Specialization Patterns in International Trade

Specialization Patterns in International Trade Walter Steingress∗ PRELIMINARY - Comments are welcome Abstract I document new facts on the pattern of ...
Author: Melanie Atkins
26 downloads 0 Views 1MB Size
Specialization Patterns in International Trade Walter Steingress∗ PRELIMINARY - Comments are welcome

Abstract I document new facts on the pattern of specialization by looking at export and import diversification. As a result of international trade, countries normally specialize in a few sectors/varieties, which tend to get exported, and diversify the importing sectors/varieties. To measure specialization, I compute concentration indexes for the value of exports and imports and decompose the overall concentration into the extensive product margin (number of products traded) and intensive product margin (volume of products traded). Using detailed product-level trade data for 151 countries, I find that exports are more concentrated than imports, specialization occurs mainly on the intensive product margin, and larger economies have more diversified exports and imports because they trade more products. Based on these novel facts, I evaluate the ability of the Eaton-Kortum model, the workhorse model of modern Ricardian trade theory, to account for the observed patterns. The results show that specialization through comparative advantage induced by technology differences explains the qualitative and quantitative facts. Also, I evaluate the role of the key determinants of specialization: the degree of comparative advantage, the elasticity of substitution and geography.

Keywords: Ricardian Trade Theory, Comparative Advantage, Specialization, Import Concentration, Export Concentration

∗ Department of Economics, Université de Montréal, C.P.6128, succ.

(e-mail: [email protected]).

1

Centre-Ville, Montréal (Québec) H3C 3J7, Canada

1

Introduction

The pattern of specialization is at the core of international trade theory. A consequence of international trade is that countries are not required to produce all their consumption goods, instead they can specialize in the production of certain goods in exchange for others. Trade theory offers different explanations of how countries specialize in the number and sales volume of goods. Assessing the empirical relevance of the underlying theory is of vital interest since it not only allows to evaluate the gains from trade through specialization but also informs how trade affects the structure of an economy. My contribution is twofold. First, I uncover new facts on the pattern of specialization by looking at export and import concentration. I decompose the overall level of concentration into a measure for the extensive and intensive product margins and document concentration levels for exports and imports on all margins. The extensive product margin indicates the degree of specialization in the number of goods traded. The concentration index on the intensive margin measures specialization in the volume of goods traded. The second contribution consists of testing the Eaton and Kortum (2002) model’s ability to account for the observed specialization patterns. Specifically, I test the model based on three basic questions about specialization: What explains the level of specialization in exports and imports? What determines the gap between specialization in exports and imports? Does specialization occur on the intensive or extensive product margin? To start with, I establish a new set of qualitative and quantitative facts of observed crosscountry specialization patterns. Based on detailed product-level trade data for 151 countries, the results show that, on average, countries specialize more in exports relative to imports, with Gini coefficients of 0.98 and 0.92 respectively. The decomposition reveals that specialization occurs predominately on the intensive margin. Countries concentrate their exports and imports in the value of few products and, at the same time, trade a fairly wide range of goods. The difference between the concentration levels of exports and imports is explained by the number of products traded. Countries specialize in exporting few goods and diversify on imports by acquiring various products from abroad. Focusing on cross-country differences, I find that larger economies have more diversified imports and exports. This is mostly along the extensive margin, i.e. large economies export and import a wider product range. Having documented the observed specialization pattern, I employ a standard Ricardian trade model developed by Eaton and Kortum (2002) to evaluate its ability to reproduce the stylized facts. My analysis relies on the Alvarez and Lucas (2007) general equilibrium extension of the Eaton and Kortum model. A key implication of this model is that it uncovers how comparative advantage due to technology differences determines specialization endogenously on both the extensive and the intensive product margins. Furthermore, it identifies geography together with the degree of comparative advantage and the elasticity of substitution as the main determinants of specializa-

2

tion. The degree of comparative advantage heightens the sensitivity of concentration to changes in unit costs, thereby dictating specialization on both margins. Trade costs decrease comparative advantage and increase specialization on the extensive and intensive margin. A higher elasticity of substitution provides for better substitution between intermediate goods and consequently increases concentration on the intensive margin. The model characterizes import specialization on all margins. To calibrate the model, I follow Waugh (2010) and use data and the structure of the model to infer trade costs and the elasticities of trade and substitution. Not surprisingly, the simulated results show that the model produces the observed specialization pattern qualitatively with countries being specialized in exports and diversified in imports on all margins. More importantly, the model delivers an excessive degree of specialization on the extensive margin. As a result, the simulated concentration levels are too high and specialization is dominated by the extensive rather than the intensive product margin as observed in the data. Focusing on the variation across countries, the simulated model reproduces the fact that larger economies are more diversified but fails to account for the observed levels of concentration. Countries trade a too small range of products and specialize excessively on the extensive margin in both trade directions. This paper contributes to the international trade literature that analyses the relationship between the pattern of trade and specialization in commodities. Leamer (1984), who tested whether the structure of trade can be explained by the availability of resources, started an empirical literature on specialization relating the pattern of trade to factor endowments motivated by the Hechscker-Ohlin theorem, see, for example, Bowen, Leamer, and Sveikauskas (1987), Trefler (1995) and Schott (2003). On the other hand, MacDougall (1951), Balassa (1963), Golub and Hsieh (2000) and Costinot, Donaldson, and Komunjer (2012) use trade data to test the Ricardian prediction that countries export relatively more of the commodities they are relatively more productive in. Unlike these papers, the analysis in this paper does not intend to explain why countries specialize in a certain commodity, rather it uses the level of concentration in trade data to shed light on the factors that specify specialization in the number and the volume of goods traded. The levels of concentration in each trade direction contain information on the pattern of trade and as such they provide a new quantitative test of the extent of specialization observed in the data. The analysis presented in this paper is also related to a growing literature in quantifying the importance of Ricardian comparative advantage in explaining trade patterns using the EatonKortum framework, see, for example, Chor (2010), Shikher (2011), Levchenko and Zhang (2011) and Costinot, Donaldson, and Komunjer (2012). These papers specify a multi-sector Ricardian model with both inter- and intra-industry trade in order to derive implications on the sectorial level. In contrast, I abstract from intra-industry trade and attach a sectoral interpretation to the continuum of traded goods within the standard Eaton-Kortum framework. Given this notion, the number of traded sectors arises endogenously and is not assumed to be fixed as in the previous

3

papers. While the standard model has been primarily used to explain bilateral trade flows and trade volume, (see, for example, Eaton and Kortum (2002), Alvarez and Lucas (2007) and Waugh (2010)), I focus on the model’s implications on the pattern of trade and analyze how trade induces countries to specialize in narrow sectors. In particular, I characterize the model’s predictions on export and import concentration on the intensive and extensive product margin and highlight the implications on the specialization pattern. A merit of this approach is that it uncovers how preferences, technological differences as well as geography effect specialization. At this point, it is important to note that the Ricardian model shares with other models of international trade, most notably monopolistic competition models based on Krugman (1980) and Armington models like Anderson and van Wincoop (2003), the ability to develop quantitative predictions about specialization patterns on the intensive and extensive margin. However, the underlying mechanisms of generating the specialization pattern differ. In monopolistic competition and Armington models, tradable goods are differentiated by location of production since each country is the sole producer of a good. Thus, countries are completely specialized in disjoint sets of goods. In the Ricardian model of Eaton and Kortum (2002), production competes with imports because countries produce and export the same goods. As such, the Eaton and Kortum model generates incomplete specialization. Finally, my investigation relates to the empirical growth literature analyzing the relationship between economic growth and trade patterns on the intensive and extensive margins, see Hummels and Klenow (2005) and Cadot, Carrèère, and Strauss-Kahn (2011). Like the previous papers, I study cross-country differences by decomposing the trade pattern into extensive and intensive margins. However, my focus is not to analyze cross-country linkages between exports and economic growth by quantifying the contribution of each export margin to explain GDP differences. Instead, I apply the decomposition to both exports and imports and use the resulting empirical evidence to test Ricardian trade theory based on the Eaton-Kortum model. My analysis shows that the Eaton-Kortum model offers a structural framework that can reconcile Hummels and Klenow’s finding that larger economies export more goods and Cadot, Carrèère, and Strauss-Kahn’s result that cross-country concentration differences in exports are predominately driven by the extensive margin. The novel approach of linking cross-country variation of export and import concentration to test the Eaton-Kortum model sheds light on how the interaction between preferences, technology and geography establishes trade patterns on the intensive and extensive margin. As such, the Eaton-Kortum framework can provide theoretical guidance for future work. The rest of the paper is organized as follows. Section 2 describes the data and presents the empirical evidence of import and export concentration. Section 3 lays out the theoretical framework. Section 4 describes the simulation results and the resulting pattern of specialization. Section 5 concludes.

