FCP < 0, then a political bias towards some lobby group may lead to either negative (b > 1, c < 1), or positive (b < 1, c > 1) change in the total contributions. If sROW > esROW0 (sH0 ) and FCP > 0, then with b < 1, c < 1 the FTA would a¤ect FC negatively, while with b > 1 or c > 1 the e¤ect would be positive. Another issue related to FTAs with foreign lobby is a possible endogeneity of the structural parameters of the model. In this model, the option to form an FTA becomes available exogenously, while in practice FTAs are more likely to be observed between countries with close economic, political or cultural ties. Moreover, the trade agreement with a prospective partner country may 27Pro t of one rm equals to the revenue devided by the elasticity of substitution, so that the toal pro t of all rms summ up to the consumer expenditure devided by the elasticity of substitution. 28When all lobbyists are politically active, competition among them for protection is the most intense and the government captures all of the protection bene ts leaving interest groups indi¤erent between lobbying or not. 56 a¤ect the government perception of political contributions that come from the FTA partner rms. In either case, the domestic government may place a higher value on contributions from the partner country relative to the ROW in its objective function, which reinforces a distorting e¤ect of foreign lobby on the political process. Proposition 6 also allows to draw inferences about FC for any combination of two active lobbyists. Knowing that the change in contributions from three organized industries implied by the FTA sum up to zero, FC 6= 0 whenever at least one group of producers is not participating in the political game, except for the case with speci c restriction on parameters b and c. For example, with active home and ROW lobbying the total political contributions would increase for sROW < esROW0 (sH0 ) and decrease for sROW > esROW0 (sH0 ) when b = c = 1. Similarly, absence of a domestic lobbying would imply FC > 0 when other two lobbyists are active, and lack of lobbying from third countries would lead to FC < 0. Once the e¤ect of the FTA on the total amount of political contributions is identi ed, it becomes possible to analyze welfare implications of FTAs, endorsed by politically motivated governments in the presence of foreign lobbying. Condition (16) determines the change in the governments objective function as a result of the agreement. Whenever FG > 0, the government will choose to form the agreement, while the agreement is welfare e¢ cient if and only if FW > 0. It follows that whenever aFC = a IHFC H + bIPFC P + cIROWFC ROW is positive, the government may choose to enter a welfare-reducing trade agreement at the extent that FW > a IHFC H + bIPFC P + cIROWFC ROW (18) In other words, when the FTA leads to more political contributions, the government may tolerate a small reduction in welfare and endorse a socially ine¢ cient agreement. Similarly, if the government anticipates a reduction in the amount of political contributions as a result of the FTA, it would approve such agreement only if it is welfare-improving. In order for such FTA to be signed, an increase in welfare should be big enough to compensate for contribution reduction (the right-hand side of (18), which becomes positive). In this case another source of ine¢ ciency identi ed in Ornelas (2005a) may appear: the government may decide to block the welfare-improving trade agreement for political economy reasons if the FTA does not generate big enough welfare gain. These considerations establish the following proposition with records the main 57 result of this section. Proposition 7 . When FC > 0, the government may choose to enter a welfare-reducing trade agreement if FW 2 [ aFC; 0]. When FC < 0, the government will never enter a welfare- reducing trade agreement but may block a welfare-improving FTA if FW 2 [0; aFC]. Propositions 5 and 7, together with Proposition 6 and the discussion that follows, describe all possible welfare consequences of an FTA. If only the domestic industry is politically organized, aFC < 0 from Proposition 5 and the right-hand side of (18) is positive. On the other hand, Proposition 2 implies that with the domestic lobbying only, the external tari¤ would fall under the FTA towards the optimal level, which, in conjunction with distortion removal for the partner country imports, guarantees that the agreement improves national welfare and FW > 0. There- fore, condition (18) implies that with domestic lobbying alone, the government will never enter a welfare-reducing trade agreement but may block a welfare-improving FTA if FW < aFCH , i.e. if welfare gain generated by the FTA is not enough to cover the loss in campaign contributions. When only ROW exporters are organized into lobbying group, aFC > 0 from Proposition 5, and the external tari¤ would fall below the optimum level. This implies that FW can be either positive or negative, and in the latter case the second type of ine¢ ciency may appear: if 0 > FW > acFCROW , the government may neglect the negative e¤ect of a trade agreement on the national welfare in exchange for increased political contributions from the ROW lobby. Finally, when only partner country exporters form a lobbying group, both types of ine¢ ciency may appear. Proposition 3 indicates that in the presence of a partner country lobbying the external tari¤ would raise above the optimal level and aggregate welfare may decrease. It also follows from Proposition 5 that the change in contributions from the partner country is indeterminate. Hence, when the share of domestic rms on the home market is small relative to the ROW exporters and FC P > 0, it is possible that 0 > FW > abFCP and welfare-reducing trade agreement would be signed. On the other hand, when the share of the domestic rms is bigger than the share of the ROW exporters and FCP < 0, the government may refuse to form a welfare-improving trade agreement if 0 < abFCP < FW . In a more general case, the welfare e¤ect of an FTA depends on the governments valuation of contributions originated from di¤erent countries (b and c). The discussion of Proposition 6 58 describes the e¤ect of these parameters values on the total amount of contributions and possible type of ine¢ ciency that an FTA may induce. The result that the foreign lobby may cause a country to form a welfare-reducing trade agree- ment is the main one in this section and have not been identi ed before. Proposition 7 states that the foreign lobby represents a threat for the national welfare when a trade agreement is negotiated and justi es introduction of legal bans on foreign contributions from the perspective of trade policy issues. Moreover, what the next result shows is that the FTA, formed in the presence of lobby- ing by partner country rms, may also preclude the government from entering a welfare-e¢ cient multilateral trade agreement. Proposition 8 . When only domestic producers are politically organized, an FTA makes welfare- improving multilateral trade liberalization more likely. When only partner country exporters are politically organized, an FTA makes welfare-improving multilateral trade liberalization less likely. Proposition 8 says that in the presence of a home country lobby only, the FTA may stimulate the local government to adopt multilateral trade liberalization (MTL) and eliminate all tari¤s. Intuitively, this is so because the home country lobbying represents a distortion in the government decision-making process regarding trade arrangements, and weakening of the domestic lobbying due to the free-riding and pro t-shifting e¤ects reduces political economy considerations in the governments decision-making process. Since the government loses all political contributions as a result of MTL, transition to the global free trade from the FTA would cause a smaller reduction in the governments objective function and thus more likely. At the same time, the FTA may preclude from MTL that was otherwise feasible when a partner country industry is politically active. The reason is just opposite to the one described above: in presence of the partner country lobbying, an increased ow of contributions implies greater reduction in the governments utility function under the MTL, which implies that the minimum welfare increase from MTL required to compensate contribution loss will get larger under the FTA. Proposition 8 suggests an additional source of ine¢ ciency that may result from the foreign lobbying: if MTL becomes available in the presence of the FTA and organized foreign lobbying, the government may ine¢ ciently block a welfare-improving MTL that would have been adopted without the FTA. 59 To sum up, presence of politically organized industry in a partner country that can inuence the decision making processes in a home country may encourage a politically biased government to support a welfare-reducing trade agreement and block the e¢ cient multilateral trade liberalization. These results highlight the destructive e¤ect of foreign lobby on a countrys trade policy and rationalize legal restrictions on political donations from abroad. 3.4 Endogenous FTA With Lobbying for Change in a Trade Regime Thus far, the organized interest groups were not allowed to a¤ect the choice of a trade regime with political contributions. In this section, I turn to the analysis of situation when domestic and foreign lobby groups can support with monetary contributions those Home country politicians, who pronounce a commitment to the trade regime that suits interests of those groups. In the previous section interest groups were assumed to be unable to a¤ect directly the choice regarding the establishment of an FTA, and the o¤er to form the agreement was made exogenously by the prospective partner countrys government. In this section I relax these assumptions and allow interest groups to provide political contributions to either government in support of their most preferred trade regime. I show that a country can still enter a welfare-reducing trade agreement for political economy reasons in the presence of a foreign lobby, even if we allow for the direct lobby participation in a trade regime decisions. I do not consider the possibility of lobbying for or against the FTA by the ROW rms. The political structure of the model de ned similarly to the previous section. In the rst stage, each lobby group provides each government of the proposed FTA with contribution schedule contingent on the governments choice of a trade regime. Since this choice is binary, the contribution schedule consists of only two numbers. Let CijF and C ij 0 to be the contribution by group j to the government of country i if it choose to form the FTA and maintain status quo, respectively. As Grossman and Helpman (1995) show, it is optimal for a lobby to provide positive contribution only for the choice of a preferred outcome and zero if an alternative trade regime is chosen. Observing o¤ers of the organized lobby groups, governments of two countries choose to form the FTA or not in the second stage. In the third stage, lobby groups observe the current trade regime and provide each government with the contribution schedule as a function of the trade policy 60 outcome. Denote by Cijt the amount of political contribution of group j associated with the choice of the external tari¤ by the government of country i. Finally, governments choose the external tari¤s that maximize their objective functions: Gi =W i + ai Ci;Ht + b iCi;Pt + c iCi;ROWt + C i;H F + C i;H 0 + b iCi;PF + b iCi;P0 , i = H;P (19) The FTA can appear as an equilibrium outcome under several conditions. First, the FTA may raise welfare in both countries. Second, the FTA may reduce welfare in both countries. Finally, the FTA may be welfare-improving in one country and welfare-reducing in another. In the rst case, the FTA get established regardless of the political activity by the interest groups and I will not consider it in this work. I will also leave out the second case and concentrate on a third one. From now on, I assume that the FTA raises welfare in a prospective partner country P , while the e¤ect on welfare in the home country H can be either positive or negative. In this case, if the organized industry in country i expects to gain under the FTA, it will nd it optimal to lobby the government of country H in support of the agreement. Similarly, the organized interest group in country i that expects to lose under the FTA will try to block it by providing political contribution to the government of country H. When the FTA emerge as an equilibrium outcome, the total amount of contributions should make the government H indi¤erent between entering the agreement and maintaining a status quo: GHF = G H 0 . We also know that the lobby that will lose under the agreement o¤ers the whole amount of potential loss to the government H. This implies that CH;j0 = (Fj FCjt ) if (F j FCjt ) < 0, where j is the total pro t of group j earned on the whole FTA market and equals the sum of pro ts from sales in each member country: j = jH + j P . Finally, the winning lobby will never bid more than what it gains under the FTA: CH;j0 (Fj FCjt ) if (F j FCjt ) > 0. Substituting these two conditions into the equality GBF = G B 0 using (19), we obtain the following result. Proposition 9 In the presence of direct lobbying for change in a trade regime, the FTA is approved by the home country government i¤ the following condition is satis ed: FW H + aH IHF H + bHIPF P + cHIROWFC H;ROW t 0 (20) 61 Proposition 9 indicates that the FTA is approved by the home country government if it increases the sum of the national welfare and pro ts of the organized lobbyists, weighted by the governments political preferences. Condition (20) is similar to the one obtained by Ornelas (2005a) with two additional terms representing the e¤ect of foreign lobby. The condition also similar to the analysis of foreign lobby by Grossman and Helpman (1995), although in the current setup we account for the endogeneity of the FTA external tari¤s and foreign lobby heterogeneity. It follows from Proposition 9 that the necessary condition for viability of the FTA may be more or less stringent than condition (16) in Section 3, where direct lobbying for or against the FTA was not allowed. The di¤erence comes from the fact that the change in contributions for the external tari¤ is in general di¤erent from the change in pro ts implied by the FTA (FC H;j t 6= F j).29 Suppose, for example, that all three industries are organized, so that the change in contributions for the external tari¤ equals to the change in pro ts of group j on the country H market: FC H;j t = F j H . Then condition (20) is the same as (16) with an extra term on the left-hand side: aH IHF H P + b HIPF P P . Therefore, the main results of Proposition 7 for the case with three lobbyists are also valid for the case with the direct lobbying for change in a trade regime. When FCt > 0 and the weighted sum of the FTA member rmspro ts in country P is positive, the range of welfare-reducing trade agreements supported by the government gets larger. Similarly, when FCt < 0 and the weighted sum of FTA rms in country P is negative, the range of welfare-improving trade agreements blocked by the government gets larger. In a more general case, the following modi cation of Proposition 7 applies. Proposition 10 . When FC > 0, the government may choose to enter a welfare-reducing trade agreement if FWF 2 [ aFC; 0], where FC = IHF H + bHIPF P + cHIROWFC H;ROW t . When FC < 0, the government will never enter a welfare-reducing trade agreement but may block a welfare-improving FTA if FWF 2 [0; aC]. Proposition 10 reects the e¤ect of a foreign lobby on viability of the FTA. Allowing rms in a prospective partner country to lobby for a change in the trade regime raises the total contributions obtained by the government in support of the FTA if the partner country rms win from the 29From the proof of Proposition 5, FC H;j t = ajF j aj(FjA+FjB(tf j)). Therefore, FCH;jt >j Fj only when aj is big enough. 62 agreement. In this case, the foreign lobby may help to overcome the resistance to block the welfare- reducing trade agreement so that the FTA becomes more likely to emerge as an equilibrium outcome. On the other hand, if the partner country rms lose from the FTA, the government will endorse the agreement only if the welfare gain and competing lobby contributions are large enough to o¤set the opposition by the partner country to the pact. Hence, presence of the foreign lobby that opposes the FTA raises the chance that the e¢ cient FTA would be blocked by politically motivated government. Therefore, the presence of foreign lobby, either winning or losing under the FTA, introduces an additional distortion and raises the chance of ine¢ cient decision about the choice of the trade regime. As in Section 3, foreign contributions may also lead to an ine¢ cient decision regarding multi- lateral tari¤ reduction. Proposition 11 (i) Whenever domestic or partner country producers are politically organized and gain from the FTA, the agreement makes welfare-improving multilateral trade liberalization less likely. (ii) If both domestic or partner country producers are politically organized and the local government values contribution from each FTA country equally, the FTA makes welfare-improving multilateral trade liberalization less likely. Similarly to Proposition 8, the welfare-improving MTL is more likely to be blocked by the government in the presence of the FTA if political contributions rise as a result of the agreement. Therefore, whenever the FTA enhances the lobbying power of the organized industries, the loss of contributions as a result of MTL should be compensated by a larger welfare gain in order for the government to agree for trade distortions removal. As it is reected by Proposition 11, the lobbying may get stronger under the FTA for two reasons. First, either domestic or partner country rms that bene t from the FTA are politically organized. Second, if both domestic and partner country producers are organized and the governments valuation of foreign contributions is high enough, the FTA raises total contributions obtained by the government and makes it more reluctant to give up trade policy instruments. In either case, presence of foreign contributions may make a welfare-improving MTL politically infeasible that would not otherwise be so. 63 3.5 Quanti cation The central result of Section 3.4 is a viability of welfare-reducing trade agreements in the presence of foreign lobby: a politically biased government may endorse an FTA that lowers national welfare in exchange for political contributions from foreign agents. Proposition 10 stipulates that an ine¢ cient trade agreement will be signed if it bene ts organized interest groups and the change in contributions outweigh the welfare loss. The interesting question, however, is whether these conditions can be satis ed simultaneously in the current model setup and what factors make welfare-reducing trade agreement more or less likely. The numerical analysis of this section reveals that foreign lobby indeed may be responsible for establishment of ine¢ cient trade agreements and hence should be restricted from deleterious interference into trade policy negotiations. The purpose of this section is not to assess the likelihood of approval or disapproval of welfare-reducing trade agreement by the government under the pressure of foreign interests. This would require some prior knowledge about parameters that have no obvious values, such as the governments political bias and marginal costs of production. Instead, I was intended to show that there is a non-empty set of the model parameter values for which the politically motivated government driven by foreign contributions may choose to endorse welfare-reducing trade agreement. In the numerical analysis I look at the case of two symmetric economies considering formation of an FTA. The size of both countries is normalized to one (nH = nP = 1), however, the government in the prospective partner country assumed to have no political motivations (aP = 0). The