4

2

Empirical evidence and data

The starting point of my analysis is an empirical assessment of the observed specialization patterns in world trade using detailed product level trade data. Before describing the data and the empirical evidence, I illustrate the properties of the concentration measurements used, which form the basis of the qualitative and quantitative tests of the model.

2.1

Concentration measurements

I compute two measures of specialization for product level sales, the Gini coefficient and the Theil index. The Theil index has the advantage of being decomposable into an extensive and intensive product margin measure. For concreteness, I focus on exports - concentration measures for imports are entirely analogous. The two measurements are defined as follows. Let k index a product among the N products in operation in the world economy, let Rk be the corresponding export sales revenue, say, in a given country. The export Gini in this country is defined as : 2 G = ¯ Cov ( R, F ( R)) R

(1)

where Cov is the covariance between export revenues R and the cumulative distribution of export revenues F ( R). R¯ denotes the average export revenue. A Gini coefficient of zero expresses complete diversification across trade revenues, i.e. (1) a country exports all products and (2) the revenues are the same across them. An index of one expresses complete specialization in which case export revenues stem from one product only. The Theil index is a weighted average of the log difference from the mean export revenue ( R¯ ) and defined by the following formula 1 T= N



k∈ N

Rk ln R¯



Rk R¯



(2)

The index takes the value of zero in the case of complete diversification and ln( N ) in the case of complete specialization. Cadot, Carrèère, and Strauss-Kahn (2011) decompose the Theil index into a measure for the intensive and extensive product margin, T = T ext + T int . The extensive Theil index ( T ext ) captures the concentration in the number of products (extensive product margin) whereas the intensive Theil ( T int ) measures the concentration in the sales volume of products (intensive product margin). The intensive Theil index is given by: T int =

1 Nex



k ∈ Nex

Rk R ln( ¯k ) ¯ Rex Rex

(3)

and the extensive Theil index is T

ext

= ln



N Nex



(4)

Nex denotes the number of exported products and R¯ ex represents the mean value of exported products.

5

2.2

Data

To build my empirical evidence, I use the Comtrade data set collected by the United Nations and choose the 6 digit HS 1992 product classification scheme as the preferred level of disaggregation, which results into 5015 products. Import flows of the same 6-digit product from different trading partners refer to different varieties of the same product. The baseline sample covers 151 countries representing all regions and all levels of development between 1992 and 2009 (18 years), including 121 developing countries, defined by the World Bank as countries with 2006 per capita GDP under $16,000 in constant 2005 PPP international dollars. After I take out missing-year data, the sample consists of 2676 observations (country-years). Note the data includes trade within 6 digit product categories. The model I am testing is Ricardian and does not feature trade between varieties of the same product. To establish a mapping between the model and the data, I net out the within product component by considering net trade flows instead of gross trade flows. To measure the importance of trade between products and trade between varieties, I follow Grubel and Lloyd (1975) and calculate the percentage share of trade between products with respect to total trade. I find an average value of 81 percent across countries. The overall share of total net trades flows with respect to total gross trade flows is 65 percent. Both findings suggest that in this sample the majority of trade flows between countries is across products.1 Based on net trade flows at the product level, I calculate first the corresponding concentration indexes on all margins for each year and then take the average over the whole sample period. Since the concentration indexes employed are independent of scale, the calculation on a year-toyear basis avoids the need to deflate the data. Figure 1 plots the mean export against the mean import concentration for each country together with the 45 degree line. In terms of overall concentration, see Figures 1(a) and 1(b), the vast majority of observed levels lie above the 45 degree line highlighting the fact that exports are more concentrated than imports for almost all countries. The notable exceptions are the United States, China, Germany, India, Italy, Japan and Afghanistan. On the intensive product margin, see Figure 1(c), the specialization level of exports is similar to imports with slightly higher levels of concentration for exports. Figure 1(d) plots the results for the extensive product margin. With the exception of China and Germany, who lie below the 45 degree line, countries export fewer products than they import. Table 1 summarizes the sample statistics with the average year-by-year indices over the 2676 country-year pairs. As implied by Figure 1, exports are more concentrated than imports on all margins. In terms of overall concentration, the summary statistics reveal high levels of export and 1 In

the appendix I present an alternative approach to account for observed intra-industry trade in the data. The basic idea is to develop a measurement device that enables the model to characterize trade within and across products. The suggested procedure converts the measurement of product units in the model to product units in the data and allows examining specialization patterns based on gross trade flows. In the rest of the paper, I follow the net trade flow approach and present the estimation and results of the alternative procedure in the appendix.

6

Table 1: Mean concentration indexes over 2676 country-year pairs

Gini

Level of concentration % share of overall concentration

Theil Exports (X)

Theil Imports (M)

Exports

Imports

Extensive Margin

Intensive Margin

Total

Extensive Margin

Intensive Margin

Total

0.98

0.92

2.46

2.56

5.02

1.06

1.91

2.97

49%

51%

32%

68%

import concentration with a Gini coefficient of 0.98 for exports and 0.92 for imports. A possible explanation for the high concentration levels might be that countries trade very few products between each other and hence specialization is driven by the extensive margin. However, the decomposition favors the alternative explanation. Countries export and import a fairly wide range of products but concentrate their trade in the value of few products. The share of the intensive margin with respect to overall concentration is 68 percent for imports and 51 percent for exports. Focusing on the gap between export and import concentration, Table 1 shows that differences between exports and imports are primarily driven by the number of products traded. Countries export few and import a lot of products. In fact, the Theil index of 1.06 on the extensive margin of imports implies that, on average, a country imports a 34.4 percent of all products. On the other side, the extensive Theil index of 2.46 implies that a country exports 8.5 percent of the product range. Turning the attention to cross country differences, the empirical evidence shows that larger economies diversify more than smaller economies. Figure 2 plots the log of the mean levels of concentration as a function of market size including the best linear fit for all margins. Market size is measured by the log of the average relative GDP to the United States (USA = 0) over the periods 1992 to 2009. As Figures 2(a) and 2(b) show, the overall Theil index is decreasing with respect to relative GDP, i.e. smaller economies specialize more. This relationship is equally strong for exports and for imports with an R square of 0.39 and 0.38 respectively. The decomposition reveals that specialization on the intensive margin does not vary with market size for both exports and imports, see Figures 2(e) and 2(f). The main driver of specialization differences across countries is the extensive margin. Particularly robust is the linear relationship on the extensive margin for exports with an R square of 0.80. Bigger economies are more diversified because they export more products. The relationship between market size and specialization on the extensive margin of imports follows a L shape pattern. As the size of an economy increases, countries diversify on imports until reaching a certain market size after which concentration is roughly equal across countries. At this point, the key qualitative and quantitative facts have been established. First, exports are

7

more specialized than imports on all margins. Second, overall concentration is primarily driven by specialization on the intensive rather than the extensive product margin. Third, the target levels of concentration are displayed in Table 1. Fourth, the cross-country patterns imply a negative relationship between market size and specialization caused by the extensive margin. The rest of the paper evaluates the Ricardian model’s ability to account for these stylized facts. I start by presenting the relevant parts of the Alvarez and Lucas (2007) extension of the Eaton-Kortum framework.