parameter values were de ned as follows: fIH = 0; IP = 1; IROW = 0g reecting active lobbying by the partner country only, cH = cP = cROW = 1 reecting equal cost advantage across borders, H = P = 3 reecting 50% markup over the marginal costs, H = 0:3 reecting a 30% population share holding industry-speci c capital, and bH = cH = 1 reecting indi¤erence of policymakers to the origin of political contributions. I call this speci cation a benchmark and contrast later its predictions with those obtained from models with di¤erent parameter structure. I now outline the rationale for welfare-reducing trade agreements that may be signed under the pressure of foreign lobby. Figure 3.1 presents FW = 0 and FW + aFC = 0 loci in the plane aH nROW for the benchmark case. All else equal, the more biased the local government is towards political contributions, the higher is an increase in the external tari¤ from the FTA (Proposition 3) and the greater is a reduction in welfare caused by trade diversion. Similarly, 64 when the share of the FTA member rms on the home country market is small relative to the ROW, pro t-shifting and free-riding e¤ects cause the partner country industry to lobby more for protection (Proposition 5), leading to greater reduction in welfare. These two facts imply that the FW = 0 is downward sloping in aH nROW plane and the FTA raises welfare in the region below the curve. The shape of FP = 0 locus follows from Proposition 5: there is a threshold level for the ROW market share beyond which the FTA increases pro ts of the partner country rms and raises their willingness to contribute for preferential agreement. Therefore, partner country rmspro ts are positive to the right of FP = 0 when the FTA potential for trade diversion towards rms with preferential access is high, and negative to the left of FP = 0 for the opposite reason. The curve FW + aFC = 0 is a preference-weighted average of FW = 0 and FP = 0 and represents the viability condition for the FTA. Thus the FTA raises governments utility below FW + aFC = 0 when either a is not big enough to make the government too sensitive to the reduction in contributions, or the ROW market share is not too big to generate substantial trade diversion. Proposition 5 states that there always exists a threshold level for the ROW market shareesROW0 (sH0 ) such that for sROW > esROW0 (sH0 ) contributions (and pro ts) from partner country rms rise. Therefore,FW+aFC = 0 curve will always crossFW = 0 at nROW : sROW0 (n ROW ) = esROW0 (sH0 ) , and there always exist a range or parameters values such that FW < 0 and FG > 0 and the home country government would be willing to endorse a welfare-reducing trade agreement. The welfare-reducing FTA is viable if the governments political bias and the ROW market share are not too big to generate a net welfare gain from the FTA, and at the same time not too small to cause a substantial welfare reduction that would outweigh an increase in contributions. On the other hand, there is also a non-empty set of parameters values where the potentially welfare-improving FTA is blocked by speci c capital owners in a prospective partner country who are expected to lose under the agreement. This would be the case when two conditions are satis ed. First, the governments valuation of contributions is so big that the partner country rms can e¤ectively lobby for low MFN rate even without the agreement, and an FTA would imply an overall reduction in pro ts due to the loss in their domestic market share to rms with preferential access. Second, the governments valuation of political contributions is small enough in order to avoid welfare reducing trade diversion due to active lobbying by the partner country rms for an 65 increase in the external tari¤. It is important to note that to some extent the above results are driven by the ability of foreign lobby to raise the FTA external tari¤. With high political bias in governments preferences, foreign interests can successfully lobby for tari¤ increase and compensate large welfare loss from trade diversion with increased political contributions. However, this logic contradicts GATT Article XXIV(5), which precludes FTA members from raising their tari¤s above the level that was in force before the agreement was formed. Inability to raise the external tari¤ constitutes a considerable restriction on our analysis but its main implications continue to apply. As in Grossman and Helpman (1995) and Krishna (1998), a country can still enter a welfare-reducing and politically supported trade agreement if its FTA partner is small enough relative to the ROW, although the likelihood of such agreement decrease considerably relative to the case without tari¤ restrictions. The disadvantage of Grossman and Helpman and Krishna approaches is treatment of external tari¤ as being exogenously xed, while Ornelas (2005a,b) show that with endogenously determined tari¤s FTA would always reduce trade barriers and generate net welfare gain for all of its member and for the ROW. What this paper shows is that active foreign lobby may prevent tari¤s from being cut and generates the same result as in Grossman and Helpman and Krishna but with tari¤s being endogenously chosen by the government. It also shows that the GATT Article XXIV(5) restricts a potential adverse e¤ect of foreign lobby on the FTA trade policy and reduce the chance of welfare-reducing trade agreement being signed. Although the above analysis suggests that the scope for viability of welfare-reducing and in- viability of welfare improving FTAs is rather limited, part C of Proposition 5 shows that both types of ine¢ ciencies may appear with active foreign lobby. In other words, changes in underlying parameters would not change the qualitative predictions about political viability of an FTA and its welfare implications. The remaining part of this section analyses how the discrepancy between welfare impact and political viability of an FTA is a¤ected by global economy characteristics such as political bias in a partner country, relative size of FTA economies, the degree of market compet- itiveness, cross-border cost advantage, governments perception of foreign political contributions and lobby composition. I compared each counterfactual to the benchmark speci cation, holding xed other parameters of the model. First, we trace out the e¤ect of increased political bias in the prospective partner country on both types of ine¢ ciencies on Figure 3.2. In this case, the positive e¤ect of an FTA on home 66 countrys welfare is greater since a partner country was more protectionist and preferential access to its market becomes more valuable thus shifting FW = 0 to the right. The FP = 0 also shifts to the right because the higher is the political bias of the government, the weaker is the lobbying power of organized interests on their domestic market under the FTA (Proposition 2), i.e. FPP gets smaller when partner country government is politically biased. It follows that the FW + aFC = 0 intersects FW = 0 locus and the home country government can still nd it optimal either to block the welfare-e¢ cient FTA or to enter a welfare-reducing trade agreement under the pressure of foreign interests. The second counterfactual experiment is the e¤ect of tightening regulations regarding political funding from abroad. Prohibition on accepting contributions from foreign sources by domestic political parties leads to more risk associated with foreign contributions and biases the governments preferences toward domestic contributions (lowers bH). This situation is depicted in Figure 3.3. The e¤ect of the reduction in b on FW is straightforward: it weakens the governments motive for trade distortions and shifts FW = 0 locus up. As for the FP = 0, the e¤ect is twofold: smaller b lowers the political rent available both before and after the FTA becomes e¤ective. It follows that for the threshold level of the ROW market size, the net e¤ect is positive thus shifting F P = 0 locus to the left. In order to enter a welfare-reducing trade agreement, the governments political bias should be greater when b is small, however, the set of parameters values when country signs a welfare-reducing trade agreement is still non-empty. Figure 3.4 contrasts results for the benchmark speci cation to the ones with a more competitive market structure. When the elasticity of substitution gets bigger, both FW = 0 and FP = 0 loci shift to the right, and so doFW+aFC = 0 locus. The intuition comes from the fact that the equilibrium tari¤ rate depends inversely on the import demand elasticity and, therefore, on a more competitive market there are less possibilities for foreign interests to lobby for protection. This result implies that the government of a more competitive economy could enter a welfare-reducing FTA only if it has stronger preferences for political contributions. It is also true that the range of parameters values when the welfare-reducing trade agreement may be signer remains positive. The viability of the FTA with a more productive partner country, as reected by 20% higher costs of production at home, is represented by Figure 3.5. Cost advantage by the partner country leads to reallocation of market shares towards more productive rms, making foreign lobbying more e¤ective. As such, the FTA with technologically advanced and politically organized industry would 67 imply a greater upward pressure for the external tari¤, making the agreement more trade diverting and more attractive for partner country rms. Finally, we want to look at the combined e¤ect of domestic and foreign lobby on the governments decision regarding trade regime. If only domestic industry is politically organized, the FTA is always welfare-improving since the external tari¤ always falls as a result of the agreement and there is no room for trade diversion. However, the second type of ine¢ ciency is still possible: the FTA lowers contributions from domestic industry and when the governments preference for political contributions is high, this reduction outweighs an increase in welfare. Therefore, for the large values of a presence of domestic organized interest groups would prevent the government from entering a welfare-improving FTA. Figure 3.6 examines viability of the FTA in the presence of both domestic and partner country lobbying. This can also reect the case when trade agreement is formed under the pressure of binational corporations operating on both countries. Solid lines show zero isowelfare and govern- ments isoutility curves for the case when foreign and domestic political contributions are valued equally by policymakers. Cooperatively, domestic and partner country industries can successfully lobby for higher tari¤s under the FTA, which raises their aggregate pro ts at the expense of third countries producers. Therefore, FC = 0 locus coincides with nROW = 0 line and do not cross zero isowelfare curve, which has the similar shape as before. In this case, viability condition always lies above zero isowelfare curve and there exist a non-empty set of parameters values for which the government choose to sign a welfare-reducing trade agreement. However, since organized interests never loose from this FTA, they would never lobby against welfare-improving trade agreements. The equilibrium with restrictions on foreign political contributions is represented by dashed lines in Figure 3.6. Smaller bH implies that the government is less inuenced by foreign interests and is more reluctant to set higher import tari¤s. As a result, the zero isowelfare curve shifts up, while FC = 0 locus shifts to the right to reect strong negative e¤ect of pro t-shifting and free- riding. FPH and F P H are both be negative when the number of ROW rms is small because the external tari¤ would fall. Qualitatively, the outcome with domestic and foreign lobby where government is biased towards domestic contributions is similar to the one with foreign lobby only (Figure 3.1): there is a range of parameters value where government approve welfare-reducing and blocks welfare-improving trade agreement for political economy reasons. However, when bH falls too much the isowelfare curve will continue to move up and right and eventually will not cross 68 FC = 0 locus, so that the viability condition would always lie below zero isowelfare curve. In summary, this section reveals that for almost every values of parameters of the model there is a always a non-empty set of as such that the government choose to enter a welfare-reducing FTA, driven by political contributions from prospective partner country. This result implies that ine¢ cient preferential trade agreements are politically viable in the presence of foreign lobby, which is in contrast to the ndings of Ornelas (2005a,b) and Gawande, Krishna, and Robbins (2006). If foreign agents can nd the way to lobby local policymakers for their preferred trade regime and trade policy, the FTA is more likely to become trade diverting. However, such an agreement can still be supported by the government when welfare loss is compensated by foreign agents out of the protection rent. 3.6 Conclusion This paper analyzes a three-country world with two of them considering formation of an FTA. The paper examines political feasibility of the FTA and its welfare implications when cross-border lobbying is allowed by politically organized interest groups. If it is costly for organized interest groups to coordinate their e¤orts across industries for lobbying a particular trade regime, there is no direct lobbying for or against the FTA. In this case, while making the decision to form an FTA or not, politically minded governments take into account how the trade arrangement will a¤ect the stream of political contributions that lobbyists provide them with in exchange for protection. The main implication is that active lobbying by organized interest groups from a prospective partner country may encourage the local government to endorse a welfare-reducing trade agreement. With preferential access to the home country market, partner country exporters start lobbying for enhanced protection and the FTA external tari¤ may increase. However, the domestic government with strong preferences towards political contributions may neglect the welfare-reducing trade diversion e¤ect when it is compensated by increased contributions from foreign agents and enter an ine¢ cient trade arrangement for political economy reasons. This result is preserved when interest groups can directly lobby towards the choice of their most preferred trade regime: an organized foreign lobby provides the home country government 69 an incentive to engage in trade diverting preferential arrangements that generate more rents from the external tari¤. Increased foreign contributions also obscure the movement to the global free trade since it gets harder for the government to abandon trade policy as a rent-generating device. The fact that foreign lobby may be responsible for more protectionist trade policy of the FTA implies that GATT Article XXIV(5) is an e¢ cient tool in restricting a potentially adverse e¤ect of foreign lobby on the FTA trade policy that reduces the chance of ine¢ cient trade agreement being signed. However, with a direct lobbying for a change in a trade regime the foreign lobby may be responsible for a second type of ine¢ ciency regarding formation of the FTA. When the FTA generates substantial welfare gain in both countries due to the reduction in the external tari¤ but hurts partner country exporters, they may e¤ectively block such an FTA if the local government has a strong political bias. Simulations of counterfactuals illustrate a potentially adverse impact of foreign lobby on the governments choice of the e¢ cient trade regime. 70 3.7 Tables and Figures Table 3.1: The e¤ect of an FTA with di¤erent lobby structure on the external tari¤. Figure 3.1: Viability of welfare-reducing FTA: a benchmark speci cation 71 Figure 3.2: Viability of a welfare-reducing FTA with a politically biased government in a partner country (aP = 1). Figure 3.3: Viability of a welfare-reducing FTA with a governments preference for domestic contributions (bH = 0:5). 72 Figure 3.4: Viability of a welfare-reducing FTA with a more competitive market structure ( = 5). Figure 3.5: Viability of a welfare-reducing FTA under cost disadvantage (cH = 1:2). 73 Figure 3.6: Viability of a welfare-reducing FTA with domestic and partner country lobbying. 74 4 A Model of Trade Liberalization and Technology Adoption with Heterogeneous Firms 4.1 Introduction Recent models of rm heterogeneity and trade have been successful in explaining a large set of di¤erences, documented in the empirical literature, between rms that engage in exporting and rms producing solely for the domestic market. In particular, these models explain the observed di¤erences in productivity levels (Melitz, 2003, Bernard et al, 2003), size (Melitz, 2003, Luttmer, 2007), skill intensity and wage premium (Yeaple, 2005) and innovation activities (Atkeson and Burstein, 2007), and the importance of these di¤erences for the e¤ect of trade liberalization on aggregate productivity. Additionally, many rm-level empirical studies report higher capital in- tensity of exporters in countries with di¤erent income levels and factor endowments.30 Although consistently found in the data, such pattern cannot be rationalized by any of the recent theoretical models of rmsself-selection into the export market. In this paper I investigate the source of capital-labor di¤erences across rms and construct a model to address this issue. First, using French rm-level data, the paper shows that the reason for higher capital intensity of exporters is not only a di¤erence in factor prices, but also a production technology of exporters, which is more intensive in physical capital. With Cobb-Douglas production technology and two input factors, the elasticity of output with respect to capital is 10 25% higher for exporting rms. Second, the paper extends the Melitz (2003) model to capture heterogeneity in capital intensities across rms to analyze the implications of this heterogeneity for trade policy outcomes and aggregate economic variables. To replicate the di¤erences in factorsusage observed in the data, the model allows for the endogenous choice of production technologies by rms with di¤erent productivity levels. In an environment where higher productivity is associated with lower capital price, as the data suggests, more productive rms are willing to use more capital-intensive technologies, and since these are more expensive to use, only the most productive rms will get enough revenue to cover the technology adoption costs. In this model, trade liberalization has two 30The evidence can be found in Clerides, Lach, and Tybout (1998) for Colombian, Moroccan and Mexican rms, Bernard and Jensen (1999) and Bernard, Eaton, Jensen, and Kortum (2003) for US rms, Baldwin and Gu (2003) for Canadian rms, Van Biesebroeck (2005) for African rms, Alvarez and Lopez (2005) for Chilean rms, and many others. 75 e¤ects: reallocation of market shares as a result of rmsentry an exit, identi ed in Melitz (2003), and output expansion by the most productive rms due to the adoption of more capital-intensive technologies. The model simulations suggest that up to 10% of trade liberalization welfare gains come from the adoption of new technologies by existing rms. The theoretical model builds on Melitz (2003) general equilibrium model of rms self-selection into exports market. To start a new rm, the owner has to make a sunk investment before the rms productivity is revealed. The randomly drawn total factor productivity index is constant over the life cycle of the rm and determines the cross- rm di¤erences in marginal costs, given the same amount of input factors. Once the productivity is observed, the rm decides whether to exit or stay in the market, and whether to enter the foreign market or not. Following Melitz (2003), Bernard, Eaton, Jensen, and Kortum (2003), Alessandria and Choi (2007) and others, both producing and exporting require a xed amount of nal output in every period, which plays a key role in the rms entry and exporting decisions. The model is extended by allowing for two factors of production: labor, which is inelastically supplied by consumers, and capital, which can be produced from labor with constant returns to scale technology. Capital is owned by consumers, who rent it to rms at the currently prevailing market price. The literature on trade and rm heterogeneity has traditionally assumed that all rms within an industry use the same production technology but may have di¤erent productivity levels.31 This paper starts by providing evidence that rms with di¤erent productivity levels and export status operate under di¤erent production technologies. Using French rm-level data, I estimate the elasticity of output with respect to physical capital to be signi cantly higher for exporting rms. Controlling for di¤erences in size, factorsprices, and total factor productivity (TFP), the capital intensity of exporters is still 10 25% higher than for non-exporters. It is also the case that capital intensity is increasing with export intensity. The fact that the more productive exporting rms operate with di¤erent production technologies suggests that being an exporter gives an advantage in using more capital-intensive production processes. One possible explanation for this result is the well known empirical regularity that small 31Models by Melitz (2003), Helpman, Melitz, and Yeaple (2004) Ghironi and Melitz (2005), Luttmer (2007) Atkeson and Burstein (2007) among others based on linear production technology with one factor of production and rms are heterogeneous only with respect to total productivity parametor. In studies with multiple input factors by Bernard, Eaton, Jensen, and Kortum (2003), Rossi-Hansberg and Wright (2005), and Alessandria and Choi (2007) rms are assumed to share the common production technology with rm-spci c total factor productivity. 76 (and less productive) rms are more risky and, hence, are faced with higher capital prices.32 The positive relationship between rms survival probability and productivity nds strong support in the French data, and the result that the interest paid on loans is decreasing with productivity is very robust as well. As long as the most productive rms are more likely to survive, they will also be more likely to repay the loan. Therefore, they will be charged lower capital price since it will include smaller risk premium. In this context, if there are di¤erent production technologies available to rms, more productive exporting rms will set up technologies that use cheaper factor (capital) more intensively, provided that the costs of this technology does not exceed the total productivity gain. Since the data shows that di¤erent rms use di¤erent production technologies, the model was further extended by assuming that there are di¤erent types of capital, and more advanced machines require fewer number of workers to produce one unit of output. Upon entry, every rm uses the basic and the least capital-intensive technology, but once its productivity is revealed, the capital structure can be adjusted. In following periods, rms can choose from the menu of available technologies the one that maximizes the expected stream of pro ts, given the increasing adjustment costs of using more capital-intensive processes. These adjustment costs of adopting more capital-intensive technology result from the necessity to rearrange rms capital stock by simultaneous sales of old capital, that requires more worker to produce a unit of output, and purchase of new capital. Apart from direct capital costs, this implies additional adjustment costs of labor retraining, installation of new and dismantling of old capital, and other. Additionally, in the process of capital replacement the old capital may lose part of its value. Therefore, faced with di¤erent capital prices and increasing technology adjustment costs, rms have incentives to use di¤erent technologies: only the most productive rms will nd it worthwhile to install a more capital-intensive technology, which gives them an additional cost advantage over other rms. Di¤erent capital intensities then follow from rm-speci c productivity and probability of survival. This way of modelling technology adoption is novel for the literature that links trade and technology. In previous studies new technology is associated either with the development of a new product (Grossman and Helpman,1989, Coe and Helpman, 1995) or with the increase in rm- level TFP (Eaton and Kortum, 2001, Luttmer, 2007). In both cases, production technologies are 32Roll (1981) and Titman and Wessels (1988) argue that small rms are riskier and pay more for debt and equity issue. Easley and Ohara (2004) assert that small rms have to provide higher return on investments since they can reveal less information to the public. 77 identical across rms, except for TFP di¤erences. In this paper, rms also di¤er in a way they organize production process and employ a mix of input factors. Here, I abstract from endogenous investments in TFP improvement or new product development in order to focus sharply on the e¤ect of production process heterogeneity on aggregate productivity and trade. The general equilibrium trade model with rm heterogeneity and endogenous technology choice implies that higher productivity encourage adoption of an even more productive capital-intensive technology, so that the exposure to trade induces only the more productive and the more capital- intensive rms to enter the exports market. In this environment, trade liberalization not only raises the threshold level for productivity required to enter the market, but also lowers the productivity threshold that makes is pro table to install a more e¢ cient technology. In equilibrium, the reduction of trade barriers has two e¤ects. First, as in Melitz (2003), free trade causes within-industry reallocation of resources, forcing the least productive rms to exit and more productive rms to enter the exports market. Thus, trade costs reduction increases the aggregate productivity by changing the composition of rms within the industry, but does not a¤ect the rm-level productivity. The second e¤ect of tari¤ reduction, not identi ed in previous literature, is that exporters switch to using more productive and more capital-intensive technologies since they can spread the technology adoption costs over larger quantities of output. In equilibrium, the reduction in production costs by rms that install more advanced capital-intensive technologies results in reallocation of production shares towards exporters, which ampli es the initial reallocation e¤ects and leads to further increase in the aggregate industry productivity. By disregarding the second e¤ect, previous works shut down an important source of trade liberalization bene ts and attributed within-industry technological changes to changes in TFP in empirical models. The results of simulation exercise indicate that around 10% of the e¤ect of trade liberalization on productivity are due to the change in composition of technologies across rms rather than the composition of rms within the industry. Predictions of this model are related to the ndings by Atkeson and Burstein (2007). They constructed a dynamic open economy model where monopolistically competitive rms endogenously make investments decisions to decrease marginal costs of production or to establish a new rm. They show that trade openness allows exporting rms to spread innovation costs over larger output levels and stimulates investing more in innovation activities. Similar to this paper, trade raises incentives of exporting rms to use more productive technology and, hence, raises aggregate productivity. 78 Although in this paper rms do not investment in TFP improvements, trade stimulates rms to use more productive technologies, which are more intensive in the relatively cheaper factor of production. As a result, the model explains why more productive exporting plants operate with more capital-intensive technologies, and reveals an additional channel for the welfare gains from trade liberalization, arising from the change in the composition of technologies across rms. The paper is organized as follows. The next section presents an empirical evidence of within- industry technological di¤erences using French rm-level data. Section 4.3 presents a theoretical model of endogenous technology choice with heterogeneous rms. Section 4.4 describes the models calibration strategy and implications of trade barriers reduction for the aggregate economic variables and Section 4.5 concludes. 4.2 Empirical Model This section examines empirically the di¤erences in production technologies used by French export- ing and non-exporting rms by examining the di¤erences in capital productivity between them. It provides the evidence that rms selling on foreign markets operate under di¤erent production tech- nologies with larger share of capital in the nal output. In the following sections I present a model that controls for production function di¤erentials across rms and provide its quantitative version to analyze the implications of these di¤erences for the trade policy e¤ect on the economy. 4.2.1 Modeling Strategy Assume that the production function of rm i at time period t has a Cobb-Douglas form: yit = ait + K it kit + L itlit + M it mit (21) where y is log sales, a is log rm-level productivity, k is log stock of xed capital input, l is log amount of labor input, and m denotes log intermediate material input. The main hypothesis about the production function is that capital productivity of exporters is systematically higher than for non-exporters. To allow for di¤erentiated capital-intensity with respect to the export status, assume that Kit = K;0 it + K;E it I E it , where I E it is an indicator function 79 for the export status. Exporters are also allowed to have di¤erent level of productivity relative to non-exporters: ait = ai + a EIEit + 0zit + ht + uit (22) where ai is an unobserved rm-speci c productivity parameter, zit are observable variables a¤ecting productivity such as age, geographic location and legal form, and uit is a rm-speci c productivity shock. Sub-script h denotes industry, and ht stays for the industry-wide time-speci c productiv- ity shock. It controls for aggregate technology shocks, demand shocks, factor prices shocks, and any other industry-speci c shocks. Therefore, the production function to be estimated takes the following form: yit = ai + a EIEit + K;0 it kit + K;E it I E it kit + L itlit + M it mit + 0zit + ht + uit (23) and we examine whether exporting is associated with the use of a more capital-intensive technology by testing whether K;Eit = 0. A positive value of K;E it supports the hypothesis that exporting plants operate under di¤erent more capital-oriented technology. In a more general case exporters are also allowed to have di¤erent intensities with respect to labor and materials. 4.2.2 Data The data for this study is taken from the Amadeus database available from Bureau van Dijk. The data covers the panel of 103; 652 French manufacturing establishments with a rm being a unit of observation. The unbalanced panel includes annual observations for the period 1997 2005. During that period, the number of rms in the sample have increased from 59; 067 to 82; 449, and the share of rms that export have decreased from 36:4% to 31:9%. For the analyzed period 8% of all rms continuously operated on foreign markets, 50% never participated in export operations and 42% switch between exporting and not exporting. The balance sheet book value of domestic sales, xed capital, material costs and foreign sales, measured in current Euro, are used to measure rms output, capital stock, intermediate inputs, and export status, respectively.33 This paper uses the balance sheet record of employment, represented 33The current year capital stock does not include current investments that are assumed to become productive in 80 by the number of workers, as a measure of labor input. Around 40% of all rms do not report employment level, and since misreporting is not systematically related to rms sales, export status, and other characteristics, the missing data was replaced by the rms total labor expenditures divided by the industry-average wage. Firms with other missing information were removed from the sample, which reduced the number of rms in the analysis to 98; 671. In addition to employment and nancial data, each rm is classi ed by geographic region and 4-digit US NAICS code. The data on the date of establishment and legal status is also available. Table 4.1 provides summary statistics for the data. There are many small and medium rms in the data with the median employment level of only seven workers, which means that the sample is not biased towards large rms. Table 4.1 con rms several stylized facts about exporting rms: on average, they are twice older, produce three times more output, employ four times more labor, and have higher labor productivity measured by value added per worker. Table 4.1 also con rms the considerable di¤erence between exporters and non-exporters with respect to capital-labor ratio. The average exporting rm uses 30% more xed capital per worker, and this magnitude is similar to the ones documented by earlier studies. Not surprisingly, exporting rms are more investment intensive as well, investing 35% more in physical capital per worker that non-exporting rms. The amount of debt per worker and the interest rate are also vary across rms with di¤erent export status. The average exporting rm pays 0:5% less interest on capital and borrows 15% more per one worker. Additionally, exporting rms are more likely to survive: the annual survival rate among exporters is 94:5%, while for non-exporters it is 93:8%. 4.2.3 Estimation Approach The relationship between variable input factors on the one hand and rm-speci c productivity advantage (ai) and contemporaneous productivity shock (uit) on the other raises the issue of en- dogeneity problem in estimation of equation (23). OLS estimates tend to provide upward biased coe¢ cients on factor inputs that are positively related to rm-speci c productivity. Similarly, co- e¢ cient K;E estimated by OLS is also biased upwards since more productive rms are more likely to use more capital-intensive technology. If technology adoption does not depend on temporary productivity shocks and current choice the next year. 81 of input factors, then the speci cation with a full set of rm-level xed e¤ects will provide a consistent estimate of K;E . However, as long as adoption of a more capital-intensive technology is associated with the rm-speci c time-invariant productivity, as the theoretical model predicts, then the within-estimator will absorb this e¤ect completely and underestimate capital intensity of exporters. Moreover, the within-transformation does not address the issue of simultaneity between variable inputs and contemporaneous productivity shock and will deliver downward biased estimate of K;E when technology adoption is positively related to input stocks. The true value of K;E should, therefore, lie in between OLS and xed e¤ect estimates, which provides a useful check for the results from other methods of estimation. To control for endogeneity of input factors two estimation methods were used. The rst one is a two-step semi-parametric estimator developed by Olley and Pakes (1996), henceforth OP estimator. It uses rm-level choice of investments to proxy time-varying unobserved heterogeneity, and corrects for sample selection bias resulting from rms self-selection decision to exit.34 I also considered GMM system estimator proposed by Blundell and Bond (1998). 4.2.4 Empirical Results Table 4.2 presents OLS estimation results for the standard production function speci ed in equation (23). The rst four rows show exporters premium in productivity, capital intensity, labor intensity and material intensity, respectively, while the next three rows present output elasticities of three factors of production of non-exporting rms. Column (1) estimates the basic production function with a dummy variable for export participation. Shares of capital and labor in the nal output are estimated at 10% and 48%, respectively, and the scale economies is estimated at 0:96. OLS estimates point at 8:6% increase in productivity when a rm starts exporting to foreign markets. The positive coe¢ cient on export participation variable is very signi cant and persistent, which support results of many previous studies on exportersproductivity advantage over non-exporters. The second column includes interaction of xed capital and export status, which enters the equation positively and is highly signi cant. It says that capital productivity of exporters is 28% 34A potential problem with OP estimator is a large number of rms with zero investment observations, which violates investment monotonicity condition. Although the number of such observations does not exceed 15%, I also used Levinsohn and Petrin (2003) estimator that identi es productivity shocks from the demand for intermediate materials. 82 higher (2:68 percentage points relative to baseline capital elasticity of 9:41), and doubling the capital stock increases productivity by 12:1% if the rm is exporting and only by 9:4% if it operates only domestically. This result is preserved when interaction of export status with employment is introduced in the third column. Finally, column four estimates production functions for exporters and non-exporters separately and results suggest that rms that export operate with di¤erent technology. In particular, column (4) demonstrates that exporting rms have higher productivity with respect to capital and intermediate inputs and lower productivity with respect to labor. It is also important to note that in all speci cations considered the average scale economies of exporters is only marginally higher than for other rms. The last ve columns report estimates of the production function with establishment xed e¤ects. Removing rm-speci c productivity e¤ect from equation (23) reduces the bias in estimates of factor input elasticities arising from their correlation with unobserved plant-speci c productivity, but at the same time bias the estimates of export dummy and factor usage by exporters towards zero. As a result, the total factor productivity premium of exporters reduced to 1:3 3:4%, and capital productivity premium fell to 10 13%. However, all of the estimates are still of correct signs and signi cant, which con rms previous result that exporters operate with di¤erent production technology. Results are very similar when value added speci cation of the production function is used in column (10). Table 4.3 shows results for a more general speci cation of production function with a full set of input interactions and order terms (translog speci cation). These are not presented in the table, although most of them are statistically signi cant and they are always jointly signi cant. The translog speci cation imposes no restrictions on production technology and is non-homothetic, i.e. controls for the size di¤erences across rms. However, the key results on interactions of factor inputs with the export status are quite similar to the ndings discussed above. Not only exporters are more productive than non-exporters, they also use more capital-intensive and less labor-intensive technology. In previous sections it was argued that the OLS and xed e¤ect coe¢ cients reported in Tables 4.2 and 4.3 are inconsistent estimates of factor elasticities since factor usage and technology adoption are correlated with time-varying rm-speci c productivity shock, absorbed by the error term. To account for endogeneity of factor inputs I used a semi-parametric estimation technique proposed by Olley and Pakes (1996) and a System GMM estimator of Blundell and Bond (1998). The OP 83 approach also controls for endogenous exit by least productive rms, which can contribute to an upward bias of the export coe¢ cient due to the sample selection problem. As it was discussed in Section 4.2.3, the presence of time-varying rm-speci c productivity shock, that a¤ects the choice of inputs and production technology, tends to bias the coe¢ cient on K;Eit upward in the simple OLS estimation and downward in speci cation with xed e¤ect. Estimates of the true parameter should, therefore, lie in between of these values. The results for production function estimation with OP and system GMM estimators are pre- sented in Table 4.4. Olley and Pakes (1996) show that endogeneity of factor inputs bias capital coe¢ cient downwards, which is con rmed by higher estimates of capital elasticity in Table 4.4. As a result, the scale economies estimated by OP method is higher than that by OLS, and one cannot reject the hypothesis of constant returns to scale in speci cation (1) in Table 4.4. The new point estimates of the coe¢ cient on exporterscapital intensity in columns (2) to (5) are always positive and signi cant, which supports the results obtained previously with OLS and xed e¤ect regres- sions. Moreover, all of these coe¢ cients are within the desired range and closer to OLS than to xed e¤ect estimators. Exporters are estimated to be from 8:3% more capital intensive (speci cation 4) to 25:1% (speci cation 2). Again, it is important to note that scale economies of exporters are not di¤erent from scale economies of rms that operate only domestically. Controlling for endogenous exit from the sample has little e¤ect on the results, except for a small reduction in export dummy coe¢ cient. Column (5) presents OP estimation results for the case when the rm is considered as being exporting only if it sells at least 10% of its output in foreign markets. This reduces the share of exporting rms from 34 to 16% of the whole sample. According to column (5), when exporters are more narrowly de ned their production technology is essentially the same as was estimated previously. The system GMM estimates in columns (6) and (7) also indicate the statistically signi cant di¤erences between the production technology of exporters and non-exporters. Exporting rms are estimated to be 30 50% more capital intensive. However, the robustness of these results is questionable and we will stick to OP as a preferred estimation method.35 35The rejection of the null hypothesis of the Sargan-Hansen test raises the concern of the validity of using forth order lagged inputs as instruments even in the absence of third order autocorrelation in residuals. 