3

Model

The Eaton–Kortum model is Ricardian, with a continuum of goods produced under a constantreturns technology. The Alvarez and Lucas (2007) extension builds on the Eaton Kortum framework and introduces a non-tradable sector to establish the general equilibrium. I derive the relevant theoretical predictions on the pattern of trade and evaluate the importance of the key model parameters for specialization of imports and exports. Consider a world economy with I countries, where each country produces tradable intermediate goods as well as non-tradable composite and final goods. Following Alvarez and Lucas (2007), define x = ( x1 , ..., x I ) as a vector of technology draws for any given tradable good and refer to it I . The production of an intermediate good in country i is defined by: as “good x” with x ∈ R+

qi ( xi ) = xi−θ si ( xi ) β qm,i ( xi )1− β . Technology xi differs between goods and is drawn independently from a common exponential distribution with density φ and a country specific technology parameter λi , i.e. xi ∼ exp(1/λi ).

Denote the wage by wi and the price of the intermediate aggregate good by pm,i . The intermediate good sector is perfectly competitive. Intermediate good producers minimize input costs and sell the tradable intermediate good at price β 1− β

pi ( xi ) = Bxiθ wi pm,i . where B = β− β (1 − β)−(1− β) . The continuum of intermediate input goods x enters the production of the composite good qi symmetrically with a constant elasticity of substitution (η > 0) qi =

�ˆ

0



q( x )

1−1/η

φ( x )dx

�η/(1−η )

.

The produced aggregate intermediate good qi can then be allocated costless towards the production of final goods or being used as an input in the production of intermediate goods. Similarly, labor, as the only primary factor input, can be used either to produce intermediate or final goods. Finally, consumers draw their utility linearly from the final good. All markets are perfectly com-

8

petitive. Since these features are not central to the implications I derived in this paper, I omit them. The interested reader is refereed to Alvarez and Lucas (2007) for the full description of the model.

3.1

General equilibrium

Once a country opens to international goods markets, the intermediate goods are the only goods traded. Final goods and labor do not move between countries. Due to trade costs, factor prices will not equalize across countries. The intermediate goods needed to produce the composite good will be acquired from the producer of good x in the country that operates at lowest unit costs. Trading these intermediate goods between countries is costly. We define “Iceberg” transportation costs for good x from country i to country j by κi,j where κi,j < 1 ∀ i �= j and κi,i = 1 ∀i.

As in Alvarez and Lucas (2007), we also consider tariffs. ωi,j is the tariff charged by country i on goods imported from country j. Tariffs distort international trade but do not entail a physical loss of resources. Incorporating the trade costs, composite good producers in country i will buy the intermediate good x from country j that offers the lowest price 

pi ( x ) = B min  j

β 1− β

w j pm,j

κi,j ωi,j



x θj  .

(5)

Equation 5 shows that whether country i specializes in the production of good x depends on the productivity realizations, factor prices and trade costs. The set of goods where country i obtains I . If country i does not offer a good at lowest the minimum price at home is denoted by Bi,i ⊂ R+

costs in the local market, the good is imported. Following Alvarez and Lucas, the resulting price index of tradable goods in country i is 

I



 pm,i = ( AB)  ∑  j =1

β 1− β

w j pm,j

κi,j ωi,j

−1/θ 

−θ

 λj 

(6)

where A = Γ(1 + θ (η − 1)) is the Gamma function evaluated at point (1 + θ (η − 1)). Next, we calculate the expenditure shares for each country i. Let Di,j be the fraction of country i’s per

capita spending pm,i qi on tradables that is spent on goods from country j. Then, we can write total spending of i on goods from j as pm,i qi Di,j =

ˆ

Bi,j

pi ( x )qi ( x )φ( x )dx

where Bi,j defines the set of goods country j attains a minimum in equation 5. Note that Di,j is simply the probability that country j is selling good x in country i at the lowest price and calculated to be

9



Di,j = ( AB)−1/θ 

β 1− β

w j pm,j

pm,i κi,j ωi,j

−1/θ 

λj.

(7)

Equation 7 shows that in this model the sensitivity of trade between countries i and j depends on the level of technology λ, trade costs ω, geographic barriers κ and the technological parameter θ (reflecting the heterogeneity of goods in production) and is independent of the elasticity of substitution η. This result is due to the assumption that η is common across countries and does not distort relative good prices across countries. Note also that by the law of large numbers, the probability that country i imports from country j is identical to the share of goods country i imports from j. In this sense, trade shares respond to costs and geographic barriers at the extensive margin: As a source becomes more expensive or remote it exports/imports a narrower range of goods. It is important to keep in mind that the amount of input industries into the production of the composite good is fixed. Each country uses the whole continuum of intermediate goods to produce composite goods. There are no gains of trade due to an increased number of varieties. Welfare gains are realized through incomplete specialization. Domestic production competes with imports and countries specialize through the reallocation of resources made available by the exit of inefficient domestic producers. Finally, to close the model, we impose that total payments to foreigners (imports) are equal to total receipts from foreigners (exports) for all countries i Li pm,i qi

I

∑ Di,j ωi,j =

j =1

I

∑ L j pm,j q j Dj,i ω j,i

(8)

j =1

The previous equation implies an excess demand system which depends only on wages. Solving this system, describes the equilibrium wages rate for each country together with the corresponding equilibrium prices and quantities. Next, I describe the model’s predictions on export and import concentration on both margins.

3.2

Concentration of exports and imports

In the model, the pattern of trade is established by domestic producers competing with importers for selling intermediate goods in the local market. If foreign producers selling a particular good at a lower price than domestic ones, the good will be imported from the cheapest source. Given the equilibrium price, p( x ), and quantity, q( x ), the total expenditure that country i spends (c.i.f.) on imported good x, Ri,M ( x ), is: Ri,M ( x ) = Li pi ( x )qi ( x )

x∈ / Bi,i

Equivalently, domestic producers export the good to all the foreign markets where they obtain the minimum price. The set of exporting goods is simply a collection of the set of goods country

10

i exports to any destination j, x ∈ ∪ jI�=i B j,i . As a result, (f.o.b.) export revenue sales of good x, Ri,X ( x ), are given by :

Ri,X ( x ) =

I

∑ Lk pk (x)qk (x)κk,i ωk,i

k � =i

x ∈ ∪ jI�=i B j,i

Given the described pattern of trade, the concentration index for imports is identified. To show this, I decompose the overall concentration into a concentration measure for the intensive and extensive product margin. Using equation 3, the Theil index for the concentration of imports on the intensive production margin can be written as: int Ti,M

=

ˆ

x∈ / Bi,i

Ri,M ( x ) ln R¯ i,M



Ri,M ( x ) R¯ i,M



φ( x )dx

In the appendix I show that the distribution of import expenditures follows a Fréchet distribution with shape parameter 1/θ (η − 1) and scale parameter si . Solving the integral, the intensive Theil index of imports for country i becomes:

int Ti,M = ln (Γ(1 + θ (1 − η ))) −

ˆ

0

1

� � ln u(−θ (1−η )) e−u du

(9)

where Γ(.) stands for the Gamma function. Import specialization on the intensive margin is independent of equilibrium prices, trade costs, geography and the level of technology λ. It is solely determined by preferences (i.e the elasticities of substitution) and heterogeneity in production (i.e. the degree of comparative advantage). A higher elasticity of substitution (η ) increases specialization by allowing producers in the aggregate intermediate good sector to better substitute less for high productive goods. Similar, an increase in the degree of comparative advantage (θ ), which corresponds to a higher variance of productivity realizations across products, heightens the degree of specialization. To compute the concentration of imports on the extensive margin, I use the fact that the set of goods produced is disjoint form the set of goods imported. Consequently, we can express the share of goods imported as 1 minus the share of goods produced, (1 − Dii ). The Theil index for the extensive margin of imports is equal to : ext Ti,M

= ln



N Ni,M



= − ln(1 − Dii )

where Di,i = ( AB)−1/θ



wi pm,i

�− β/θ

λi

and depends on the level of technology and equilibrium prices.