84 4.2.5 Robustness Tests Selection of exporting rms into more capital-intensive industries. A potential problem with the measure of exporterscapital intensity is a possible higher capital productivity of rms operating in more export-intensive industries. This is particularly likely to be the case for capital- abundant countries like France, which has a comparative advantage in producing more capital- intensive products. In this case parameter K;Eit would capture industry-speci c capital intensity rather than a characteristic of an exporting rm. To deal with this problem, the model (23) was augmented by including the interactions of capital and labor variables with the share of exporters in an industry (Table 4.4, columns 8 and 9). Although the capital interaction is positive and labor interaction is negative and signi cant, suggesting that export-oriented industries in France are indeed more capital-intensive, the coe¢ cient K;Eit remains positive and statistically signi cant. The e¤ect of export intensity on production technology of exporters. Another interest- ing empirical issue is how export intensity a¤ects the production technology. To understand the responsiveness of technology adoption to trade policy it is critical to understand whether capital in- tensity is a¤ected by export status or by export intensity. Column (10) of Table 4.4 adds continuous measure of the export intensity to the estimation of production function of exporters, where export share is measured as a ratio of export sales to total revenue. Positive and signi cant coe¢ cient on the interaction of export share with capital stock would imply a continuous relationship between the choice of capital intensity and productivity of exporter. The estimation results in column (10) of Table 4.4 show that higher export intensity is associated with higher productivity: one percentage point increase in export intensity leads to 0:065% increase in productivity. The e¤ect is quantitatively small36 but statistically signi cant and supports the theoretical prediction of the Melitz (2003) model that more productive rms export larger fraction of their output. The coe¢ cient on the capital stock interacted with the exports share is positive and signi cant, suggesting that rms that export more of their output abroad use more capital-intensive technologies.37 Firms that export all of their output are 7:5% more capital intensive than rms that receive only 1% of revenue from foreign sales. Moreover, coe¢ cients on labor and materials 36The export-intensity of an average exporting rm is 6.5%, which implies only 0.4% productivity advantage over non-exporting rms arising from the intensive margin of exporting. 37At the same time, the relationship between capital intensity and export intensity is concave since the coe¢ cient on the squared interaction term is negative and signi cant. 85 interacted with the export share are negative and positive, respectively. This result implies that the relationship between export status and technology adoption, identi ed in Section 4.2.4, is ampli ed by the export intensity. However, for the rm with the median export intensity of 8:6% the size of the latter e¤ect becomes very small, suggesting that the extensive margin of exporting is more important for the choice of production technology than intensive margin. Industry-level evidence. As another robustness check I veri ed that the main result of this section holds not only at the economy level but at the industry level as well. Table 4.5 presents OP estimation results for ten individual NAICS3 manufacturing industries with largest total output.38 In all industries under consideration exporters are characterized by productivity advantage, except for chemical products manufacturing (NAICS 325). In non-metallic mineral products, paper, pri- mary metal, fabricated metal products, food, and transportation equipment exporters are estimated to use more capital intensive technologies, while in plastic and rubber, computer and electronics and chemical products there is not enough evidence to draw the same conclusion in spite of the positive K;E coe¢ cient. Only for machinery the coe¢ cient K;E is negative but statistically insigni cant. Overall, out of 21 NAICS3 industries the coe¢ cient on the interaction of capital and export status was estimated to be positive and signi cant for twelve industries, positive and insigni cant for six and for the remaining two it is negative but insigni cant. So the evidence supports that in more than half of all industries exporting rms operate with the more capital-intensive production technology. Production technology before and after exporting. In this section we want to investigate how rms adoption of a new technology is related to its decision to enter foreign markets. To analyze the transition of a technological process I estimated the production function of rms before and after they start exporting. Columns (1) to (5) of Table 4.6 contrast production function of current exporters and rms that will become exporters in the next year. Across all speci cations future exporters are more productive than rms selling only domestically, and signi cantly less productive than exporting rms, which supports many previous rm-level studies of exporting decision. Coe¢ cients on capital intensity of future exporters are mostly positive and signi cant, but smaller than capital-intensity 38Estimation results for the seventh largest industry Petroleum and Coal (NAICS 324) are not reported due to the small number of rms in that industry and inability to identify all parameter estimates. 86 of current exporters and the di¤erence is statistically signi cant. In terms of labor and material intensity of future exporters, regressions produce mixed results but in general one can safely reject the hypothesis of identical production functions for future exporters and non-exporters. Similarly, the hypothesis of identical technologies of current and future exporters is strongly rejected. In other words, the characteristics of exporting rms can, to some extent, be observed at the rms that will enter the exports market next year. In columns (6) to (10) of Table 4.6 future exporters are re-de ned as rms that would enter exports market in two years. Two years before becoming exporters, rms are still more productive than other non-exporting rms, however, OP estimator suggest that future exporters are not dif- ferent from other non-exporters with respect to any input factor usage. These results imply that rms start adopting more capital-intensive technologies approximately one year prior to entering exports market. In a similar way Table 4.7 estimates production function for new and old exporters using di¤erent econometrics techniques. In columns (1) to (5) the production function of rms who just entered the exports market is compared to all other exporters. We see that rst-year exporters are still slightly less productive and use almost as much capital as other exporters, but the hypothesis of identical production functions is always strongly rejected. In columns (6) to (10) the same production function is estimated where new exporters are de ned as rms that sell on foreign markets for two consecutive years. These second-year exporters are not statistically di¤erent from other exporters in terms of productivity premium and, more importantly, the hypothesis of identical production functions cannot be rejected in neither speci cations. That is rms with two years of exporting experience operate with the same technology as other exporting rms. Hence, these results are consistent with the idea that it takes some time for rms to restructure their production technologies to compete on global markets. Firms start using more capital-intensive technology one year before becoming exporters and this transitions process takes for about three years. Note that during the whole transition period the more capital-intensive technology operates with the same scale economies, which is true for both speci cations in sales and value added. Robustness to the choice of a functional form. A possible concern with the higher estimated capital intensity of exporters is the restriction on the elasticity of substitution being equal to one, imposed by the Cobb-Douglas production function. If the substitution elasticity were greater than 87 one, the di¤erence in capital-labor ratios implied by the di¤erence in factor prices of exporters and non-exporters would lead to the overestimated capital intensity of the former ones. To tackle this issue I estimated technological di¤erences between exporters and non-exporters, assuming a constant elasticity of substitution (CES) production function of the form: Yit = Ait h KK it + LL it + MM it i + uit (24) controlling for time and time-industry xed e¤ects. Equation 24 was estimated by non-linear least squares and the results are presented in Table 4.8. We see that all estimates are statistically signi cant and economically plausible. The return-to-scale parameter is essentially equal to one and the assumption of constant returns to scale is veri ed. Negative estimates of suggest the values of the elasticity of substitution = 11+ of around 3, irrespective of the export status. However, the coe¢ cient on capital K is always considerably greater for exporting rms relative to the whole sample. Although this coe¢ cient does not show the capital share in the nal output for CES speci cation, this share is still increasing in K ,39 con rming the result that exporters are more capital intensive than non-exporters. 4.3 A Model of Technology Adoption and Trade It was established in the previous section that French rms participating in exporting activities ap- pear to operate under more capital-intensive production technology than rms selling only at home. This section presents a simple general equilibrium model of rmsdecision to export and the choice of technology that could potentially explain the rm-level di¤erences in production technologies observed in the data. The model is based on Melitz (2003) model of monopolistic competition with heterogeneous rms and extended to allow rms to choose optimally the production technology. Then this model will be used to analyze the impact of trade policy on the incentives to use more capital intensive technologies and aggregate productivity gains. Throughout the rest of the essay I will distinguish TFP from rm-level productivity. By TFP I would assume the Solow residual measure of productivity shock, or the fraction of the output not explained by capital or labor inputs. At the same time, the simple productivityof the rm 39For CES production function the capitals share of output is sK = KK it KK it + LL it + MM it , and @s K @K > 0. 88 would imply the unit output e¢ ciency, or the inverse measure of the unit cost of production. In the standard literature where all rms use the same production technology there will be no di¤erence between these two measures of e¢ ciency, however in the model developed here the production technology may imply a lower per unit cost of production, holding the TFP parameter constant. Consider two countries, home and foreign, where foreign country variables are labeled with a star. Following the literature, the variables related to domestic market I denote by h and the vari- ables related to exporting by x. There are two sectors in the economy of each country: homogeneous good sector and di¤erentiated goods sector. 4.3.1 Demand Side A representative consumer in each country has a Cobb-Douglas preferences over homogeneous and di¤erentiated goods and CES preferences over N varieties of a di¤erentiated product: U = z1 0@ NX j=1 x 1 j 1A 1 where is the share of consumers expenditure on di¤erentiated products and > 1 is the elasticity of substitution between varieties. Each country is endowed with non-depreciating labor stock L, supplied inelastically by con- sumers on the competitive labor market at price !. Maximizing this utility function subject to the standard budget constraint, we obtain aggregate industry-wide price index (P ), and the demand and revenue functions faced by each rm in the economy (xj and ej , respectively): P = 0@ NX j=1 p1 j 1A 11 xj = I pj pj P 1 (25) ej = I pj P 1 (26) where I = !L is the consumers labor income. 89 4.3.2 Production Side There is a continuum of potential rms that can freely enter the market for homogeneous product. They all share the same linear production technology that uses only labor. Labor is supplied on a perfectly competitive market, which pins down the equilibrium wage rate. For simplicity, assume that both countries are equally productive in homogeneous goods sector such that there are no cross-country wage di¤erentials. To enter the di¤erentiated goods market each rm has to make a sunk entry costs fe, measured in the units of nal output. Upon entry to the market, a rm receives a productivity draw A from the distribution function G(A), which is constant over the life cycle of the rm, and decides whether to exit or stay in the industry. Firms produce nal output using labor and capital according to the Cobb-Douglas production function with constant returns to scale and rm-speci c productivity Aj : Yj = AjKj L 1 j . Solution to the cost-minimization problem provides the following total cost function Cj : Cj = j (Dj + F ) j = 1 Aj r ! 1 1 (27) where Dj is the quantity demanded of variety j, F is the xed cost of production measured in terms of nal output, and r and ! are capital and labor prices, respectively. It follows that the marginal costs of production are constant and equal to j . Each period a rm receives an extreme cost shock with probability that forces it to leave the industry. Since rms exit is associated with a negative shock to its productivity, the more productive rm is less likely to be hit with a bad shock large enough that makes the rms value non-positive. It follows that the probability of a bad shock (A) that forces a rm to exit is thus a decreasing function of productivity: @(A)@A < 0. 40 4.3.3 Capital Market Capital is produced from labor with a constant returns to scale technology: K = Lk, where Lk is the amount of labor allocated to capital production. Consumers own capital and inelastically 40 In Section 4.4.1 I provide an empirical evidence on the positive relationship between rms survival probability and productivity. 90 supply it to a risk-neutral perfectly competitive nancial intermediary at the market price r. The intermediary in turn lends capital to rms. Perfect competitiveness implies that the price of capital observed by rms should include a risk premium, and since the chance of capital being repaid by a rm is decreasing with its productivity, the risk premium is lower for more productive rms.41 It is also assumed that productivity of each rm that sells on the market is perfectly observable to the nancial intermediary since without any uncertainty the productivity parameter A can be revealed from the observable pro t function and rms size. Therefore, a rm j that operates on the market hires labor on the competitive market and borrows capital at the competitive price rj = r+(Aj).42 However, the productivity of rms that enter the market is unobservable and the risk premium for the new rms simply equals to the ex-ante probability of survival: re = r + (E(A)). 4.3.4 The Choice of Technology Prior to entry, each rm chooses the same production technology Yj = AjKj L 1 j , which is optimal given the expected productivity draw. However, once productivity is revealed, a rm may choose to re-organize its production: Yj = AjK +T j L 1 T j where 2 (0; 1 ), and T 2 [0; 1] is a set of available technologies normalized to a zero-one interval.43 Let h(T ) be a technology adjustment cost function measured in terms of a nal output with h0(T ) > 0 such that more capital-intensive technologies require more investments. Then each rm chooses T to maximize the discounted stream of pro ts: T j = argmax ( 1X t=0 (1 j)tj ) (28) tj = etj(j(Tj)) j [F + (Aj)h(Tj)] (29) 41Another reason why larger and more productive rms face lower interest rates that would lead to similar e¤ect is that they have better access to the capital market and can reveal more company information to nancial intermediates. 42There is an extensive literature that documents a negative relatioship between interest rate and rm size, e.g. Bond (1983), Berkowitz and White (2004). In section 4.4.1 I show that for the French data the elasticity of interest rate, calculated as a ratio of debt payment over long term debt, with respect to output is 0:11 with t-statistics 119:5. Similarly, the elasticity of the interest rate with respect to total factor productivity is 0:28 with t-statistics 76:48. 43 In line with the empirical result that both exporters and non-exporters operate with the same economies of scale, it is assumed that all available technologies are CRS. 91 j(Tj) = 1 Aj rj + Tj +Tj ! 1 Tj 1 Tj (30) The rst-order condition for maximization problem (28) is given by: (1 ) e ((Aj)) (Aj)(F + (Aj)h(T j )) ln 1 T j + T j rj w ! @h(T j ) @T j = 0 (31) The expression in square brackets shows the increase in revenue ( rst term) and decrease in the marginal technology adoption costs function implied by productivity increase (second term) from a marginal increase in T , which in equilibrium should equal to the absolute change in technology adoption costs. 