11

(10)

To assess the level of specialization in exports, I simulate the model within a discrete product space. I calculate the export concentration index on the intensive margin according to equation 3. The extensive Theil index on the extensive margin is given by the inverse share of the number of goods exported, Ni,X , with respect to the total number of simulated goods, N. ext Ti,X

= ln



N Ni,X



Having outlined the pattern of trade and the corresponding implications on the specialization pattern of exports and imports, the next section discusses the simulation of the model. It contains special cases of equilibria designed to spell out step-by-step the main implications of the model on export and import concentration and in further instance on specialization.

4

Calibration and simulation

To simulate the theoretical model, which assumes an infinite amount of goods, I "discretize" the Fréchet distribution of total factor productivity and calculate the respective trade value for each product x. Concerning the remaining parameters of the model, I use the same values as Alvarez and Lucas (2007). I assume a variance of individual productivities θ = 0.12, an elasticity of substitution of intermediate goods η = 2, an efficient labor share in the production of non-tradable final goods α = 0.75 and an efficient labor share in the tradable goods sector of β = 0.5. I simulate I = 151 countries. In the following subsections, I analyze import and export concentration in special cases of the equilibrium by assuming different trading schemes. Doing so builds intuition of how taste, technology and geography determine specialization. To illustrate the impact of each factor separately, it is instructive to start the analysis by assuming symmetric countries and introduce heterogeneity across countries afterwards. Finally, I show that for a particular configuration of trade costs the Eaton-Kortum model is able to replicate the specialization patterns observed in the data.

4.1

Symmetric countries

All countries are identical with respect to their size Li = L and technology parameter λi = λ. Trade costs are symmetric and set to κi,j = κ ∀ i �= j with κi,i = 1 and ωi,j = 1 ∀i, j. I solve

for the equilibrium wage in each country, all good prices and the value of imports and exports. Due to symmetry, wages and composite good prices equalize across countries. Trade costs distort international trade. In comparison to free trade, firms will not be able to buy good x from the cheapest producer world wide and rely more on home production. The corresponding trade share matrix D is symmetric and the (i, j) element is given by: Di,j =

(κ )1/θ 1 ∀i �= j and Di,i = 1/θ 1 + ( I − 1)(κ ) 1 + ( I − 1)(κ )1/θ

12

where Di,j is the set of goods country j exports to country i and Dii the set of goods country i produces at home. In free trade, κ = 1, each country’s intermediate good producers specialize in a distinct set of goods equal to the relative size of the economy and export all products produced, Dii = Dij = 1/I. The corresponding share of imported products is 1 − Dii = ( I − 1)/I. In this case,

Ricardian specialization forces are strongest and the gap between export and import concentration reaches a maximum.

Concentration on the Extensive Margin The concentration index of imports equals the share of goods country i imports from all countries in the world and is given by: ext Ti,M = − ln((1 − Dii )) = ln(1 + ( I − 1)(κ )1/θ ) − ln(( I − 1)(κ )1/θ )

Concentration at the extensive margin of imports depends negatively on the number of trading partners I − 1 and positively on the degree of comparative advantage θ and geographical barriers κ. In the case of exports, I simulate the model and compute the set of exporting products

numerically. The simulations show that the concentration of exports increases with geographical barriers, the degree of comparative advantage and the number of trading partners. In general, a larger number of trading partners increases competition between production and imports in the domestic market resulting in the production of fewer goods at home and an increase in the number of goods imported. Also, more trading partners increase competition among exporters in foreign markets forcing countries to specialize more on the extensive margin of exports. Impediments to trade, i.e. a reduction in κ, and a higher degree of comparative advantage, θ, reduce import competition and, as a result, fewer goods are exported and imported. Notice that in the special case of free trade all goods produced are exported and concentration of production equals concentration of exports. With trade costs, countries export a subset of produced goods leading to more concentration on the export side relative to production.

Concentration on the Intensive Margin

As noted previously, the distribution of im-

ports follows a Fréchet distribution where the concentration indexes depend only on the elasticity of substitution (η) and the degree of comparative advantage (θ). Consequently, given θ and the concentration of imports observed in the data, I can pin down the elasticity of substitution. Concerning the distribution of export revenues across products, the simulation shows that it depends positively on the elasticity of substitution (η), the degree of comparative advantage (θ) and geographical barriers (κ ). The number of trading partners has non-monotone effects on the concentration of exports at the intensive margin. Few trading partners increase concentration because high revenue generating exports are sold in more markets whereas products with low revenue are exported only to few countries. However, as the number of trading partners increases, the degree of competition in the export markets increases and countries export only high revenue generating products. Thus, after a threshold level, concentration among export revenues reduces with the number of trading partners. In the case of free trade, countries export all their goods to all desti-

13

Table 2: Simulated export and import concentration indexes for benchmark parameters. Gini

Symmetric countries

Theil Exports (X)

Parameters

Exports

Imports

Extensive

Intensive

Margin

Margin

Theil Imports (M) Total

Extensive

Intensive

Margin

Margin

Total

(η = 2,κ = 1)

0.99

0.09

5.01

0.01

5.02

0.01

0.01

0.02

(η = 7.3,κ = 1)

0.99

0.72

5.01

1.91

6.92

0.01

1.91

1.92

(η = 7.3, κ = 0.71)

0.99

0.77

5.04

1.18

6.22

0.10

1.91

2.01

(η = 7.3, κ = 0.71, NT=8)

0.98

0.86

2.47

2.45

4.92

1.07

1.91

2.98

Data

0.98

0.92

2.46

2.56

5.02

1.06

1.91

2.97

nations and, given that preferences are identical, export and import concentration on the intensive margin equalize.2 The results presented in Table 2 show that the free trade calibration of Alvarez and Lucas (2007) is able to replicate the qualitative fact that, overall, exports are more concentrated than imports. While the simulated level of export concentration attains the degree of specialization observed in the data, in the benchmark free trade parametrization countries diversify excessively in imports. Focusing on the decomposition reveals the underlying reason: countries import too many goods and the value of those goods is too evenly distributed. Using the fact that for a given value of θ, import concentration on the intensive margin is solely determined by the elasticity of substitution, I calibrate η = 7.3 to match the level observed in the data3 . As we can see in row 2 of Table 2, this alters the level of import concentration by allowing composite good producers to better substitute between intermediate inputs. To reduce the gap between export and import concentration caused by the extensive product margin, I introduce 40 percent symmetric trade costs to all trading partners, κ = 0.71. Row 3 of Table 2 features the results. Impediments to trade reduce the number of products exported and imported. Consequently, concentration on the extensive margin increases for both exports and imports. Concentration on the intensive margin of exports decreases. Because of trade costs, only efficient producers remain exporters allowing them to distribute their export revenues more evenly across products and trade partners. Although the gap between export and import concentration 2 The

intuition behind this result is that preferences are such that the import expenditure distribution is the same for each trading partner. In the appendix, I show in detail that the expenditure distribution of bilateral trade, Eij (qp), between importer i and exporter j is the same for each source country j, i.e. Eij (qp) = Ei (qp), ∀ j ∈ I. Furthermore, the bilateral import expenditure distribution, Ei (qp), is Fréchet with common shape parameter 1/(θ (η − 1)) and country specific scale parameter si . The shape parameters are identical because preferences are common across countries. Note that the bilateral import expenditure distribution of country i, Ei (qp), equals the export revenue distribution of country j. In free trade with symmetric countries the exporting country ships the exact same goods to all countries. As a consequence, overall export revenue distribution of country j is equal to the import expenditure distribution of each country scaled up by the number of trading partners. Since the concentration indexes are independent of scale, the concentration of exports and imports on the intensive margin equalize. 3 This value is consistent with previous ones found in the literature, see Broda and Weinstein (2006).

14

narrows slightly, the difference is still substantial. The reason is that the degree of competition countries face in export and domestic markets is too high. Introducing infinite trade costs by limiting the amount of trading partners (NT) reduces competition in all markets and improves the fit of the model on all dimensions, see the fourth row of Table 2. Less competition in the domestic market increases the survival rate of domestic industries and reduces the amount of goods imported. Infinite trade costs reduce the number of countries competing in a particular market and increases the probability to export to any of them. As a result, the gap between export and import concentration diminishes. Note that revenues of exporting industries are now geographically more concentrated and hence specialization on the intensive margin of exports intensifies. In sum, with the introduction of symmetric trade costs, the model can replicate the mean levels of concentration observed in the data. The key parameters are the elasticities of substitution η and the trade cost function κ. In particular, by limiting the amount of trading partners, which amounts to introduce zeros in the bilateral trade matrix, we can calibrate the model to explain the pattern of specialization at the mean.

4.2

Asymmetric countries

In this section I analyze the effects of cross-country heterogeneity on specialization. The empirical facts imply a negative relationship between specialization and market size. For this reason I introduce heterogeneity in technology λi and size Li to reflect observed GDP differences in the data. To do so, I consider the model’s free trade equilibrium relationship between wages, size and technology. Under the assumption of free trade, the equilibrium wage is given by : wi =



λi Li

�θ/( β+θ )

(11)

Using equation 11, I back out the level of technology, λi = (wi Li )( β+θ )/θ Li

− β/θ

, as a function of

GDP (wi Li ) and labor endowment ( Li ). To calibrate λ, I use GDP and endowment data from the Penn World table. I proxy labor endowment with data on the country’s level of population. I then normalize the obtained parameters for λi and Li relative to the United States.

Concentration on the Extensive Margin Plugging in the equilibrium wage into equation 7, I get the corresponding trade share matrix D with the (i, j) element given by: D j,i =

wi L i I ∑ k =1 w k L k

∀j

(12)

Equation 12 shows that under the assumption of free trade countryi’s share of number of products exported is equal to its relative level of GDP with respect to world GDP. Hence, larger economies export more products and import less products compared to small economies. This result is at odds with the empirical evidence. In the data, larger economies are more diversified

15

on exports and on imports. In the next section I introduce trade costs and argue that trade costs have to be asymmetric in order to replicate the empirical observations.

Concentration on the Intensive Margin On the intensive margin, under the assumptions of homogenous tastes across countries and free trade, export and import concentration equalize. In this case, export and import concentration on the intensive margin depend only on θ and η and are unaffected by the introduction of heterogeneity in technology and country size.

4.3

Asymmetric trade costs

To reconcile the cross-country concentration differences for imports, I introduce asymmetric trade costs. In particular, I consider trade costs as a function of an exporter (ex j ) or an importer (imi ) fixed effect. The next paragraphs show the different implications of each effect.

Importer fixed effect Under the assumption of an importer fixed effect, country i faces the same cost of importing independent of the origin country j. The trade cost matrix becomes κi,j = imi ∀ j �= i and κi,i = 1 ∀ j = i.

Due to asymmetric trade costs, wages and composite good prices do not equalize. The trade share matrix is asymmetric and given by:

D j,i = ( AB)

−1/θ



β 1− β �−1/θ

wi pm,i

pm,j im j

λi ∀i � = j

and

Di,i = ( AB)

−1/θ



wi pm,i

�− β/θ

λi

Focusing on the expression for the share of goods that country j imports from country i, D j,i , shows that higher import costs (imi ↓) reduce the number of goods country j imports from i. Solving

for the equilibrium and assuming that price differences across countries are approximately equal to the import cost differences, one can show that the corresponding share of goods imported is approximately: � � −1 (1 − Di,i ) ≈ 1 − C1 imi θ wi Li

(13)

where C1 is a constant independent of country i. Equation 13 shows that an importer fixed effect can counterbalance the fact that larger economies import less under the assumption that they face lower costs to import. Lower import costs increase the share of goods imported, (∂(1 −

Dii )/∂imi > 0), and lead to a reduction in the unit cost of production through a lower price

index of tradable goods.

Exporter fixed effect In the case of an exporter fixed effect, each country pays a country specific cost to export, which is independent of the importing country i, κi,j = ex j ∀ j �= i and κi,i = 1 ∀ j = i. The trade share

16

matrix is asymmetric and given by:

D j,i = ( AB)

−1/θ



β 1− β �−1/θ

wi pm,i

pm,j exi

λi ∀i � = j

Di,i = ( AB)−1/θ

and



wi pm,i

�− β/θ

λi

Here the expression for D j,i implies that a higher export cost (exi ↓) reduces the number of goods country i exports to any destination j. Solving for the equilibrium and assuming that compos-

ite good prices across countries are approximately equal, one can show that the share of goods imported is approximately given by: �

−1 θ

(1 − Di,i ) ≈ 1 − C2 exi wi Li



(14)

where C2 represents a constant independent of country i. Equation 14 shows that the share of goods imported is decreasing in the country specific exporting costs, (∂(1 − Dii )/∂exi > 0). Lower

exporting costs lead to higher domestic wages, increasing unit costs of production and result in a larger share of imported goods. Hence, an exporter fixed effect can reconcile the fact that larger economies import more by assuming that (1) larger economies face lower costs to export and (2) the effect of GDP on the export cost offsets the effect of GDP on the share of goods imported,

(∂(1 − Dii )/∂wi Li > 0) ⇒ ((∂(1 − Dii )/∂exi ) (∂exi /∂wi Li ) > ∂(1 − Dii )/∂wi Li ). The main difference between the importer and the exporter fixed effect lies in the implication on the price level of tradable goods. The exporter fixed effect implies a nearly constant price level of tradable goods across countries. As a result, unit cost differences between countries are driven predominantly by wage differences. On the contrary, the importer fixed effect leads to large crosscountry price level differences with smaller economies facing higher a higher price level. In this case, unit cost differences are driven by wage and price level differences. Based on Waugh (2010)’s results that countries have similar price levels of tradable goods, I focus in the rest of my analysis only on the case of a country specific export costs. In sum, the introduction of asymmetric trade costs in form of a country specific cost to export or import allows the model to replicate the import specialization pattern across countries, in particular when larger economies face relative lower costs to either export or import. Waugh (2010) argues that trade costs have to be asymmetric, with poor countries facing higher costs to export relative to rich countries, in order to reconcile bilateral trade volumes and price data. While both our approaches highlight the importance of asymmetric trade costs in explaining trade data, our analysis differs. Waugh uses the Eaton Kortum model to explain bilateral trade volumes and price data whereas I look on the model’s implications on the specialization pattern of exports and imports. In this respect, the results presented in this paper provide further evidence on the importance of asymmetry in trade costs when studying trade volumes and trade patterns across countries.