4.3.5 Entry, Exit and Exporting Decisions If the rm decides to export its product to another country, in addition to the xed costs of production it needs to pay xed costs of exporting fex and variable costs of an iceberg form, so that an amount > 1 of a nal product must be shipped in order to sell one unit of output abroad. Fixed costs of exporting are modelled as an amortized per-period payment fj;x = (Aj) fex. Observing isoelastic demands for its product, each rm sets pro t-maximizing prices on the domestic and foreign markets as a constant markup over its marginal costs: phj = 1j pxj = 1j = p h j (32) Then, from equation (26), the revenue earned by a rm on the foreign market is a fraction of its revenue from domestic sales: ehj = 1exj . Similarly, from (29), the rms pro t from domestic and foreign sales can be represented as hj = ehj j [F + (Aj)h(T )] xj = exj j (Aj) fex (33) 92 And demand for the factors of production are given by Kj = Dj+F Aj ! (1 )r 1 = r (Dj + F )j Lj = Dj+F Aj (1 )r ! = 1 ! (Dj + F )j (34) Once the rm made an entry investment, its productivity is revealed and the decision should be made whether to stay on the market or not. If productivity is not high enough to cover the xed costs, a rm would decide to withdraw immediately. Therefore, the rm makes an exit decision by maximizing the rms value, which is the highest of its closing value, assumed to be zero, and the discounted stream of pro ts: (A) = max ( 0; 1X t=0 (1 (A))t (A) ) = max 0; (A) (A) Since (A) is an increasing function of productivity, denote the lowest (cuto¤) value for A required for the rm to earn non-negative pro t on the domestic market by A = inf fA : (A) 0g. Similarly, the productivity level Ax = inf fA : A A and x(A) > 0g identi es the cuto¤ value required for entering the foreign market. Given the ex-ante probability distribution function of productivities is G() with a positive support on a [0;1] interval, the fraction of exporting rms (the probability of being exporter) is equal to px = 1 G(A x) 1 G(A) . The number of exporting rms, Mx, is thus a fraction p x of all rms that operate on the home market, M , and the total number of available varieties of a product is Mv =M +Mx = (1 + p x)M 4.3.6 Equilibrium Distribution of Productivities The rmsequilibrium entry and exit rules determine the evolution of the distribution function of observed productivities: t+1(A;A ) = s(1 (A))R1 A(1 (A))t(A;A)dA t(A;A ) + (1 s)g(A) 93 with s = MM+Me denoting the share of the incumbent rms on the market, and g(A) = G 0(A) denoting the probability density function of productivities of new entrants. The rst term reects the change in the distribution of productivities due to the fact that less productive rms are more likely to exit. The second term captures the contribution of new entrants to t+1. The resulting equilibrium steady state distribution function of observed productivities is given by: (A;A) = (1 s)g(A) 1 sC (1 (A)) (35) where the constant C is determined from the equation C = Z 1 A (1 (A))t(A;A)dA = Z 1 A (1 s)(1 (A))g(A) 1 sC (1 (A)) dA 4.3.7 Open Economy Equilibrium I impose the following time structure on the problem of a rm that considers entering the market. At the rst stage, the rm makes a sunk entry investment fe. On a second stage, its productivity is revealed, which becomes a common knowledge and determines the interest rate for capital. Observing the capital price, the rm chooses how much to invest into production technology and whether to export or not. Next, the rm decides how much to produce and what price to charge on each market, and how many workers to hire given the market wage rate. Finally, the realization of the exit shock (A) is revealed. Denote by e = e(A) the average marginal costs of all domestic rms, weighted by output shares, and ex = e(Ax) the marginal costs averaged across all exporting rms: e(A) = 241Z A (A)1 (A;A)dA 35 11 fx(A) = 2641Z Ax (A)1 (A;A)dA 375 1 1 (36) Let et denote the industry-average marginal costs in the open economy with variable trade costs : et = 1 Mv Me 1 +Mx(ex) 1 1 1 (37) 94 Average revenue and pro t from domestic and foreign sales can also be expressed through average cost function and corresponding productivity cuto¤ value. From (26), enem = (An) (Am) 1 for any pair of rms n and m. Therefore, average revenue of the rm that sells domestically can be expressed as eh = eh(e) = e(A) (A) !1 e((A)) and from (33) average domestic pro t is h = h(e) = e(A) (A) !1 e((A)) e(A) hF + (Aj)heTi Following Melitz (2003), the zero cuto¤ pro t condition (ZCP) requires the pro t of a rm with the marginal productivity A to be equal to zero, which is satis ed when e((A)) = F(A), or equivalently: h(e) = Fe(A)k(A) (1 G(A))M ee(A)heT e k(A) = h (A)e(A) 1i (38) where M e is the number of new rms that paid entry costs, eT e is the output-weighted technology adoption by new entrants, and (1 G(A))M e is the number of successful entrants. Furthermore,eT e is completely determined by the threshold productivity level A as it can be observed from (31). Additionally, the pro t of an exporting rm with the cuto¤ productivity value for exporting must be zero as well: x(ex) = fexe(Ax)kex(Ax) kex(A x) = h (Ax) (Ax)e(Ax) e(Ax)ie(x) = R1x (A) q()q(e) (A)dA (39) Combining (38) and (39), the aggregate ZCP condition in the open economy becomes: = h(e) + pxx(ex) (40) Condition (40) de nes the rst equilibrium relationship between average pro t and a cuto¤ productivity level A. Note that in (40) the cuto¤ productivity level for exporting is an im- 95 plicit function of A. From (26), enem = pn pm 1 for any pair of rms n and m. In particular, eh(A ) ex(Ax) = (A) (Ax) 1 . Furthermore, from zero pro t condition for a rm with a threshold value of productivity the following equalities must hold: eh(A ) = (A )F and ex(A x) = (A x)(A x)fex. These conditions imply that the cuto¤ value for entering exports market can be expressed as a function of a domestic market cuto¤ value: (A) = 1 (Ax) fex(A x) F 1 (41) If the rm decides to enter the market, it should make a sunk entry cost fe. The free en- try condition requires that the rm value of a potential entrant should be equal to zero, e = 1 G(A) [E(AjAA)] fe = 0, and provides a second equilibrium relation between average pro t and marginal productivity: = [E(AjA A)] 1 G(A) fe (42) where (1 G(A)) is the probability of successful entry and [E(AjA A)] is the ex-ante proba- bility of a bad shock. Conditions (40) and (42) determine the open economy equilibrium and are similar to condition (12) in Melitz (2003) but in the presence of capital in the production function, variable probability of survival, and endogenous technology choice by rms. In the same way, these conditions identify a unique equilibrium values of A and . Note that the variance in technologies across rms does not a¤ect the equilibrium conditions explicitly. Once the industry average marginal costs is known, it is possible to derive all other long-run equilibrium aggregate economic variables for the open economy. Knowing the pricing rule (32), the equilibrium price index can be expressed in terms of e: P = 1M 1 1 e =M 11 p(e) Similarly, consumption index, aggregate revenue and pro t can be expressed through the 96 industry-average marginal costs: Q =M 1 q(e) E =Me(e) =M e(e) e hF + (Aj)heTi =M(e) Factor demand conditions (34) imply that total payment to labor and capital used in the production must be equal to total costs (the revenue net of the pro t), while market clearing condition for the nal product requires the equality between total revenue and total expenditure: E = I = !L. Together, these conditions imply that total payment to investment workers should equal to aggregate pro t: !Le + rKe = M efe = . The equality of aggregate revenue and con- sumers income determines the total number of incumbent rms on the market, while the equality of aggregate pro ts and investment costs pins down the equilibrium number of new entrants: M = !Le M e = Mfe (43) Normalizing the wage rate to one, the equilibrium risk-free capital price follows from the no- arbitrage condition on factors market that labor income in nal goods sector and in the production of capital should be the same: ! = 1 = r (44) Conditions (40) and (42) characterize an open economy equilibrium, which determines equi- librium average pro t, cuto¤ productivity level and relative factors prices. Conditions (31), (35), (41), (43) and (44) complete characterization of the stationary equilibrium. 4.3.8 The E¤ect of Trade Barriers Reduction Now we can perform some comparative statics analysis of the models stationary equilibrium and look at the e¤ect of a trade policy change on the open economy equilibrium outcome. Our prime interest is the e¤ect of a reduction in trade barriers followed from a decline in tari¤ or non-tari¤ barriers. The e¤ect of a tari¤ reduction can be observed through comparative static analysis of (40) and (42) conditions with respect to . Denote x and x0 the value of a variable x before and after policy change, respectively. Inspection 97 of conditions (40) and (42) reveals that the reduction in variable trade costs from to 0 < will deliver a new ZCP curve that lies above the old one, keeping the FE curve constant. When trade barriers are reduced, domestic rms would be faced with increased competition from the side of the most e¢ cient foreign exporters and will lose some of their domestic market share to foreign rms. The least productive domestic rms with productivity level close to A would not make enough revenue to cover xed costs of production and would be forced to exit, so that the threshold level A would increase. On the other hand, the most productive rms will enter the exports market: from (41), a reduction in import tari¤ followed by the increase in A implies a decrease in productivity threshold required to enter the exports market: Ax > A 0 x . From (40) and (42) it directly follows that the average pro t and average revenue will increase in the economy with lower trade barriers, while the number of domestic rms will decrease.44 At the same time, the total number of varieties available in each country will expand (Mv > Ma) and the increase in the number of exporters outweighs the number of exiting rms. The decline in the output-weighted average costs e0 < e can be interpreted as a reduction in aggregate costs (increase in aggregate productivity) when trade distortions are reduced and reects a countrys gain from within industry resource reallocation as a result of openness to international trade. We now want to look at the e¤ect of trade on the adoption of more capital-intensive technologies. First note that the FOC (31) and the fact that trade rises average revenue and lowers average costs implies that the industry-average technology adoption eT necessarily increases in the economy with smaller variable costs of trade for any technology adjustment cost function h (T ). The reason is that the rms market share reects its ability to exploit the bene ts of a new technology adoption: the greater revenue raises the returns to the new technology and results in an upward shift of a marginal bene t curve of technology adoption as a function of T . Therefore, as long as openness to trade raises the average size of the rm, the average capital intensity would increase as well. The average increase in capital intensity is comprised of two e¤ects. The rst one, the compo- sition e¤ect, follows from reallocation of market shares from the least productive rms to the most productive ones who have already installed the more capital-intensive technologies. Therefore, exit 44To see that recall that from (26) en em = (An) (Am) 1 for any pair of rms n and m. Con- trasting average revenue before and after tari¤ reduction we get e e0 = ee0 1 . Since @e(A)@A = 1 1 e(A) R1 A (A) 1 A(A;A )dA (A)1 (A) < 0, the result that A0 > A implies e0 < e and e < e0. Then using (43) we obtain the required result: M > M 0. 98 of the least capital-intensive rms and the following change in the composition of rms on the market is one of the reasons why eT increases in the open economy. The second e¤ect is due to the change in the structure of technology adoption across survived rms. To better understand the structural e¤ect we need to analyze how trade a¤ects a rms incentives to change its production technology. Inspection of the FOC (31) reveals that trade a¤ects the choice of T j by rm j only through the e¤ect on revenue. As it was previously discussed, if rms revenue increases as a result of a policy change, the increase in the marginal bene t of technology adoption raises T j , unless T j = 1 or marginal bene t of technology adoption is non-positive. Therefore, using the insight of Melitz (2003), who showed that openness to trade raises revenue of exporting rms and decrease that of non-exporting, the e¤ect of trade on the structure of technology adoption can be summarized by the following proposition: Proposition 12 When import tari¤s go down, technology adoption by exporters in the economy with reduced tari¤s would (weakly) increase and (weakly) decrease by non-exporters. The idea of Proposition 12 can be illustrated with Figure 4.1, where rms productivity is mapped to technology adoption rate and marginal costs. Pre-trade liberalization relationships are shown as solid lines. As Figure 4.1 shows, a rm should rich a certain level of productivity A0 when capital price becomes low enough for investment in technology adoption to be worthwhile, i.e. a rm with productivity A A0 will choose T = 0. Furthermore, all rms with productivity above A1 will choose T = 1. Figure 4.1 shows three possible e¤ects of tari¤ reduction on a rms technology adoption rate. Panel A represents the case when technology adjustment costs are relatively high so that only the most productive rms would choose positive T and none of the non-exporters has an incentive to change its technology after tari¤ cut. In this case, the reduction in will shift both A0 and A1 towards Ax, and the distribution of capital intensities and marginal costs will change from the solid to the broken line. Non-exporters and exporters with productivity above A1 continue to use their old technologies after policy change. It is the rms with productivity levels in the interval h A 0 0;A1 i who respond to the increase in the revenue with increased investments in capital-intensity. The ability to use cheaper factor of production more intensively allows these rms to cut down production costs and represents 99 an additional source of increase in revenue and pro ts of exporters. Furthermore, a change in the distribution of marginal costs leads to the reduction of the industry-average marginal costse and contributes to an increased welfare per worker. Thus, the e¤ect of trade on productivity through reallocation of market shares towards more productive rms, identi ed by Melitz (2003), is ampli ed by the rmsdecisions to invest in technology improvements. Panel B represents the opposite case when technology adjustment costs are relatively low or survival probability is high, in which case the productivity threshold level for entering exports market Ax is above A1, and trade liberalization does not a¤ect the adoption of technology by the exporters. However, tari¤ reduction does lower revenues of non-exports, making more capital- intensive technology less pro table for domestic rms, and both A0 and A1 shift towards Ax. As a result, rms with productivity levels in the range h A0;A 0 1 i would switch to a less capital-intensive and less productive technology, raising the industry-average marginal costs e and decreasing welfare per worker. Finally, panel C of Figure 4.1 represents an intermediate case in which the productivity threshold level for exporting lies in-between A0 and A1. In this case, exporters with productivities in the interval [Ax;A1] would install more productive and more capital-intensive technologies once trade distortions are reduced, while non-exporters with productivities in the range [A0;Ax] would choose to have lower T . Although the exact e¤ect of technology adoption on aggregate productivity is indeterminant, we already know that the industry-average technology adoption T necessarily increases in the open economy. In fact, even if the average productivity of survived rms would decrease in the open economy, the e¤ect from reallocation of market shares from less capital- intensive rms to more capital-intensive exporters will dominate. 4.4 Quantitative Analysis This section describes the quantitative version of the model from Section 4.3. The quantitative analysis allows to measure the relative strength of the technology adoption e¤ect on aggregate productivity relative to the original e¤ect of trade on within-industry reallocation of labor identi ed by Melitz (2003). It explores the quantitative implications of trade barriers on the aggregate economic variables in a series of counterfactual experiments of trade liberalization. 100 4.4.1 Parametrization In this section we discuss the functional forms and parameter values used in the simulation exercise. The endowment of labor in each country is normalized to L = 100, and is set to be equal to 3 to reect 50% producersmarkup over marginal costs. The xed costs of exporting were set to match the fact that 34% of French rms operate on foreign markets. This implies that in the steady state a xed export costs equal to approximately 16% of the export revenue for an average exporter. Since all statistics are invariant to proportional changes in xed costs of entry, export and production, sunk entry costs and xed costs of production were both normalized to one: fe = 1; F = 1. Variable trade costs was measured in two ways. First, was proxied by the trade-weighted tari¤ for French imports, which was equal to 0 = 1:08 in 1997 and 1 = 1:03 in 2005. Second, 0 and 1 were set to deliver export intensity of manufacturing industries observed in 1997 and 2005 that have increased from 0:24 to 0:33. This gives the values of 0 = 1:53 and 1 = 1:3. Following a general approach in the literature, I parameterize the distribution function of pro- ductivity draws by new entrants G(A) to be Pareto with shape parameter and lower bound normalized to 1. The shape parameter was set to match 10% productivity advantage of French exporters observed in the data. This delivers the value of = 3, which is very close to 3:4 obtained by Bernard, Eaton, Jensen, and Kortum (2003) for US rms. The probability of exit shock is para- metrized with the logistic function of rms productivity (A) = (1+exp(d0+d1A)) 1 where d0 and d1 are scalars estimated from the data.45 The risk-inclusive capital price (r + (A)), measured in the units of labor, was obtained from the estimated survival probability by scaling it proportionally in such a way that even marginal rm would have incentive to increase capital intensity. To solve the model, I also parameterize the technology adoption cost function h(T ) in a way that it could reproduce the technology adoption patterns observed in the data. In particular, since marginal cost is a decreasing function of capital intensity, h(T ) should be an increasing and convex function of technology adoption T in order to be able to generate heterogeneity in technologies across rms. The speci c functional form that I consider is a power function h(T ) = h0T h1 , where the curvature parameter h1 > 1 reects an increasing cost of capital that requires less labor in the 45Parameters d0 and d1 were estimated from the logit model of the rms survival probability on productivity, measured by value added per worker, controlling for rms age, legal status, geographical location and industry. Alternatively, productivity was measured with tted residuals from OP estimator of the production function (without export dummy and factor interactions) and adjusted to account for di¤erent scales by matching the survival probability of 25th and 75th percentiles of productivity in simulation and logit model prediction. Both methods give very similar statistically signi cant estimates of d0 = 2:9 and d1 = 0:08. 101 production process, and h0 is a scale parameter. These two parameters jointly determine average capital intensity of exporting and non-exporting rms, which in Section 4.2.4 were estimated to be 0:22 and 0:24, respectively (Table 4.4, speci cation 4). Choosing h0 and h1 jointly to match these statistics for every given and allows to pin them down, and for the benchmark calibration yields the values of h0 = 3:6 and h1 = 3. Finally, we discuss the choice of parameters and . As it was illustrated in Section 4.3.8, these coe¢ cients are very important for the e¤ect of tari¤ reduction on the incentives to adopt more capital-intensive technologies. If prior to trade liberalization all domestic rms used the basic technology ( = 0:22), then only exporters will respond to the reduction in trade barriers by switching to more capital-intensive production processes (Figure 4.1a). On the other hand, if all exporters already used the most capital-intensive technology available, then only non-exporters will react to tari¤ reduction by switching to more labor-intensive technologies. The latter case can be safely ruled out since it was estimated that higher export intensity, and hence higher productivity, is associated with the use of the more capital-intensive production process (Table 4.4, speci cation 10). Moreover, capital intensity of the most productive rms that export most of their output is 0:0118 higher relative to other exporters. This condition was used to obtain the parameter , which for the benchmark speci cation equals to 0:1. As for the parameter , it cannot be calibrated from the data and, given the lack of any direct evidence on the size of this coe¢ cient, I tried several simulations. In the benchmark simulation I set = 0:22 to isolate the positive e¤ect of technology adoption for exporters, and then checked how sensitive are the nal results to changes in this parameter. A reduction in the variable trade costs leads to greater reduction in capital intensity of non-exporters when get smaller. 4.4.2 Steady State Distribution of Capital Intensities The main objective of this paper is to provide a model that would explain how technological dif- ferences across rms a¤ect aggregate productivity. In this context, it is necessary to check how close to the real data is the distribution of technologies, generated by the calibrated model, since these di¤erences in technologies play a key role in the e¤ect of technology adoption on aggre- gate productivity. This section describes the match between the within-industry distribution of capital-intensities implied by the model and that for French rms, and what is the contribution of 102 technology adoption e¤ect to the ability of the model to reect distribution patterns observed in the data. Figure 4.2 shows the distribution of ln(K=L) ratio observed in the data and implied by the models with and without endogenous technology choice.46 Note that the log right tail of this distribution is virtually on the straight line.47 If there were no di¤erences in factor prices, than the capital-labor ratio would be identical across rms and the slope of the distribution would be equal to zero. Allowing for capital price to be negatively related to productivity but assuming that all rms use the same production technology (the model with exogenous technology) can successfully replicate the right-tale distribution of capital intensities, once we adjust the slope of the former such that both lines should have the same slope to control for di¤erences in scale. However, the distribution of capital intensities implied by that model departs from the data for rms with the lowest K=L ratio. On the other hand, the distribution of capital intensities implied by the model with endogenous technology choice resembles real data very closely. For the lowest values of K=L the distribution function follows that of the model with exogenous technology choice since these are the least productive non-exporting rms. In the benchmark calibration of the model all non-exporters use the basic technology with capital intensity equal to , and the di¤erence in K=L ratio comes only from di¤erences in factor prices. The rst kink in the distribution corresponds to the increase in K=L due to entering foreign market. This raises the bene t of more capital-intensive technology, because it allows to spread the costs of technology adoption over the larger quantities of output, and results in adoption of technologies that are more productive in capital. As productivity increases and price of capital goes down, exporting rms invest more in technology adoption until capital intensity reach its maximum. After that, higher productivity would only imply lower capital price and distribution will go parallel to the one without endogenous technology choice. Therefore, the model with endogenous technology adoption by rms is more successful in generating the discrepancy in the slopes of the distribution of capital intensities and to reproduce the atter segment for small ln(K=L) ratio. 46Parameters of the model are those from the benchmark speci cation with variable trade costs of 8%. 47The data covers the entire period 1997-2005, but there is almost no change in distribution across time. 103 4.4.3 The E¤ect of Trade Liberalization Now we can focus on the implications of a symmetric worldwide trade liberalization, de ned either as a 5% decline in import tari¤ or as a reduction in variable costs of trade that leads to 9% increase in export intensity of the economy. We are particularly interested in the e¤ect trade liberalization has on the incentive to adopt more capital-intensive technologies, and in the contribution of this e¤ect to the aggregate productivity gain. Panel A of Table 4.9 presents the absolute change in selected aggregate variables followed from the reduction in variable costs of trade for the benchmark speci cation with no technology adoption by non-exporting rms (columns 1 and 4). The strength of the e¤ect of technology adoption may be understood by comparing predictions of the model with endogenous technology choice and that of the model where the set of available technologies is singleton with = 0:225.48 Columns (2) and (5) of Table 4.9 present the e¤ect of trade liberalization episodes in the Melitz model with exogenous production technology and Columns (3) and (6) show the fraction of the growth rate that is due to technology composition e¤ect. Liberalization of world trade decreases the average capital intensity of exporters but increases the economy-average capital intensity. The rst e¤ect is mostly due to the reduction in the pro- ductivity threshold level for exporting and change in the export status of the most productive non-exporters that use the basic technology. For example, when variable trade costs fall by 23%, the trade-weighted average capital intensity of continuing exporters have increase by 0:009. How- ever, non-exporters continue to use the basic technology in the benchmark model and as a result, output-weighted average capital intensity of the economy have increased only by 0:0053. Next, we can examine how these changes in technology adoption a¤ect other aggregate variables. The decrease in tari¤ lowers the export price and raises the revenue of exporters. The increased attractiveness of foreign market lowers the export productivity cuto¤ and raises the number of exporting rms by 4:4% if there is a 5% tari¤ reduction, and by 20:9% when variable trade costs fall by 23%. At the same time, Columns (3) and (6) show that these e¤ects would be 11 and 13:7% larger, respectively, in the model without endogenous technology.49 The reason is that larger exporters bene t more from adoption of more capital-intensive technologies and capture greater 48 = 0:225 is the estimator of output elasticity with respect to capital in OP speci cation with value added under the assumption that exporters and non-exporters operate under the same technology. 49 i.e., growth rates of exporting rms would be 4:9 and 24:2% in the model with exogenous technology. 104 share of foreign market, raising the productivity threshold level for exporting. For the same reason, the total number of rms will fall by 9:3% more in the model where rms can adjust the production technology. The model with endogenous choice of technology also has additional implications for the impact of trade costs reduction on the rm and industry-level productivity. The average variable costs of production for non-exporters have fallen by 3:2% in response to 23% decline in trade costs, and this e¤ect would be 12:9% smaller (2:8%) in the model without endogenous technology choice. In the benchmark speci cation when all non-exporters use the basic technology, the additional increase in productivity of non-exporters comes from increased capital-intensities of exporting rms. The following reallocation of market shares from non-exporters to exporters forces the least productive rms to shut down and raises the productivity threshold for successful entry together with the average productivity of non-exporters. The e¤ect of trade liberalization on productivity of exporting rms is negative due to the expansion in export participation and reduction in the export productivity threshold. However, tari¤ cuts raise productivity of continuing exporters through adoption of more capital-intensive technologies, and together with the e¤ect of labor reallocation from exporters to non-exporters, technology adoption e¤ect is responsible for 10:2% increase in industry-average productivity. Thus, the possibility to adjust technological process in response to di¤erent factor prices ampli- es the long-run e¤ects of trade on comparative advantage of exporters, identi ed by Melitz (2003), and contributes to the increased export-intensity of the economy. Endogeneity of technology ad- justment decision by rm also ampli es the impact of trade on industry-average productivity and output, and magni es the total e¤ect of trade liberalization on welfare per worker, de ned as a real household income, by 11:1%. Now consider the model where even the threshold rm has an incentive to change its production technology to the more capital-intensive one upon entry. In this case openness to trade has an additional negative e¤ect on aggregate productivity since non-exporting rms switch to cheaper and less capital-intensive technologies. The simulation results with = 0:2 are presented in Panel B of Table 4.9. Now the change in the average capital-intensity of both exporters and non-exporters is negative. For non-exporters the e¤ect of the removal of the least capital-intensive rms is o¤set by increased incentives to use less capital-intensive technologies and by the reduction in the export 105 productivity threshold. For rms that export the increased capital-intensity of continuing exporters is dominated by the change in the export status of the most productive and less capital-intensive non-exporters. However, output-weighted industry-average capital intensity increases even more than in the benchmark case for two reasons. First, increased variable costs of production by non- exporters magnify the e¤ect of labor and market shares reallocation towards rms that export. This, in turn, raises the value of more capital-intensive technologies for the latter ones, increasing the output and exports of exporting rms while rms that do not export shrink. Thus it is the case that if non-exporting rms also invest into technology improvement, the e¤ect of trade on the aggregate variables identi ed in the benchmark case is reinforced, which is con rmed by comparing the last three rows of Panels A and B in Table 4.9. For example, comparing results for 23% reduction in variable trade costs (Columns 4 to 6) we observe that not only the industry-average productivity has increased by 0:1% up to 3:8%, but also the contribution of technology adjustment activities to this e¤ect have increased from 10:2 to 13:8%. Consequently, there is an additional increase in welfare per worker that is slightly o¤set by the reduction in the number of available varieties. Therefore, reallocation of technology adjustment incentives across rms represents an additional channel for welfare gain followed from tari¤ reduction. A series of counterfactual experiments described in this section suggest that taking into account the possibility of using di¤erent production technologies by rms with di¤erent survival probabilities can amplify the welfare and productivity gains from trade liberalization by 10 15%. However, these e¤ects could be even larger had the di¤erence in capital-intensities between exporters and non-exporters were more pronounced.50 4.5 Conclusions This paper documents substantial di¤erences in production technologies for exporting and non- exporting rms, with the former using the more capital-intensive ones. The main objective of the paper is to o¤er a theoretical framework that could explain observed di¤erences in production technologies across rms. It extends the traditional Melitz (2003) heterogeneous rms model by 50Speci cally, if the initial capital intensity or capital price premium of exporters were larger, the e¤ect of the mechanism that ampli es productivity gain from within-industry resource reallocation could be even greater. Em- pirical evidence suggest that these di¤erences may in fact be stronger in other countries. For example, Bernard and Jensen (1999) report 19% capital to labor ratio premium for US exporters, which is three times greater than that for French exporters. Alvarez and Lopez (2005) report 60% capital-labor ratio premium of Chilean exporters. 106 allowing for the endogenous technology choice by rms in an economy where survival probability and factors prices are rm-speci c. The model emphasizes that trade liberalization raises the incentives for exporting rms to install more productive and more-capital-intensive technologies, which reinforces their competitive advantage relative to non-exporters and contributes further to the economy-wide productivity increase. Simulation results highlight that up to 10% of welfare gain from trade liberalization is a result of the change in production technology composition within the industry. These ndings point to the need to better understand the mechanism and determinants of technology choice by rms. 107 4.6 Tables and Figures Figure 4.1: The e¤ect of trade openness on technology adoption and productivity. Figure 4.1a. 108 Figure 4.1b. Figure 4.1c. 109 Figure 4.2: Distribution of capital intensities across French rms (1997-2005). 110 T ab le 4. 1: D es cr ip ti ve St at is ti cs . E X P O R T E R S N O N E X P O R T E R S A L L F IR M S M ea n M ed ia n SD M ea n M ed ia n SD M ea n M ed ia n SD F re qu en cy Sa le s 17 74 5 19 09 27 69 53 55 10 39 1 29 49 57 96 75 61 5 28 90 11 68 79 12 E m pl oy ee s 80 16 10 47 20 5 48 6 41 7 73 1 67 30 10 M at er ia ls 83 15 61 9 16 26 56 33 08 10 7 22 75 36 50 59 17 8 20 71 81 66 23 78 F ix ed C ap it al 54 41 19 3 14 18 51 14 97 54 80 31 9 28 30 79 10 52 39 69 29 18 E xp or t In te ns it y 0. 19 6 0. 08 6 0. 24 3 0. 06 7 0 0. 17 0 68 58 05 V al ue ad de d p er w or ke r 52 .9 43 .3 30 9. 8 47 .5 40 .2 24 4. 2 49 .9 41 .5 27 5. 9 41 95 99 Sa le s p er w or ke r 18 5. 7 11 8. 2 20 19 .6 12 5. 1 86 .3 10 21 .3 14 7. 8 96 .4 14 76 .5 59 78 59 C ap it al p er w or ke r 37 .2 12 .7 11 02 .6 28 .8 11 .1 66 6. 5 32 .0 11 .7 85 7. 0 59 27 06 M at er ia ls p er w or ke r 87 .3 40 .3 11 01 .6 51 .7 23 .7 51 1. 3 65 .2 28 .9 78 9. 6 58 28 27 In ve st m en t p er w or ke r 8. 1 2. 4 25 9. 9 6. 0 1. 6 12 3. 7 6. 8 1. 9 18 8. 0 45 84 19 L on g- te rm D eb t p er w or ke r 4. 8 0 37 8. 3 4. 2 0 79 .7 4. 4 0 25 1. 1 50 56 94 In te re st on D eb t 0. 02 5 0. 02 0 0. 02 6 0. 03 2 0. 02 5 0. 03 1 0. 02 9 0. 02 2 0. 02 9 43 54 99 A ge 20 .7 15 19 .4 9. 4 6 14 .9 12 .2 8 16 .8 93 45 87 N ot es : A ll st at is ti cs co ve r th e p er io d 19 97 -2 00 5. A ll m on et ar y va lu es ar e in th ou sa nd s of E ur o (y ea r 20 00 pr ic es ). E xp or t in te ns it y is th e sh ar e of ex p or ts in to ta l sa le s of th e r m . In te re st on de bt is th e ra ti o of in te re st pa id to lo ng te rm de bt . Sh ar e of ex p or te rs in th e sa m pl e is 0. 34 1. 111 T ab le 4. 2: O L S es ti m at es of th e pr od uc ti on fu nc ti on fo r ex p or ti ng an d no n- ex p or ti ng r m s. D ep en d en t va ri ab le ln (S al es ) ln (S al es ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) ln (S al es ) ln (S al es ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) 0. 15 56 ** * 0. 11 38 ** * 0. 11 21 ** * 0. 06 23 ** * 0. 06 48 ** * 0. 03 47 ** * 0. 02 79 ** * 0. 02 68 ** * 0. 02 25 ** * 0. 03 20 ** * E x p o rt er (0 .0 02 3) (0 .0 03 2) (0 .0 03 4) (0 .0 03 5) (0 .0 04 1) (0 .0 01 3) (0 .0 02 1) (0 .0 02 1) (0 .0 02 3) (0 .0 03 1) 0. 02 68 ** * 0. 03 92 ** * 0. 02 33 ** * 0. 02 78 ** * 0. 00 16 ** 0. 00 69 ** * 0. 00 47 ** * 0. 00 55 ** C ap it al * E x p o rt er (0 .0 01 1) (0 .0 02 3) (0 .0 02 3) (0 .0 03 1) (0 .0 00 8) (0 .0 01 3) (0 .0 01 3) (0 .0 02 3) -0 .0 21 5* ** -0 .0 63 9* ** -0 .0 18 1* ** -0 .0 10 2* ** -0 .0 16 6* ** -0 .0 05 5 L ab o r* E x p o rt er (0 .0 03 2) (0 .0 04 5) (0 .0 04 5) (0 .0 01 9) (0 .0 02 5) (0 .0 03 8) 0. 05 98 ** * 0. 00 93 ** * M at er ia ls * E x p o rt er (0 .0 03 7) (0 .0 02 2) 0. 10 40 ** * 0. 09 41 ** * 0. 09 06 ** * 0. 09 62 ** * 0. 16 09 ** * 0. 06 41 ** * 0. 06 35 ** * 0. 06 20 ** * 0. 06 27 ** * 0. 10 59 ** * F ix ed C ap it al (0 .0 01 1) (0 .0 01 2) (0 .0 01 2) (0 .0 01 2) (0 .0 02 1) (0 .0 01 0) (0 .0 01 0) (0 .0 01 0) (0 .0 01 0) (0 .0 01 9) 0. 47 72 ** * 0. 47 55 ** * 0. 48 21 ** * 0. 49 60 ** * 0. 78 14 ** * 0. 26 14 ** * 0. 26 14 ** * 0. 26 42 ** * 0. 26 61 ** * 0. 54 97 ** * L ab o r (0 .0 02 3) (0 .0 02 3) (0 .0 02 4) (0 .0 02 5) (0 .0 02 9) (0 .0 02 3) (0 .0 02 3) (0 .0 02 3) (0 .0 02 3) (0 .0 04 8) 0. 38 32 ** * 0. 38 29 ** * 0. 38 30 ** * 0. 36 25 ** * 0. 41 32 ** * 0. 41 32 ** * 0. 41 31 ** * 0. 41 00 ** * M at er ia ls (0 .0 02 0) (0 .0 02 0) (0 .0 02 0) (0 .0 02 2) (0 .0 03 0) (0 .0 03 0) (0 .0 03 0) (0 .0 03 0) 0. 00 29 -0 .0 25 9* ** -0 .0 24 4* ** -0 .0 38 0* ** -0 .3 09 4* ** -0 .2 79 4* ** -0 .2 81 0* ** -0 .2 80 6* ** -0 .2 83 4* ** -0 .6 00 9* ** C o n st an t (0 .0 02 4) (0 .0 02 7) (0 .0 02 7) (0 .0 02 8) (0 .0 03 3) (0 .0 05 1) (0 .0 05 1) (0 .0 05 1) (0 .0 05 1) (0 .0 05 0) F ir m -l ev el x ed e¤ ec ts N O N O N O N O N O Y E S Y E S Y E S Y E S Y E S R 2 0. 93 8 0. 93 8 0. 93 8 0. 93 9 0. 90 0 0. 65 3 0. 65 3 0. 65 3 0. 65 3 0. 34 5 N 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 43 4, 75 0 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 43 4, 75 0 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . A ll sp ec i ca ti on s in cl ud e as ad di ti on al co nt ro ls : fo ur di gi t in du st ry du m m ie s in te ra ct ed w it h a fu ll se t of ye ar du m m ie s, ag e, re gi on du m m ie s, an d le ga l st at us du m m ie s. C ol um ns (6 )- (1 0) in cl ud e r m -l ev el x ed e¤ ec ts . St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. 112 T ab le 4. 3: O L S es ti m at es of th e tr an sl og pr od uc ti on fu nc ti on fo r ex p or ti ng an d no n- ex p or ti ng r m s. D ep en d en t va ri ab le ln (S al es ) ln (S al es ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) ln (S al es ) ln (S al es ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) 0. 11 92 ** * 0. 06 91 ** * 0. 06 25 ** * 0. 04 28 ** * 0. 07 91 ** * 0. 02 22 ** * 0. 02 12 ** * 0. 02 15 ** * 0. 01 37 ** * 0. 03 39 ** * E x p o rt er (0 .0 01 9) (0 .0 03 0) (0 .0 03 1) (0 .0 03 1) (0 .0 03 7) (0 .0 01 2) (0 .0 01 9) (0 .0 01 9) (0 .0 02 0) (0 .0 03 1) 0. 02 21 ** * 0. 03 96 ** * 0. 03 86 ** * 0. 04 55 ** * -0 .0 00 4 0. 00 31 ** * 0. 00 64 ** * 0. 01 28 ** * C ap it al * E x p o rt er (0 .0 01 1) (0 .0 01 9) (0 .0 02 1) (0 .0 02 7) (0 .0 00 7) (0 .0 01 1) (0 .0 01 2) (0 .0 02 2) -0 .0 31 3* ** -0 .0 34 1* ** -0 .0 56 3* ** -0 .0 06 7* ** 0. 00 30 -0 .0 19 8* ** L ab o r* E x p o rt er (0 .0 02 6) (0 .0 03 0) (0 .0 04 0) (0 .0 01 7) (0 .0 02 0) (0 .0 03 5) 0. 00 42 -0 .0 14 2* ** M at er ia ls * E x p o rt er (0 .0 02 7) (0 .0 01 7) 0. 07 22 ** * 0. 06 20 ** * 0. 05 39 ** * 0. 05 44 ** * 0. 17 78 ** * 0. 05 07 ** * 0. 05 09 ** * 0. 04 93 ** * 0. 04 76 ** * 0. 11 94 ** * F ix ed C ap it al (0 .0 02 0) (0 .0 02 0) (0 .0 02 0) (0 .0 02 0) (0 .0 03 9) (0 .0 01 8) (0 .0 01 9) (0 .0 01 9) (0 .0 01 9) (0 .0 03 1) 0. 40 66 ** * 0. 40 43 ** * 0. 41 91 ** * 0. 42 06 ** * 0. 78 41 ** * 0. 24 00 ** * 0. 24 01 ** * 0. 24 30 ** * 0. 23 83 ** * 0. 59 76 ** * L ab o r (0 .0 02 9) (0 .0 02 9) (0 .0 03 2) (0 .0 03 4) (0 .0 04 0) (0 .0 03 5) (0 .0 03 5) (0 .0 03 6) (0 .0 03 6) (0 .0 06 0) 0. 50 89 ** * 0. 50 57 ** * 0. 50 55 ** * 0. 50 31 ** * 0. 56 75 ** * 0. 56 75 ** * 0. 56 75 ** * 0. 57 53 ** * M at er ia ls (0 .0 02 5) (0 .0 02 5) (0 .0 02 5) (0 .0 03 0) (0 .0 04 3) (0 .0 04 3) (0 .0 04 3) (0 .0 04 3) F ir m -l ev el x ed e¤ ec ts N O N O N O N O N O Y E S Y E S Y E S Y E S Y E S R 2 0. 