17

Table 3: Simulated export and import concentration indexes for asymmetric countries. Gini

Theil Exports (X) Extensive

Intensive

Margin

Margin

0.73

5.75

1.91

0.99

0.85

8.08

0.98

0.85

2.29

0.98

0.92

2.46

Parameters

Exports

Imports

(η = 7.3, κ = 1)

0.99

Asymmetric

(η = 7.3, κ = ex)

countries

(η = 7.3, κ = ex, NT=8) Data

Theil Imports (M) Extensive

Intensive

Margin

Margin

7.66

0.007

1.91

1.92

1.68

9.76

1.07

1.91

2.98

2.67

4.96

1.08

1.91

2.99

2.56

5.02

1.06

1.91

2.97

Total

Total

I simulate the model with asymmetric countries, where technology differences are calibrated to reflect GDP differences, for the case of symmetric and asymmetric trade costs. Row 1 of table 3 presents the results in the case of free trade. Note that in relation to the symmetric country case, introducing technology differences increases the mean level of concentration for exports and decreases the level of concentration for imports. The underlying reason is the technology distribution being skewed towards less productive countries and these countries export fewer and import more goods. Beside these changes, the results are similar to the symmetric case. While exports are more concentrated than imports, the simulated level of concentration for exports (imports) is too high (low) compared to the data. The reason is excessive specialization (diversification) on the extensive margin of exports (imports). In terms of cross country differences, calibrated GDP differences in combination with zero or symmetric trade costs lead to the false prediction that larger economies import less goods. To reconcile the empirical evidence, I introduce a country specific cost to export with larger economies facing relative lower export costs. I calculate the implied export cost from equation 14 by replacing the share of goods produced at home by the extensive Theil index of imports observed in the Ext ). Row 2 of table 3 presents the results of the corresponding mean data , Dii = 1 − exp(− TM

concentration levels. While the model matches the cross country concentration pattern, the mean level of export concentration is twice as high as in the data. The reason being excessive competition in the export markets, which results in particular high levels of concentration on the extensive margin of exports. To counter the competition effect, I create trading blocks between countries of size 9 (i.e. number of trading partners NT = 8) by introducing infinite trade costs. In addition, I assume that countries trade within a block with countries whose economic size is similar.4 Trade blocks conditional on economic size improve the fit of the model. The obtained concentration levels match the data on all dimension. Row 3 of Table 3 presents the results. Countries are more concentrated in exports than in imports on all margins, the intensive margin dominates in terms of overall concentration for both exports and imports and the simulated concentration levels are close to the target levels. In terms of the cross country pattern figure 3 plots the simulated (in red) and the empirical (in blue) concentration levels against GDP for both margins. The figure 4 The

precise trade cost configuration is given in the appendix.

18

shows that the exporter fixed effect in combination with technology and endowment differences can replicate the across country evidence on all margins. Bigger economies are more diversified because they import/export more products and concentration patterns on the intensive margin are insensitive to market size. In the previous section I analyzed special cases of the equilibrium to study the different factors that determine specialization in the Eaton Kortum model. The key determinants are the degree of comparative advantage, the elasticity of substitution and asymmetric trade costs. However, I treated trade costs as free parameters and showed that for a particular configuration of trade costs, the model is able to reproduce concentration levels at the mean as well as the cross-country specialization pattern for both exports and imports. In the next section, I estimate trade cost and technology parameters based on bilateral trade shares using the models structure and check whether for given trade shares, the model is able to generate the observed specialization pattern in the data.

5

Estimating trade costs from bilateral trade shares

The starting point of the estimation of technology and trade costs is a structural log-linear “gravity” equation that relates bilateral trade shares with trade costs and structural parameters of the model. To derive the relationship, simply divide each country i’s trade share from country j, see equation 7, by country i’s home trade share. Taking logs yields I − 1 equations for each country i : log



Di,j Di,i



= S j − Si +

1 1 log(κi,j ) + log(ωi,j ) θ θ

(15) − β/θ −(1− β)/θ pm,i λ i ).

in which Si presents the structural parameters and is defined as Si = log(wi

In

order to estimate trade costs κ and technology λ implied by equation 15 I use data on bilateral trade shares across 151 countries. I calculate the corresponding bilateral trade share matrix by the ratio of total gross imports of country i form country j divided by absorption Absi Di,j =

Mi,j Absi

where M( xk )i,j represents total imports of of country i from country j. Absorption is defined as total GDP plus total imports N Mi minus total exports NXi . Note there are only I 2 − I informative

moments and I 2 parameters of interest. Thus, restrictions on the parameter space are necessary. To create them, I follow Eaton and Kortum (2002) and assume the following functional form of trade costs. � � log κi,j = bi,j + dk + ex j + �i,j

Trade costs are a logarithmic function of distance (dk ) a shared border effect between country i and

19

Table 4: Estimation Results Summary Statistics Observations 19833 Geographical barriers Barrier [0,375) [375,740) [750,1500) [1500,3000) [3000,6000) [6000,max) Shared border Tariff

TSS 2,05E+05

SSR 5,23E+04

R2 0.76

Paremeter estimate -5,55 -6,45 -7,49 -8,62 -9,57 -10,37 1,35 1,05

Standard error 0,11 0,07 0,05 0,04 0,03 0,04 0,1 0,18

% effect on cost 94,70% 116,72% 145,59% 181,20% 215,41% 247,26% -14,95% -11,90%

j (bi,j ) and an exporter fixed effect (ex j ). The distance function is represented by a step function divided into 6 intervals. Intervals are in miles: [0, 375); [375, 750); [750, 1,500); [1,500, 3,000); [3,000, 6,000); and [6,000, maximum]. �i,j reflects barriers to trade arising from all other factors and is orthogonal to the regressors. ˆ and strucTo recover technology, I follow Waugh (2010) and use the estimated trade costs, κ, ˆ to compute the implied tradable good prices, pˆm , by rewriting equation 6 in tural parameters, S, ˆ terms of S:

pˆ m,i = ( AB)



I

∑e

Sˆj

j =1



κˆ i,j ωi,j

�1/θ

�−θ

ˆ I get the convolution of wages and technology, From the obtained prices and the estimates in S, − β/θ

log(wi

λi ). Then, given the bilateral trade shares, Di,j , and the balanced trade condition in

equation 8, I follow Alvarez and Lucas (2007) and calculate equilibrium wages according to the following equation. L i wi (1 − s f i ) =

I

∑ Lj

j =1

w j (1 − s f j ) D j,i ω j,i Fj

where s f i is the labor share in the production of final goods sfi =

α(1 − (1 − β) Fi ) (1 − α) βFi + α(1 − (1 − β) Fi )

20

Table 5: Simulated concentration level with exporter fixed effect Gini Model Simulation Data

Theil Exports (X)

Exports

Imports

0.99

0.87

0.98

0.92

Extensive

Intensive

Margin

Margin

4.87

3.81

56%

44%

2.46

2.56

49%

51%

Theil Imports (M) Total 8.68 5.02

Extensive

Intensive

Margin

Margin

1.46

1.91

43%

57%

1.06

1.91

35%

65%

Total 3.37 2.98

and Fi is the fraction of country i spending on tradable goods net of tariff expenses. Fi =