95 3 0. 95 3 0. 95 3 0. 95 3 0. 90 4 0. 70 1 0. 70 1 0. 70 1 0. 70 1 0. 35 3 N 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 43 4, 75 0 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 43 4, 75 0 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . A ll sp ec i ca ti on s in cl ud e as ad di ti on al co nt ro ls : fo ur di gi t in du st ry du m m ie s in te ra ct ed w it h a fu ll se t of ye ar du m m ie s, ag e, re gi on du m m ie s, an d le ga l st at us du m m ie s. C ol um ns (6 )- (1 0) in cl ud e r m -l ev el x ed e¤ ec ts . St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. 113 T ab le 4. 4: R ob us tn es s of pr od uc ti on fu nc ti on es ti m at es . D ep en d en t va ri ab le ln (S al es ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) ln (S al es ) ln (S al es ) ln (V al u e A d d ed ) ln (S al es ) ln (S al es ) ln (S al es ) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) E st im at io n m et h o d O P O P O P O P O P S Y S -G M M S Y S -G M M O L S F E O P 0. 12 47 ** * 0. 08 86 ** * 0. 07 15 ** * 0. 05 02 ** * 0. 06 31 ** * 0. 03 81 ** * 0. 05 90 ** * 0. 05 64 ** * 0. 03 14 ** * 0. 04 94 ** * E x p o rt er (0 .0 02 4) (0 .0 03 7) (0 .0 04 0) (0 .0 03 5) (0 .0 04 5) (0 .0 06 8) (0 .0 27 2) (0 .0 03 5) (0 .0 02 3) (0 .0 04 3) 0. 03 52 ** * 0. 01 73 ** * 0. 01 81 ** * 0. 01 77 ** * 0. 10 79 ** * 0. 07 40 ** 0. 02 03 ** * 0. 00 42 ** * 0. 01 71 ** * C ap it al * E x p o rt er (0 .0 00 5) (0 .0 00 5) (0 .0 00 7) (0 .0 01 2) (0 .0 20 1) (0 .0 41 8) (0 .0 02 3) (0 .0 01 4) (0 .0 00 6) -0 .0 30 7* ** -0 .0 77 3* ** -0 .0 19 3* ** -0 .0 82 9* ** -0 .1 63 9* ** -0 .1 00 5* * -0 .0 57 0* ** -0 .0 15 7* ** -0 .0 68 1* ** L ab o r* E x p o rt er (0 .0 03 5) (0 .0 05 1) (0 .0 04 8) (0 .0 06 7) (0 .0 55 1) (0 .0 46 2) (0 .0 04 5) (0 .0 02 5) (0 .0 05 8) 0. 06 54 ** * 0. 06 23 ** * -0 .0 71 9* * 0. 05 89 ** * 0. 00 92 ** * 0. 06 07 ** * M at er ia ls * E x p o rt er (0 .0 04 1) (0 .0 05 4) (0 .0 24 4) (0 .0 03 7) (0 .0 02 2) (0 .0 04 7) 0. 14 96 ** * 0. 14 02 ** * 0. 14 11 ** * 0. 21 89 ** * 0. 14 81 ** * 0. 31 30 ** * 0. 13 70 ** * 0. 09 36 ** * 0. 06 06 ** * 0. 14 11 ** * F ix ed C ap it al (0 .0 01 0) (0 .0 01 1) (0 .0 01 1) (0 .0 01 5) (0 .0 01 0) (0 .0 18 6) (0 .0 22 3) (0 .0 02 0) (0 .0 01 7) (0 .0 01 1) 0. 47 45 ** * 0. 48 48 ** * 0. 50 16 ** * 0. 74 73 ** * 0. 49 09 ** * 0. 32 81 ** * 0. 85 53 ** * 0. 51 50 ** * 0. 27 12 ** * 0. 50 15 ** * L ab o r (0 .0 02 8) (0 .0 03 0) (0 .0 03 4) (0 .0 03 5) (0 .0 03 1) (0 .0 30 5) (0 .0 48 2) (0 .0 03 6) (0 .0 03 5) (0 .0 03 3) 0. 36 89 ** * 0. 36 92 ** * 0. 34 50 ** * 0. 35 99 ** * 0. 10 58 ** * 0. 36 32 ** * 0. 41 00 ** * 0. 34 45 ** * M at er ia ls (0 .0 02 2) (0 .0 02 2) (0 .0 02 6) (0 .0 02 3) (0 .0 11 6) (0 .0 02 2) (0 .0 03 0) (0 .0 02 6) 0. 01 76 ** 0. 00 86 [S h ar e o f ex p o rt er s] * ca p it al (0 .0 07 9) (0 .0 06 3) -0 .0 76 9* ** -0 .0 20 4* [S h ar e o f ex p o rt er s] * la b o r (0 .0 12 6) (0 .0 11 8) 0. 06 49 ** * E x p o rt sh ar e (0 .0 10 9) 0. 01 18 ** * C ap it al * [E x p o rt sh ar e] (0 .0 03 6) -0 .0 39 7* ** L ab o r* [E x p o rt sh ar e] (0 .0 14 3) 0. 02 04 ** M at er ia ls * [E x p o rt sh ar e] (0 .0 09 2) N 35 1, 66 2 35 1, 66 2 35 1, 66 2 24 8, 47 2 35 1, 66 2 52 8, 03 2 34 3, 31 1 64 1, 30 2 64 1, 30 2 35 1, 66 2 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . A ll sp ec i ca ti on s in cl ud e as ad di ti on al co nt ro ls : fo ur di gi t in du st ry du m m ie s in te ra ct ed w it h a fu ll se t of ye ar du m m ie s, ag e, re gi on du m m ie s, an d le ga l st at us du m m ie s. St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. "S ha re of ex p or te rs " is th e fr ac ti on of r m s w it hi n an in du st ry (N A IC S4 ) th at ex p or ts . C ol um ns (1 ) to (5 ) an d (1 0) ar e es ti m at ed us in g th e m et ho d de ve lo p ed by O lle y an d P ak es (1 99 6) . In co lu m n (5 ) ex p or te rs ar e de n ed as r m s th at ex p or t at le as t 10 % of th ei r ou tp ut . C ol um ns (6 ) an d (7 ) w er e es ti m at ed us in g Sy st em -G M M (B lu nd el an d B on d, 19 98 ) an d us e al l le ve l fa ct or in pu ts la gg ed 3 ye ar s as in st ru m en ts in th e di ¤ er en ce eq ua ti on an d la gg ed di ¤ er en ce s in th e le ve l eq ua ti on . T he L M te st fo r th ir d or de r se ri al co rr el at io n in di ¤ er en ce d re si du al s is re je ct ed in b ot h sp ec i ca ti on s at 1% cr it ic al le ve l. T he p- va lu e fo r co m m on fa ct or re st ri ct io n, im p os ed by th e m in im um di st an ce , is 0. 25 9 in sp ec i ca ti on (6 ) an d 0. 00 0 in sp ec i ca ti on (7 ). Sa rg an -H an se n te st fo r ov er id en ti c at io n is al w ay s re je ct ed at 1% co n de nc e le ve l. 114 T ab le 4. 5: E st im at es of th e pr od uc ti on fu nc ti on fo r te n la rg es t m an uf ac tu ri ng in du st ri es . In d u st ry d es cr ip ti o n N on -M et al li c P ap er P ri m ar y P la st ic an d M ac h in er y F ab ri ca te d C om p u te r an d F oo d C h em ic al T ra n sp or ta ti on M in er al P ro d . M et al R u b b er M et al P ro d . E le ct ro n ic s P ro d u ct s E qu ip m en t N A IC S 3 in d u st ry co d e 32 7 32 2 33 1 32 6 33 3 33 2 33 4 31 1 32 5 33 6 0. 11 10 ** * 0. 03 34 * 0. 01 88 0. 02 46 0. 05 90 ** * 0. 05 94 ** * 0. 07 81 ** * 0. 06 63 ** * -0 .1 15 0* * 0. 01 65 E x p o rt er (0 .0 24 3) (0 .0 19 2) (0 .0 22 7) (0 .0 22 4) (0 .0 13 3) (0 .0 06 8) (0 .0 24 5) (0 .0 08 2) (0 .0 34 1) (0 .0 44 8) 0. 05 60 ** * 0. 02 68 ** * 0. 04 53 ** * 0. 00 11 -0 .0 06 1 0. 00 62 ** 0. 00 24 0. 01 67 ** * 0. 00 13 0. 02 67 ** * C ap it al * E x p o rt er (0 .0 03 5) (0 .0 02 8) (0 .0 05 0) (0 .0 02 7) (0 .0 03 8) (0 .0 01 9) (0 .0 01 8) (0 .0 05 0) (0 .0 03 1) (0 .0 04 1) -0 .0 70 9* ** -0 .0 28 0 -0 .0 46 3 -0 .0 01 0 -0 .0 03 7 -0 .0 10 6 0. 00 20 -0 .0 11 9 -0 .0 48 7* -0 .0 33 0 L ab o r* E x p o rt er (0 .0 20 8) (0 .0 22 3) (0 .0 09 1) (0 .0 20 7) (0 .0 13 0) (0 .0 07 8) (0 .0 13 2) (0 .0 08 4) (0 .0 28 8) (0 .0 34 3) 0. 11 64 ** * 0. 09 46 ** * 0. 10 71 ** * 0. 12 25 ** * 0. 14 33 ** * 0. 14 45 ** * 0. 10 10 ** * 0. 10 24 ** * 0. 17 00 ** * 0. 12 22 ** * F ix ed C ap it al (0 .0 08 4) (0 .0 07 6) (0 .0 01 3) (0 .0 06 2) (0 .0 45 3) (0 .0 03 1) (0 .0 06 0) (0 .0 15 1) (0 .0 10 5) (0 .0 14 1) 0. 49 53 ** * 0. 41 25 ** * 0. 33 48 ** * 0. 38 47 ** * 0. 49 98 ** * 0. 56 46 ** * 0. 49 14 ** * 0. 35 58 ** * 0. 35 89 ** * 0. 56 67 ** * L ab o r (0 .0 22 0) (0 .0 23 7) (0 .0 08 0) (0 .0 22 4) (0 .0 11 3) (0 .0 04 7) (0 .0 12 4) (0 .0 69 9) (0 .0 28 5) (0 .0 31 2) 0. 39 46 ** * 0. 46 20 ** * 0. 52 51 ** * 0. 44 75 ** * 0. 36 19 ** * 0. 26 71 ** * 0. 37 34 ** * 0. 54 44 ** * 0. 48 83 ** * 0. 38 40 ** * M at er ia ls (0 .0 13 2) (0 .0 16 5) (0 .0 09 1) (0 .0 13 2) (0 .0 08 9) (0 .0 04 6) (0 .0 08 2) (0 .0 25 3) (0 .0 14 8) (0 .0 18 4) N 8, 41 1 5, 08 9 2, 19 2 7, 41 0 19 ,4 82 39 ,7 77 11 ,3 34 61 ,2 12 6, 36 9 3, 35 4 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . A ll sp ec i ca ti on s w er e es ti m at ed w it h O lle y an d P ak es (1 99 6) ap pr oa ch an d in cl ud e in du st ry -y ea r x ed e¤ ec ts , ag e, re gi on an d le ga l st at us du m m ie s. St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. 115 T ab le 4. 6: E st im at io n of th e pr od uc ti on fu nc ti on fo r cu rr en t an d fu tu re ex p or te rs . D ep en d en t va ri ab le : ln (S al es ) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) E st im at io n m et h od : O L S F E T ra n sl og T ra n sl og O P O L S F E T ra n sl og T ra n sl og O P E x p o rt er 0. 06 72 ** * 0. 02 53 ** * 0. 04 13 ** * 0. 01 39 ** * 0. 06 94 ** * 0. 06 67 ** * 0. 02 21 ** * 0. 04 25 ** * 0. 01 46 ** * 0. 06 62 ** * (0 .0 03 5) (0 .0 02 4) (0 .0 03 2) (0 .0 02 1) (0 .0 03 3) (0 .0 03 5) (0 .0 02 3) (0 .0 03 2) (0 .0 02 3) (0 .0 04 0) F u tu re E x p o rt er 0. 04 24 ** * 0. 00 85 ** 0. 02 34 ** * -0 .0 00 6 0. 04 31 ** * 0. 04 31 ** * -0 .0 00 9 0. 06 41 ** * -0 .0 01 0 0. 04 15 ** * (0 .0 05 8) (0 .0 03 1) (0 .0 04 9) (0 .0 02 8) (0 .0 06 5) (0 .0 06 6) (0 .0 03 6) (0 .0 05 7) (0 .0 03 6) (0 .0 07 2) C ap it al * E x p o rt er 0. 02 39 ** * 0. 00 46 ** * 0. 03 94 ** * 0. 00 65 ** * 0. 01 76 ** * 0. 02 36 ** * 0. 00 43 ** * 0. 03 90 ** * 0. 00 43 ** * 0. 01 74 ** * (0 .0 02 4) (0 .0 01 4) (0 .0 02 1) (0 .0 01 2) (0 .0 00 5) (0 .0 02 4) (0 .0 01 3) (0 .0 02 1) (0 .0 01 3) (0 .0 00 5) C ap it al * (F u tu re E x p o rt er ) 0. 01 89 ** * -0 .0 00 7 0. 02 14 ** * -0 .0 00 1 0. 00 62 ** * 0. 01 36 ** * -0 .0 06 9* * 0. 01 40 ** * -0 .0 06 9* * 0. 00 45 (0 .0 03 7) (0 .0 02 2) (0 .0 03 1) (0 .0 02 1) (0 .0 01 3) (0 .0 04 6) (0 .0 02 9) (0 .0 04 2) (0 .0 02 9) (0 .0 02 7) L ab o r* E x p o rt er -0 .0 63 2* ** -0 .0 15 9* ** -0 .0 33 2* ** 0. 00 43 * -0 .0 77 2* ** -0 .0 63 8* ** -0 .0 15 7* ** -0 .0 33 5* ** -0 .0 15 7* ** -0 .0 77 1* ** (0 .0 04 5) (0 .0 02 5) (0 .0 03 1) (0 .0 02 0) (0 .0 05 1) (0 .0 04 5) (0 .0 02 5) (0 .0 03 1) (0 .0 02 5) (0 .0 05 1) L ab o r* (F u tu re E x p o rt er ) 0. 00 32 0. 00 62 0. 00 93 0. 01 10 ** -0 .0 03 4 -0 .0 07 0 0. 01 35 ** * 0. 01 42 ** 0. 01 35 ** * 0. 00 16 (0 .0 08 0) (0 .0 03 8) (0 .0 05 3) (0 .0 03 4) (0 .0 09 7) (0 .0 08 8) (0 .0 04 1) (0 .0 05 5) (0 .0 04 1) (0 .0 11 2) M at er ia ls * E x p o rt er 0. 06 03 ** * 0. 00 88 ** * 0. 00 40 -0 .0 15 7* ** 0. 06 67 ** * 0. 06 05 ** * 0. 00 94 ** * 0. 00 42 0. 00 94 ** * 0. 06 60 ** * (0 .0 03 8) (0 .0 02 3) (0 .0 02 8) (0 .0 01 8) (0 .0 04 2) (0 .0 03 7) (0 .0 02 2) (0 .0 02 7) (0 .0 02 2) (0 .0 04 1) M at er ia ls * (F u tu re E x p o rt er ) 0. 00 38 -0 .0 04 2 -0 .0 10 8* -0 .0 11 9* ** 0. 01 58 ** 0. 01 78 ** * 0. 00 11 -0 .0 03 2 0. 00 11 0. 01 06 (0 .0 06 0) (0 .0 03 2) (0 .0 04 6) (0 .0 02 6) (0 .0 06 8) (0 .0 06 9) (0 .0 03 6) (0 .0 05 3) (0 .0 03 6) (0 .0 07 9) F ix ed C ap it al 0. 09 56 ** * 0. 06 27 ** * 0. 05 36 ** * 0. 04 76 ** * 0. 14 03 ** * 0. 09 59 ** * 0. 06 30 ** * 0. 05 41 ** * 0. 06 30 ** * 0. 14 20 ** * (0 .0 01 2) (0 .0 01 0) (0 .0 02 0) (0 .0 01 9) (0 .0 01 0) (0 .0 01 2) (0 .0 01 0) (0 .0 02 0) (0 .0 01 0) (0 .0 01 0) L ab o r 0. 49 53 ** * 0. 26 57 ** * 0. 41 95 ** * 0. 23 74 ** * 0. 50 15 ** * 0. 49 59 ** * 0. 26 56 ** * 0. 41 98 ** * 0. 26 56 ** * 0. 50 13 ** * (0 .0 02 5) (0 .0 02 3) (0 .0 03 5) (0 .0 03 6) (0 .0 03 3) (0 .0 02 5) (0 .0 02 3) (0 .0 03 5) (0 .0 02 3) (0 .0 03 2) M at er ia ls 0. 36 21 ** * 0. 41 03 ** * 0. 50 33 ** * 0. 57 65 ** * 0. 34 37 ** * 0. 36 19 ** * 0. 40 99 ** * 0. 50 30 ** * 0. 40 99 ** * 0. 34 44 ** * (0 .0 02 2) (0 .0 03 0) (0 .0 03 1) (0 .0 04 3) (0 .0 02 6) (0 .0 02 2) (0 .0 03 0) (0 .0 03 1) (0 .0 03 0) (0 .0 02 6) F ir m -l ev el x ed e¤ ec ts N O Y E S N O Y E S N O N O Y E S N O Y E S N O te st e x p o rt er = f u tu re ex p o rt er , p va lu e 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 te st C ap it al * E x p = C ap i- ta l* (F u tu re E x p ), p va lu e 0. 20 9 0. 01 7 0. 00 0 0. 00 2 0. 00 0 0. 04 2 0. 00 0 0. 00 0 0. 00 0 0. 00 0 te st p ro d u ct io n fu n ct io n s fo r cu r- re n t an d fu tu re ex p o rt er s ar e th e sa m e, p va lu e 0. 00 0 0. 00 0 0. 00 0 0. 00 8 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 te st p ro d u ct io n fu n ct io n s fo r fu - tu re ex p o rt er s an d n o n -e x p o rt er s ar e th e sa m e, p va lu e 0. 00 0 0. 16 4 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 18 5 N 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 35 1, 66 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 35 1, 66 2 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . In co lu m ns (1 ) to (5 ) fu tu re ex p or te rs ar e r m s th at do no t ex p or t in a cu rr en t ye ar bu t ex p or t in th e ne xt ye ar . In co lu m ns (6 ) to (1 0) fu tu re ex p or te rs ar e r m s th at do no t ex p or t in th e cu rr en t ye ar bu t b ec om e ex p or te rs in tw o ye ar s. C ol um ns (1 )- (4 ) an d (6 )- (9 ) ar e es ti m at ed us in g O L S w it h co lu m ns (2 ), (4 ), (7 ) an d (9 ) in cl ud in g r m -l ev el x ed e¤ ec ts . C ol um ns (5 ) an d (1 0) ar e es ti m at ed us in g O lle y an d P ak es (1 99 6) . A ll sp ec i ca ti on s in cl ud e as ad di ti on al co nt ro ls : fo ur di gi t in du st ry du m m ie s in te ra ct ed w it h a fu ll se t of ye ar du m m ie s, ag e, re gi on du m m ie s, an d le ga l st at us du m m ie s. St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. 116 T ab le 4. 7: E st im at io n of th e pr od uc ti on fu nc ti on fo r ol d an d ne w ex p or te rs . D ep en d en t va ri ab le : ln (S al es ) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) E st im at io n m et h od : O L S F E T ra n sl og T ra n sl og O P O L S F E T ra n sl og T ra n sl og O P E x p o rt er 0. 06 84 ** * 0. 02 35 ** * 0. 04 49 ** * 0. 01 46 ** * 0. 06 78 ** * 0. 06 52 ** * 0. 02 34 ** * 0. 04 42 ** * 0. 01 67 ** * 0. 06 59 ** * (0 .0 03 8) (0 .0 02 6) (0 .0 03 4) (0 .0 02 3) (0 .0 04 3) (0 .0 03 6) (0 .0 02 3) (0 .0 03 9) (0 .0 02 5) (0 .0 04 1) N ew E x p o rt er 0. 04 47 ** * 0. 01 44 ** * 0. 03 54 ** * 0. 00 40 * 0. 04 27 ** * 0. 05 63 ** * 0. 02 31 ** * 0. 04 14 ** * 0. 01 32 ** * 0. 06 43 ** * (0 .0 04 0) (0 .0 02 4) (0 .0 03 4) (0 .0 02 1) (0 .0 05 0) (0 .0 04 9) (0 .0 02 8) (0 .0 03 2) (0 .0 02 0) (0 .0 05 2) C ap it al * E x p o rt er 0. 02 45 ** * 0. 00 70 ** * 0. 04 29 ** * 0. 00 94 ** * 0. 01 65 ** * 0. 02 36 ** * 0. 00 44 ** * 0. 03 59 ** * 0. 00 85 ** * 0. 01 71 ** * (0 .0 02 7) (0 .0 01 5) (0 .0 02 4) (0 .0 01 3) (0 .0 00 5) (0 .0 02 5) (0 .0 01 4) (0 .0 02 5) (0 .0 01 6) (0 .0 00 5) C ap it al * (N ew E x p o rt er ) 0. 01 87 ** * 0. 00 08 0. 02 46 ** * 0. 00 17 0. 02 06 ** * 0. 02 09 ** * 0. 00 69 ** * 0. 03 89 ** * 0. 00 62 ** * 0. 01 83 ** * (0 .0 02 6) (0 .0 01 6) (0 .0 02 2) (0 .0 01 5) (0 .0 01 0) (0 .0 03 3) (0 .0 01 8) (0 .0 02 2) (0 .0 01 2) (0 .0 01 2) L ab o r* E x p o rt er -0 .0 70 5* ** -0 .0 22 6* ** -0 .0 39 1* ** 0. 00 10 -0 .0 87 1* ** -0 .0 65 4* ** -0 .0 16 0* ** -0 .0 36 5* ** -0 .0 06 5* -0 .0 79 1* ** (0 .0 05 1) (0 .0 02 9) (0 .0 03 4) (0 .0 02 3) (0 .0 05 5) (0 .0 04 6) (0 .0 02 5) (0 .0 04 1) (0 .0 02 7) (0 .0 05 3) L ab o r* (N ew E x p o rt er ) -0 .0 41 3* ** -0 .0 07 6* ** -0 .0 20 0* ** 0. 00 44 ** -0 .0 33 7* ** -0 .0 52 4* ** -0 .0 20 5* ** -0 .0 33 8* ** 0. 00 46 * -0 .0 66 3* ** (0 .0 05 0) (0 .0 02 6) (0 .0 03 3) (0 .0 02 2) (0 .0 07 0) (0 .0 06 9) (0 .0 03 3) (0 .0 03 2) (0 .0 02 0) (0 .0 07 5) M at er ia ls * E x p o rt er 0. 06 55 ** * 0. 01 07 ** * 0. 00 32 -0 .0 18 2* ** 0. 07 47 ** * 0. 06 15 ** * 0. 00 97 ** * 0. 00 72 * -0 .0 11 5* ** 0. 06 71 ** * (0 .0 04 3) (0 .0 02 5) (0 .0 03 1) (0 .0 01 9) (0 .0 04 6) (0 .0 03 9) (0 .0 02 2) (0 .0 03 4) (0 .0 02 1) (0 .0 04 3) M at er ia ls * (N ew E x p o rt er ) 0. 03 85 ** * 0. 00 60 ** * 0. 00 64 ** -0 .0 08 6* ** 0. 02 14 ** * 0. 04 75 ** * 0. 00 70 ** * 0. 00 38 -0 .0 14 7* ** 0. 05 51 ** * (0 .0 04 0) (0 .0 02 2) (0 .0 02 9) (0 .0 01 8) (0 .0 05 1) (0 .0 05 3) (0 .0 02 7) (0 .0 02 8) (0 .0 01 7) (0 .0 05 7) F ix ed C ap it al 0. 09 62 ** * 0. 06 24 ** * 0. 05 39 ** * 0. 04 68 ** * 0. 14 23 ** * 0. 09 62 ** * 0. 06 27 ** * 0. 05 44 ** * 0. 04 76 ** * 0. 14 23 ** * (0 .0 01 2) (0 .0 01 0) (0 .0 02 0) (0 .0 01 9) (0 .0 01 0) (0 .0 01 2) (0 .0 01 0) (0 .0 02 0) (0 .0 01 9) (0 .0 01 0) L ab o r 0. 49 59 ** * 0. 26 66 ** * 0. 42 12 ** * 0. 23 86 ** * 0. 50 16 ** * 0. 49 60 ** * 0. 26 61 ** * 0. 42 06 ** * 0. 23 81 ** * 0. 50 16 ** * (0 .0 02 5) (0 .0 02 3) (0 .0 03 5) (0 .0 03 6) (0 .0 03 3) (0 .0 02 5) (0 .0 02 3) (0 .0 03 4) (0 .0 03 6) (0 .0 03 3) M at er ia ls 0. 36 25 ** * 0. 40 96 ** * 0. 50 32 ** * 0. 57 62 ** * 0. 34 49 ** * 0. 36 25 ** * 0. 41 00 ** * 0. 50 32 ** * 0. 57 54 ** * 0. 34 49 ** * (0 .0 02 2) (0 .0 03 0) (0 .0 03 0) (0 .0 04 3) (0 .0 02 6) (0 .0 02 2) (0 .0 03 0) (0 .0 03 0) (0 .0 04 3) (0 .0 02 6) F ir m -l ev el x ed e¤ ec ts N O Y E S N O Y E S N O N O Y E S N O Y E S N O te st e x p o rt er = n ew ex p o rt er , p va lu e 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 02 2 0. 87 3 0. 87 6 0. 05 0 0. 24 4 te st C ap it al * E x p = C ap i- ta l* (N ew E x p ), p va lu e 0. 07 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 43 3 0. 10 4 0. 21 6 0. 08 0 0. 72 4 te st p ro d u ct io n fu n ct io n s fo r o ld an d n ew ex p o rt er s ar e th e sa m e, p va lu e 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 15 0 0. 15 9 0. 14 6 0. 00 9 0. 25 7 te st p ro d u ct io n fu n ct io n s fo r n ew ex p o rt er s an d n o n -e x p o rt er s ar e th e sa m e, p va lu e 0. 00 0 0. 00 0 0. 00 0 0. 00 1 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 0. 00 0 N 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 35 1, 66 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 64 1, 30 2 35 1, 66 2 N ot es : * si gn i ca nt at 10 % , ** si gn i ca nt at 5% , ** * si gn i ca nt at 1% . In co lu m ns (1 ) to (5 ) ne w ex p or te rs ar e r m s th at en te re d ex p or ts m ar ke t la st ye ar . In co lu m ns (6 ) to (1 0) ne w ex p or te rs ar e r m s th at en te re d ex p or ts m ar ke t tw o ye ar s ag o. C ol um ns (1 )- (4 ) an d (6 )- (9 ) ar e es ti m at ed us in g O L S w it h co lu m ns (2 ), (4 ), (7 ) an d (9 ) in cl ud in g r m -l ev el x ed e¤ ec ts . C ol um ns (5 ) an d (1 0) ar e es ti m at ed us in g O lle y an d P ak es (1 99 6) . A ll sp ec i ca ti on s in cl ud e as ad di ti on al co nt ro ls : fo ur di gi t in du st ry du m m ie s in te ra ct ed w it h a fu ll se t of ye ar du m m ie s, ag e, re gi on du m m ie s, an d le ga l st at us du m m ie s. St an da rd er ro rs in br ac ke ts ar e cl us te re d by r m . T im e p er io d co ve re d is 19 97 -2 00 5. 117 Table 4.8: Estimation of the CES production function exporters and non-exporters. Estimated equation: Yit = Ait h KK it + LL it + MM it i All rms Exporters Non-exporters All rms Exporters Non-exporters (1) (2) (3) (4) (5) (6) 0.0078*** 0.0334*** -0.0238*** -0.0244 -0.4652 -0.0453 Constant (0.0016) (0.0024) (0.0021) (0.0000) (0.0000) (0.0000) -0.6705*** -0.6872*** -0.6486*** -0.6705*** -0.6872*** -0.6486*** (0.0022) (0.0035) (0.0027) (0.0022) (0.0035) (0.0027) 0.0959*** 0.1266*** 0.0872*** 0.0980*** 0.1785*** 0.0885*** K (0.0004) (0.0008) (0.0005) (0.0005) (0.0012) (0.0006) 0.3537*** 0.3360*** 0.3623*** 0.3614*** 0.4737*** 0.3675*** L (0.0006) (0.0009) (0.0008) (0.0007) (0.0013) (0.0009) = 1 K L 0.5625*** 0.7574*** 0.5583*** M (0.0009) (0.0025) (0.0012) 0.9942*** 0.9980*** 0.9814*** 0.9942*** 0.9980*** 0.9814*** (0.0003) (0.0005) (0.0004) (0.0003) (0.0005) (0.0004) N 641,302 228,774 412,528 641,302 228,774 412,528 Notes: * signi cant at 10%, ** signi cant at 5%, *** signi cant at 1%. In columns (1) to (3) the constraint K+L+M= 1 is imposed. All speci cations include as additional controls: four digit industry dummies interacted with a full set of year dum- mies, age, region dummies, and legal status dummies. Standard errors in brackets are clustered by rm. Time period covered is 1997-2005. 118 Table 4.9: The e¤ect of trade barriers reduction, steady state comparison. 5% trade costs reduction 23% trade costs reduction Endogenous Exogenous Relative Endogenous Exogenous Relative Panel A (alfa=0.22) technology technology change technology technology change (1) (2) (3) (4) (5) (6) Capital intensity of non-exporters 0.000 0.000 Capital intensity of exporters -0.041 -0.089 Average capital intensity 0.004 0.005 % Number of rms -0.045 -0.049 0.930 -0.108 -0.116 0.931 % Number of exporting rms 0.044 0.049 0.890 0.209 0.242 0.863 % MC of non-exporters -0.014 -0.012 1.120 -0.032 -0.028 1.129 % MC of exporters 0.021 0.023 0.902 0.094 0.101 0.930 % Average productivity 0.019 0.017 1.083 0.037 0.034 1.102 Export intensity 0.024 0.021 1.124 0.091 0.079 1.152 % Welfare per worker 0.017 0.015 1.091 0.034 0.030 1.111 Panel B (alfa=0.2) Capital intensity of non-exporters -0.003 -0.005 Capital intensity of exporters -0.017 -0.046 Average capital intensity 0.005 0.009 % Number of rms -0.046 -0.050 0.932 -0.110 -0.114 0.960 % Number of exporting rms 0.043 0.048 0.893 0.204 0.235 0.869 % MC of non-exporters -0.014 -0.013 1.106 -0.032 -0.029 1.112 % MC of exporters 0.024 0.027 0.887 0.099 0.108 0.916 % Average productivity 0.019 0.017 1.101 0.038 0.034 1.138 Export intensity 0.025 0.022 1.130 0.093 0.079 1.180 % Welfare per worker 0.017 0.016 1.112 0.035 0.030 1.147 Notes: Columns (1) to (3) show the e¤ect of tari¤ reduction from 8% to 3%, Columns (4) to (6) show the e¤ect of the reduction in variable trade costs from 53% to 30%. Columns (1) and (4) show the e¤ect of trade liberalization implied by the model with endogenous technology choice, while columns (2) and (5) show the e¤ect implied by the Melitz model with exogenous production technology. In Columns (3) and (6) are the ratios of the change in the variable for the model with endogenous technology choice relative to the model with exogenous technology. 