I

∑ Dj,i ω j,i

j =1

The obtained equilibrium wages together with tradable good prices, determine the implied technology levels λˆ for each country given the structural estimates of the gravity equation. Table 4 summarizes the regression outcome of the gravity equation. In terms of fitting bilateral trade flows, I obtain an R2 of 0.76 slightly lower than the R2 of 0.83 reported by Waugh. The obtained coefficients on trade costs are consistent with the gravity literature, where distance is an impediment to trade. The magnitudes of the coefficients reported in Table 4 are similar to those in Eaton and Kortum (2002) and in Waugh (2010), which consider a similar sample of countries without tariffs. The overall size of the trade costs in terms of percentage are similar to those reported in Anderson and Van Wincoop (2004). Having identified trade costs and technology, see Table 6 for the estimated technology levels, I simulate the Eaton and Kortum model to test whether the calibrated version can replicate the concentration levels observed in the data. Table 5 presents the mean concentration levels for the simulated countries. The results show that the calibrated model replicates the fact that countries are more specialized in exports than in imports on all margins. Focusing on the simulated concentration levels reveals that countries concentrate excessively with respect to the data. The simulated concentration levels of exports and imports are almost twice as high as the ones observed in the data. The underlying reason is that countries concentrate too much on the extensive margin. Mean export (import) concentration on the extensive margin is 5.78 (2.46) compared to 2.07 (1.06) in the data. This implies that simulated countries export (import) only 0.2% (10.4%) of the product space compared to 8.5% (34.4%) in the data. Figure 4 plots the corresponding cross country pattern for simulated and empirically observed concentration levels against the simulated log of GDP. The results show that the model replicates the empirical pattern with export concentration decreasing in market size. However, the simu-

21

lated concentration levels on the extensive margin are too high, particularly for smaller economies. Countries excessively specialize in exporting few products. On the importing side, the model is unable to replicate the L shape relationship between market size and concentration on the extensive margin. In the calibrated model all countries import relative few goods. Turning the attention to the intensive margin, Figures 4(e) and 4(f), the results show that consistent with the data the model predicts no relationship between concentration and market size. Overall, the calibrated model is able to replicate the qualitative pattern for exports but produces excessive concentration levels relative to the data, particularly on the extensive margin. The underlying reasons for the excessive concentration levels lie in high trade costs and the structure of the model. Note that simulated trade shares are closely related to bilateral trade shares observed in the data. While the model reproduces the bilateral trade volumes, it fails to capture the underlying distribution of trade volumes across products. To shed light on why countries trade too few products, I plot the share of the number of exported and imported products against the number of trading countries and compare it with the data. Figure 5 shows the results. In the case of exports, simulated countries export their goods to too many destinations. For example, in the data (in blue) more than half of the products are exported by 16 or more countries. In the simulation (in red) no product is exported by more than 16 countries. This result shows that the assumed productivity distribution generates extremely efficient producers that even firms facing high trade costs can sell their products to many destinations in the world. As a consequence, the number of exporting countries per product is small. Turning the attention to imports, Figure 5(b) shows that, similar to exports, the majority of products is imported by few countries. In the model, 90 percent of products are imported by less than 35 countries whereas in the data the same number of products is imported by 25 or more countries. The cause are high import costs, which prevent countries from taking advantage of cheaper goods from abroad.

6

Robustness checks

The first part addresses concerns on the robustness of the empirical observed concentration indexes. In particular, the level of disaggregation as well as the classification scheme chosen may affect the empirical concentration measures and the decomposition of the intensive and extensive margin. For this reason, I re-calculated the concentration indexes on all margins by defining a product to correspond to a 4 digit SITC code instead of a 6 digit HS code. The implied product space changes significantly as it comprises only 792 products compared to 5015 products using 6 digit HS codes. Also, the SITC classification scheme differs from the HS classification scheme by grouping products based on economic functions rather than their material and physical properties. The empirical estimates of the SITC industry classification are very similar to the 6 digit HS code sample. The correlation coefficient between the SITC and HS concentration indexes is 0.9 for exports and 0.7 for imports. The obtained concentration levels are slightly lower because of the higher level of aggregation. However, the core results remain the same. Exports are more con-

22

centrated than imports on all margins and the intensive margin dominates concentration for both exports and imports. The obtained shares of the intensive margin in terms of overall concentration are almost identical to the standard sample with 55 percent for exports and 66 percent for imports. Also, the cross country concentration patterns feature a negative log linear relationship between concentration on the extensive margin and market size for both exports and imports. In sum, the obtained results on the 4 digit SITC level support the empirical evidence based on the 6 digit HS classification and highlight the level of generality my results apply. Another concern may be that the results are driven by trade in agricultural products and natural resources. Oil and Gas are the products with the largest trading volume and can therefore drive concentration on the intensive margin. To address this issue, I consider only manufacturing products (i.e. products with an HS code above 280000 and below 970000) neglecting trade in primary sector products and services. Based on the resulting 4438 product lines, I calculate the corresponding concentration indexes on all margins and compute the correlation coefficient between the original data and the manufacturing sample. The results show that the correlation coefficients are above 0.8 for almost all indexes with the exception being the concentration index on the intensive margin of exports with a coefficient of 0.46. As expected, neglecting trade in primary products reduces concentration on the intensive margin, in particular for oil producing countries. However, the other concentration levels and the cross country concentration patterns are almost identical to the ones based on the full 6 digit HS sample. The decomposition implies an intensive margin share of 48 percent for exports and 61 percent for imports compared to 51 and 68 percent in the data. Overall, the results based on manufacturing products do not alter the conclusions derived from the full sample including all sectors.5 Finally I want to address the discrepancy of the product space between the data and the model caused by intra-industry trade. In the main part of the paper I establish correspondence between the model and the data by netting out within product trade and considering only trade across products. This approach leaves valuable information unused and may bias the results because intra-industry trade flows occur predominantly between OECD countries and to a lesser extend between developing countries. In an alternative approach, I deal with intra-industry trade by developing a “measurement device” that enables the model to characterize trade within and across products. The basic idea is that in reality the true state of the world is indeed Ricardian, i.e. varieties are in fact products, but the data are not sufficiently disaggregated to capture the true product level. Instead, these “Ricardian products” are aggregated into sectors according to a classification scheme, i.e. HS codes. The suggested procedure converts the measurement of product units in the model to product units in the data and allows to examine gross trade flows. Because the classification scheme is unobserved, I assume that varieties are randomly assigned to an HS code following a Poisson process. Using the structure of the model, I can then estimate the Poisson parameter and characterize the “measurement device” completely. I obtain a value of 7 for the Poisson parameter 5 The

appendix contains detailed results for both approaches.

23

implying that, on average, 7 “Ricardian products” comprise an HS product category. Based on this result, I apply the Poisson process to group simulated Ricardian products randomly into artificial HS codes for which I calculate the implied concentration indexes. The results, presented in detail in the appendix, show that this approach leads to similar results as the net trade flow approach. In particular, it implies a similar value for the elasticity of substitution, η = 8 (compared to η = 7.3 in the net trade sample), and an exporter fixed effect to reconcile the cross country concentration pattern on the intensive margin. However, mean trade costs are with κ = 0.6 significantly higher than the κ = 0.71 mean trade costs in the net trade flow case.