119 References  Alessandria, G., and H. Choi (2007): Do Sunk Costs of Exporting Matter for Net Export Dynamics?,The Quarterly Journal of Economics, 122(1), 289336.  Alvarez, R., and R. Lopez (2005): Exporting and Performance: Evidence From Chilean Plants,Canadian Journal of Economics, 38(4), 13841400.  Andrews, D., Stock, J., (2005) Inference with Weak Instruments,Cowles Foundation Dis- cussion Paper, No. 1530.  Andrews, D. W., and J. H. Stock (2005): Inference with Weak Instruments,NBER Technical Working Papers 0313.  Angrist, J. D. (1991): Grouped-Data Estimation and Testing in Simple Labor-Supply Mod- els,Journal of Econometrics, 47(2-3), 243266.  Angrist, J. D., G. W. Imbens, and A. Krueger (1999): Jackknife Instrumental Variables Estimation,Journal of Applied Econometrics, 14(1), 5767.  Atkeson, A., and A. Burstein (2007): Innovation, Firm Dynamics, and International Trade, NBER Working Papers 13326.  Bagwell, K., and R.W. Staiger (1997): Multilateral Tari¤ Cooperation During the Formation of Free Trade Areas,International Economic Review, 38(2), 291319.  Baldwin, J., and W. Gu (2003): Export-Market Participation and Productivity Performance in Canadian Manufacturing,Canadian Journal of Economics, 36(3), 634657.  Bekker, P. A. (1994): Alternative Approximations to the Distributions of Instrumental Vari- able Estimators,Econometrica, 62(3), 657681.  Berkowitz, J., and M. J. White (2004): Bankruptcy and Small FirmsAccess to Credit, RAND Journal of Economics, 35(1), 6984.  Bernard, A., J. Eaton, B. Jensen, and S. Kortum (2003): Plants and Productivity in Inter- national Trade,American Economic Review, 93(4), 12681290.  Bernard, A., and B. Jensen (1999): Exceptional Exporter Performance: Cause, E¤ect, or Both?,Journal of International Economics, 47(1), 125.  Blundell, R., and S. Bond (1998): Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,Journal of Econometrics, 87(1), 115143.  Bohara, A. K., K. Gawande, and P. Sanguinetti (2004): Trade Diversion and Declining Tari¤s: Evidence from Mercosur,Journal of International Economics, 64(1), 6588.  Bombardini, M. (2008): Firm Heterogeneity and Lobby Participation,Forthcoming in the Journal of International Economics.  Bond, E. (1983): Trade in Used Equipment with Heterogeneous Firms,Journal of Political Economy, 91(4), 688705. 120  Bond, E. W., R. G. Riezman, and C. Syropoulos (2004): A Strategic and Welfare Theoretic Analysis of Free Trade Areas,Journal of International Economics, 64(1), 127.  Bound, J., D. Jaeger, and R. Baker (1995): Problems with Instrumental Variables Estima- tion When the Correlation Between Instruments and the Endogenous Explanatory Variable Is Weak,Journal of the American Statistical Association, 90(430), 443450.  Broda, C., and D. E. Weinstein (2006): Globalization and the Gains from Variety, The Quarterly Journal of Economics, 121(2), 541585.  Clerides, S. K., S. Lach, and J. R. Tybout (1998): Is Learning By Exporting Important? Micro-Dynamic Evidence From Colombia, Mexico, And Morocco,The Quarterly Journal of Economics, 113(3), 903947.  Coe, D. T., and E. Helpman (1995): International R&D spillovers, European Economic Review, 39(5), 859887.  Davidson, R., and J. MacKinnon (1993): Estimation and Inference in Eeconometrics. New York, Oxford University Press.  Deaton, A. (1985): Panel Data from Time Series of Cross-Sections,Journal of Econometrics, 30(1-2), 109126.  Easley, D., and M. Ohara (2004): Information and the Cost of Capital,Journal of Finance, 59(4), 15531583.  Eaton, J., and S. Kortum (2001): Technology, Trade, and Growth: A Uni ed Framework, European Economic Review, 45(4-6), 742755.  Feenstra, R. C. (1994): New Product Varieties and the Measurement of International Prices, American Economic Review, 84(1), 157177.  Gawande, K., and U. Bandyopadhyay (2000): Is Protection for Sale? Evidence on the Grossman-Helpman Theory of Endogenous Protection,The Review of Economics and Statis- tics, 82(1), 139152.  Gawande, K., P. Krishna, and M. J. Robbins (2006): Foreign Lobbies and US Trade Policy, The Review of Economics and Statistics, 88(3), 563571.  Ghironi, F., and M. J. Melitz (2005): International Trade and Macroeconomic Dynamics with Heterogeneous Firms,The Quarterly Journal of Economics, 120(3), 865915.  Goldberg, P. K., and G. Maggi (1999): Protection for Sale: An Empirical Investigation, American Economic Review, 89(5), 11351155.  Grossman, G. M., and E. Helpman (1994): Protection for Sale,American Economic Review, 84(4), 833850.  Grossman, G. M., and E. Helpman (1995): "The Politics of Free-Trade Agreements", American Economic Review, 85(4), 667690.  Grossman, G., and E. Helpman (2005): Product Development And International Trade, Journal of Political Economy, 97(6), 126183. 121  Hansen, C., J. Hausman, and W. Newey (2006): Estimation with Many Instrumental Vari- ables,Discussion paper.  Hansen, W., and N. Mitchell (2000): Disaggregating and Explaining Corporate Political Activity: Domestic and Foreign Corporations in National Politics,American Political Science Review, 94(4), 891903.  Helpman, E., M. J. Melitz, and S. R. Yeaple (2004): Export versus FDI with Heterogeneous Firms,American Economic Review, 94(1), 300316.  Kee, H. L., M. Olarreaga, and P. Silva (2003): Market Access for Sale: Latin Americas Lobbying for U.S. Tari¤ Preferences, The World Bank Policy Research Working Paper Series, No. 3198.  Kelejian, H. (1971): Two-Stage Least Squares and Econometric Systems Linear in Parameters but Nonlinear in the Endogenous Variables,Journal of the American Statistical Association, 66(334), 373374.  KPMG Consulting Inc. (2001): Study on Compliance Under the Lobbyists Registration Act, Prepared for O¢ ce of the Ethics Counsellor. Ottawa. September 14, 2001.  Krishna, P. (1998): Regionalism And Multilateralism: A Political Economy Approach,The Quarterly Journal of Economics, 113(1), 227250.  Luttmer, E. G. J. (2007): Selection, Growth, and the Size Distribution of Firms,The Quar- terly Journal of Economics, 122(3), 11031144.  Maggi, G., and A. Rodriguez-Clare (2005): A Political-Economy Theory of Trade Agree- ments,CEPR Discussion Papers 5321.  Melitz, M. J. (2003): The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity,Econometrica, 71(6), 16951725.  Mitchell, N. (1995): The Global Polity: Foreign Firms Political Activity in the United States,Polity, 27(3), 447463.  Olley, G. S., and A. Pakes (1996): The Dynamics of Productivity in the Telecommunications Equipment Industry,Econometrica, 64(6), 126397.  Ornelas, E. (2005a): Endogenous Free Trade Agreements and the Multilateral Trading Sys- tem,Journal of International Economics, 67(2), 471497.  Ornelas, E. (2005b): Rent Destruction and the Political Viability of Free Trade Agreements, The Quarterly Journal of Economics, 120(4), 14751506.  Richardson, M. (1993): Endogenous Protection and Trade Diversion, Journal of Interna- tional Economics, 34(3-4), 309324.  Roll, R. (1981): A Possible Explanation of the Small Firm E¤ect,Journal of Finance, 36(4), 87988.  Rossi-Hansberg, E., and M. L. Wright (2005): Firm Size Dynamics in the Aggregate Econ- omy,NBER Working Papers 11261. 122  Staiger, D., and J. H. Stock (1997): Instrumental Variables Regression with Weak Instru- ments,Econometrica, 65(3), 557586.  Stock, J. H., J. H. Wright, and M. Yogo (2002): A Survey of Weak Instruments and Weak Identi cation in Generalized Method of Moments,Journal of Business & Economic Statistics, 20(4), 518529.  Titman, S., and R. Wessels (1988): The Determinants of Capital Structure Choice,Journal of Finance, 43(1), 119.  Treer, D. (1993): Trade Liberalization and the Theory of Endogenous Protection: An Econo- metric Study of U.S. Import Policy,Journal of Political Economy, 101(1), 138160.  Van Biesebroeck, J. (2005): Exporting Raises Productivity in Sub-Saharan African Manufac- turing Firms,Journal of International Economics, 67(2), 373391.  Yeaple, S. R. (2005): A Simple Model of Firm Heterogeneity, International Trade, and Wages,Journal of International Economics, 65(1), 120. 123 Appendices Appendix A: Derivation of the Equilibrium Trade Policy The equilibrium trade policy is derived by maximizing the sum of the government and lobbying groupswelfare functions (9): = NX i=1 IHi W H i + aW + NX i=1 IPi W P i + NX i=1 IROWi W ROW i WHi = n H i H i + i(TR+ CS) - welfare of the domestic industry i W ji = n j i j i , j = fP;ROWg - welfare of the partner country and ROW industry i from trade W = P i(n H i H i ) + TR+ CS - national welfare Before deriving the optimal trade policy, calculate the responsiveness of equilibrium prices and quantities to the change in the ROW tari¤. An increase in the ROW tari¤will not a¤ect equilibrium prices of domestic and partner country rms directly, but will increase equilibrium quantities qji through an increase in the aggregate price index Pi. Firms from the ROW will have a direct negative e¤ect on equilibrium quantity through an increase in consumer price (6) and an indirect positive e¤ect through an increase in the aggregate industry price index. @qji @ fi = @qji @Pi @Pi @ fi = (i 1) q j i Pi @Pi @ fi for j = H;P (1a) @qROWi @ROWi = @qROWi @pROWi @pROWi @ROWi + @qROWi @Pi @Pi @ROWi = i i i 1 qROWi pROWi + (i 1)q ROW i Pi @Pi @ROWi (2a) For future convenience, de ne the share of country j rms on Canadian market for the product i: sji = njip j ix j i PiXi = njip j ix j i !i (3) = njid j i pji Pi !1 i (3a) The e¤ect of the ROW tari¤ rate on industry price index can be expressed as: @Pi @ROWi = @Pi @pROWi @pROWi @ROWi (4) = nROWi d ROW i Pi pROWi i i i 1 (3a) = i i 1s ROW i Pi pROWi (4a) Now we can calculate the e¤ect of the ROW tari¤ change on welfare terms in the objective function (9). The element Xi @Pi @ fi = i i 1n ROW i x ROW i (5a) is a common factor that I want to isolate in every equation. Tari¤ revenue: TR = X i nPi P i x P i + n ROW i ROW i x ROW i 124 @TR @ROWi = nPi P i @qPi @ROWi + nROWi x ROW i + n ROW i ROW i @qROWi @ROWi = = Xi @Pi @ROWi nPi P i (i 1)qPi PiXi + i 1i + ROWi pROWi i + (i 1)sROWi = = Xi @Pi @ROWi (i 1) P i pPi sPi + i 1 i + ROWi pROWi (i 1)sROWi i (6a) Consumer surplus: U (1) = X0 + X i !i lnXi Indirect utility: V (2) = Y Pi !i +Pi !i ln!iPi = Y +Pi !i(ln!i 1) Pi !i lnPi CS = P i !i(ln!i 1) P i !i lnPi @CS @ROWi = !i Pi @Pi @ROWi = Xi @Pi @ROWi (7a) Pro t functions: nji @ji @ROWi (7;1a) = nji i pji (i 1) q j i Pi @Pi @ROWi = i 1i njip j i q j i PiXi Xi @Pi @ROWi = = i 1i s j iXi @Pi @ROWi ; for j = H;P nji @ROWi @ROWi (7) = nROWi i 1 q ROW i + nROWi p ROW i i @qROWi @ROWi (2a;4a) = = nROWi i qROWi p ROW i Xi h sROWi + (i 1) isROWi i Xi @Pi @ROWi = i 1i sROWi 1 Xi @Pi @ROWi (8a) Substituting (6a), (7a) and (8a) into the derivative of (9) with respect to the ROW tari¤ rate we obtain: @ @ROWi = (a+ ) (i 1) P i pPi sPi + i 1 i 1 + ROW i pROWi (i 1)sROWi i Xi @Pi @ROWi + + (IHi + a) i 1 i sHi + bI P i i 1 i sPi + cI ROW i i 1 i sROWi 1 Xi @Pi @ROWi = 0 Rearranging and isolating the ROW tari¤ rate on the left-hand side we the obtain optimal trade policy: "i ROWi pROWi = (i 1) P i pPi sPi 1 i + IHi + a a+ i 1 i sHi + bIPi a+ i 1 i sPi + cIROWi a+ i 1 i sROWi 1 (45) Where "i is the price elasticity of demand for the ROW imports: "i = @qROWi @pROWi pROWi qROWi = @qROWi @pROWi + @qROWi @Pi @Pi @pROWi pROWi qROWi = i (i 1)sROWi 125 Appendix B: Estimation of the Elasticities of Substitution To simplify notation, time index is omitted in this section. We start with the description of supply- demand system. From equation (3), total demand for the imports of product i imported from country j equals: xji = !id j i pji pji Pi !1 i (3) Since using expenditure shares instead of quantities reduces measurement error, the above equation is transformed into shares: sji = pjix j i PiXi = pjix j i !i = dji pji Pi !1 i (1b) Taking logs and time di¤erencing, we obtain the new demand equation: ln sji = (i 1) lnPi (i 1) ln pji + "ji (2b) The error term in (1b) is likely to be correlated with the market shares due to simultaneity in determination of market shares and product prices. To model this simultaneity, following Feenstra (1994) let the supply equation for variety j of good i to take the form: pji = exp( j i )(x j i ) ei where ei is the inverse supply elasticity and j is a technology parameter. Taking logs and time di¤erencing we obtain a modi ed supply equation: ln pji = ei ln(x j i ) + j i (3b) To derive the supply equation in the form of expenditure shares, substitute the expression for quantity supplied from (1b) into (3b): ln pji = ei ei + 1 ln(!i) + ei ei + 1 ln(sji ) + j i ; j i = 1 ei + 1 ji (4b) To facilitate estimation it is convenient to eliminate product-speci c xed e¤ects from supply and demand equations through taking the di¤erence in (2b) and (4b) with respect to some reference country k: esij = (i 1)epij + e"ijepij = eiei+1esij + eij (5b) Where exij = ( lnxji lnki ) for any variable x. Although in the model I aggregated all non- NAFTA countries into the ROW, in this section I treat the output of each country as a separate product and assumed that the number of varieties equals to one (Canadian product) plus the number of importing countries. With the assumption that the ROW rms are symmetric, treating product of each country as a separate variety does not a¤ect predictions of the model. Therefore, after di¤erencing, the number of cross-sections in equation (5b) equals to the number of countries that import product i into Canada. 126 For consistent estimation of substitution elasticity parameters, the assumption of independent error terms in (5b) is critical, i.e. technological factor is assumed to be independent of the taste for variety parameter: Assumption 1: p lim T!1 1 T P t eije"ij = 0 for 8 product i With this assumption, isolate error terms on the right-hand side of (5b) and multiply them through: Yij = 1X 1 ij + 2X 2 ij + uij (6b) Yij = ep2ij X1ij = es2ij X2ij = epijesij uij = (i 1) 1e"ijeij 1 = ei (ei + 1)(i 1) ; 2 = iei 1 (ei + 1)(i 1) (7b) Several authors51 showed that a consistent estimates of 1 and 2 can be obtained by the IV estimator, where instruments are country indicators. To demonstrate this result, append (6b) across varieties for each product: Yi = 1X 1 i + 2X 2 i + ui (8b) The total number of observations in (8b) is T Ci, where T is the number of periods in the sample and Ci is the number of countries that import product i into Canada. Introduce TCi Ci matrix of instrumental variables where each column j is an indicator variable for the importing country j, such that vector ej is a T -dimensional vector of 1s: Zi = 26664 e1 0 0 0 e2 0 ... ... . . . ... 0 0 eCi 37775 Then, for the xed number of importing countries Z is a set of valid instruments as T !1. First, Assumption 1 guarantees that p lim T!1 1 T P t Z 0 iui = 0 holds for any product i. The product 1 T Z 0 iui is the Ci-dimensional vector with elements 1T eije"ij , each approaching zero under the Assumption 1. Independence of supply and demand error terms implies that instruments Z are uncorrelated with the error in (8b). Feenstra (1994) also showed that p lim T!1 1 T P t Z 0 iXi has a full column rank under a technical restriction that the demand and supply curves should vary across countries, i.e. for each product there should be some variance either in taste or in productivity parameters across varieties. Therefore, the rank condition for identi cation is satis ed. Together with p lim T!1 1 T P t Z 0 iui = 0, this condition guarantees that the IV estimator with instrument matrix Z is consistent. To make the estimator identical to the one used in Feenstra (1994) and Broda and Weinstein (2006), pre-multiply equation (8b) with Z(Z 0 Z) 1Z 0 to obtain: Y i = 1X 1 i + 2X 2 i + ui (9b) 51For example Angrist (1991) and Angrist and Imbens (1999). 127 where upper bars denote time averages. The OLS estimator on (9b) is identical to the IV estimator on (8b), thus, a consistent estimator of the reduced form parameters 1 and 2 can be obtained by averaging equation (7b) over time and estimating it with the weighted OLS for each product i. i = 0@ CiX j=1 tijX 0 ijXij 1A 10@ CiX j=1 tijX 0 ijYij 1A (10b) Weights tij , the number of years country j was exporting product i to Canada, are used to com- pensate for the di¤erent number of observations (number of years) for each importing country. Even though estimator (10b) is consistent as T goes to in nity, several authors point to its bad performance in small samples where Assumption 1 may not hold.52 The data set used in this work covers the period from 1988 to 2004, while the average number of importing countries per product equals 80. The number of cross-sections is, probably, too large relative to the number of years to rely on asymptotic theory. The problem of a small sample can be solved using a correction method similar to the Deaton (1985) error-in-variables estimator. Let Y i and X i to be the unobserved population means, and Y i and Xi to be their sample estimates. Assume the following sample structure: Y i Xi N Y i Xi ; y xy xy The normality assumption is not critical here. Let to be a moment matrix for Xi , XX a sample moment matrix for Xi, and XY a cross moment matrix for Y i and Xi. The following equalities must hold: E( XX) = + E( XY ) = + xy The OLS estimator for can be written as follows: bi = (X0i Xi ) 1(X0i Y i ) = ( XX )( XX xy) = (X 0iXi Ci) 1(X 0iY i Cixy) (11b) The di¤erence between this estimator and WLS estimator (10b) is that the sample counterparts for variances in xy and can be calculated using all T Ci observations, and estimator (11b) is consistent for Ci !1 holding T xed. Finally, the estimator I used is the WLS estimator (10b) with the small sample correction (11b): bi = 0@ CiX j=1 tij X 0 ijXij bij 1A 10@ CiX j=1 tij X 0 ijYij bxy 1A (12b) And bxy and bij are the sample counterparts for xy and : bij = 1tij 1Ptijk=1 Xijk Xij0 Xijk Xijbxy;ij = 1tij 1Ptijk=1 Xijk Xij0 Yijk Y ij 52Staiger and Stock(1997) and Bound, Jaeger and Baker (1995). 128 The matrix of covariates Xi also includes a constant term to address the problem of measurement error in prices. On the left-hand side of (6b) we have second moment for prices, which equals to the variance of true prices plus the variance of the measurement error. Thus, the constant term will absorb the measurement error in prices. Since market shares are not correlated with unit values errors, explanatory variable will not be a¤ected by the measurement error. Therefore, inclusion of a constant term in Xi makes the estimator (12b) robust to the measurement error in prices. Once we obtain the vector of estimates bi, we can derive the structural parameter of interest i from (7b): i = 1 + 2 2 + q 22 + 41 (13b) To avoid the imaginary values, the condition 22 + 41 0 must hold. In fact, this condition was satis ed every single time the estimate for i was statistically signi cant. 129 Appendix C: Proofs Proof of Proposition 1. Without political economy factors, equation (14) implies the following relationship between the external and internal tari¤s: "tROW = 1 + ( 1)tP sP + 1 1 + 1 sH (1A) Totally di¤erentiating (1A) and assuming constant elasticity of import demandh "
The Open Collections website will be unavailable July 27 from 2100-2200 PST ahead of planned usability and performance enhancements on July 28. More information here.
UBC Theses and Dissertations
Essays in international trade, political economy of protection and firm heterogeneity Stoyanov, Andrey 2008
Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 24-ubc_2008_fall_stoyanov_andrey.pdf [ 2.35MB ]
- JSON: 24-1.0066465.json
- JSON-LD: 24-1.0066465-ld.json
- RDF/XML (Pretty): 24-1.0066465-rdf.xml
- RDF/JSON: 24-1.0066465-rdf.json
- Turtle: 24-1.0066465-turtle.txt
- N-Triples: 24-1.0066465-rdf-ntriples.txt
- Original Record: 24-1.0066465-source.json
- Full Text