7

Conclusion

Export and import concentration in combination with a decomposition into an extensive and intensive product margin concentration measure provide new quantitative and qualitative evidence on specialization patterns in world trade. Based on detailed trade data, my calculations show that exports are more concentrated than imports on all margins and specialization is dominated by the intensive product margin. The extensive product margin explains the gap between export and import concentration and drives specialization differences across countries. Larger economies diversify more because they export and import more products. These qualitative findings together with the calculated concentration levels provide a new test for trade theory. Evaluating the empirical performance of the underlying theory is important because the derived theoretical predictions have decisive implications on the effect of trade on production and further instance on welfare. In this paper, I evaluate the ability of the Eaton Kortum model, the workhorse model of modern Ricardian trade theory, to account for the observed patterns. My findings show that the calibrated model is able to replicate the empirical evidence qualitatively but fails to account for the levels of specialization. In particular, the model produces an excessive degree of specialization on the extensive margin for both exports and imports. The main reasons why countries export and import too few products are high trade costs and the underlying productivity distribution. The Fréchet distribution generates extremely efficient producers of tradable goods enabling them to capture the whole world market and resulting in too few exporters per product. High trade costs prevent countries from importing more goods. However, if trade costs are chosen arbitrarily, the model is flexible enough to replicate the empirical results on all margins. Finally, I want to stress that my analysis can be readily applied to other trade theories as well. In this paper I study specialization patterns in the Ricardian framework assuming that specialization patterns emerge through comparative advantage induced by technological differences. Other models of international trade, most notably monopolistic competition based on Krugman (1980) and Armington models, also develop quantitative predictions about specialization patterns. However, in both types of models, goods are differentiated by location of production and countries

24

completely specialize in disjoint sets of goods requiring an adaption of the product space in the empirical analysis. Nevertheless, it would be interesting to compare the performance of both types of models to the results in this paper in order to gain further insights on the relevance of each trade theory.

References A LVAREZ , F., AND R. J. L UCAS (2007): “General equilibrium analysis of the Eaton Kortum model of international trade,” Journal of Monetary Economics, 54(6), 1726–1768. A NDERSON , J., AND E. VAN W INCOOP (2003): “Gravity with gravitas: a solution to the border puzzle,” American Economic Review, 93(1), 170–192. A NDERSON , J., AND E. VAN W INCOOP (2004): “Trade costs,” Journal of Economic Literature, 42(3), 691–751. B ALASSA , B. (1963): “An empirical demonstration of classical comparative cost theory,” The Review of Economics and Statistics, 45(3), 231–238. B OWEN , H., E. L EAMER , AND L. S VEIKAUSKAS (1987): “Multifactor, multicountry tests of the factor abundance theory,” American Economic Review, 77(5), 791–809. B RODA , C., AND D. W EINSTEIN (2006): “Globalization and the Gains from Variety,” Quaterly Journal of Economics, 121(2), 541–585. C ADOT, O., C. C ARRÈÈRE , AND V. S TRAUSS -K AHN (2011): “Export Diversification: What’s behind the Hump?,” The Review of Economics and Statistics, 93(2), 590–605. C HOR , D. (2010): “Unpacking sources of comparative advantage: A quantitative approach,” Journal of International Economics, 82(2), 152–167. C OSTINOT, A., D. D ONALDSON , AND I. K OMUNJER (2012): “What goods do countries trade? A quantitative exploration of Ricardo’s ideas,” Review of Economic Studies, 79, 581–608. E ATON , J., AND S. K ORTUM (2002): “Technology, Geography, and Trade,” Econometrica, 70(5), 1741–1779. G OLUB , S., AND C. H SIEH (2000): “Classical Ricardian theory of comparative advantage revisited,” Review of International Economics, 8(2), 221–234. H UMMELS , D., AND P. K LENOW (2005): “The Variety and Quality of a Nations Exports,” American Economic Review, 95(3), 704–23. K RUGMAN , P. (1980): “Scale Economies, Product Differentiation, and the Pattern of Trade,” The American Economic Review, 16(4), 670–675.

25

L EAMER , E. (1984): Sources of comparative advantage. MIT Press„ Cambridge, Mass. L EVCHENKO , A., AND J. Z HANG (2011): “The evolution of comparative advantage: Measurement and welfare implications,” Discussion paper, National Bureau of Economic Research. M AC D OUGALL , G. (1951): “British and American exports: A study suggested by the theory of comparative costs. Part I,” The Economic Journal, 61(244), 697–724. S CHOTT, P. (2003): “One size fits all? Heckscher-Ohlin specialization in global production,” The American Economic Review, 93(3). S HIKHER , S. (2011): “Capital, technology, and specialization in the neoclassical model,” Journal of International Economics, 83(2), 229–242. T REFLER , D. (1995): “The case of the missing trade and other mysteries,” The American Economic Review, pp. 1029–1046. WAUGH , M. E. (2010): “International trade and income differences,” American Economic Review, 100(5), 2093–2124.

26

Figures

Gini index − HS 6 digit

Total Theil index − HS 6 digit

Mean of the index from 1992−2009

Mean of the index from 1992−2009

6 4 2

Export Concentration

.95 .9 .85

ITA GER

NGA TCD YEM OMN GAB COG IRN KWT GNB SAU QAT MLI COM ARE BFA WSM BWA AZEBDI RWA KIR BMU BEN SDN MWI GIN STP CAF VEN DZA ZMB MRT NER VUT GMB CUB JAM ECUSYR MOZ PNG ETH UGA SLEBHS ATG CMR GUY CPV ERI GHA PRY MNG NOR TTO TGO ARM BHR KAZ DJI BOL ISL BLZ MLT BRB KGZ TZA MUS CRI COL ZWE FJI GEO SEN NIC KHM RUS PER PAN KEN SWZ CHL MDGNPL JOR HND EGY SLV LBN LVA GTM BGD CYPPHL ALB VNM AUS DOM IRL ARG LTU MDA BIH ISR MEX URY AFG MYSBLR CAN NZL MKD TUN MAR ZAF GRC LKA EST PAK BRA FIN UKR SRB HRV SVK KOR PRT HUN IDN ROM BGR DNK SWE THA ESP SVN IND TUR POL CHE FRA GBR CZE BEL JPN AUT NLD USA CHN GER ITA

0

.75

.8

Export Concentration

8

TCD GAB KWT COM YEM COG GNB QAT BDI KIR RWA MLI MRT WSM DZA OMN GIN BMU CAF BEN BWANGA VUT SDN SAU ARE STP MWI BFA PNG NER CUB GMB ZMB AZE CPV MOZ BHS CMR GUY IRN ETH JAM ERI GHA VEN ECU UGA TTO PRY MNG TGO BHR BOL ISL DJI SLE ATG BRB ARMATG ARM SEN KAZ TZA BLZ NIC MLT MUS KGZ FJI KHM SYR GEO MDG ZWE HND JOR PER NORCOL CRI SWZ CHL BGD KEN SLV LBN DOM PAN ALB GTM CYP MDA NPL URY RUS LVA EGY IRL BIH AUS ARG NZL LTU MARVNM MKD PHL GRC LKA BLR AFG PAK MYS MEX TUN EST SRB CANHRV ISR ZAF FIN PRTROM BRA UKR SVK HUN BGR KOR DNK IDN POLSVNSWE TUR THA ESPCHE AUT CZE BEL NLD GBR IND FRA JPN USA CHN

1

8

.75

.8

.85

.9

.95

1

0

2

4

Import Concentration

(a) Gini coefficient

8

(b) Theil index Mean of the index from 1992−2009 6

Extensive Theil index − HS 6 digit

Mean of the index from 1992−2009 6

Intensive Theil index − HS 6 digit

4 0

PAN

KIR

TCDSTP GNB WSM ERI CPV BDI VUT RWA BMU BENGMB CAF DJI GAB COG MRT ATG BFA GIN MWI SDN BWA BHS PNG MLI MOZNER ETH DZA GUY BRB CUB UGA QAT YEM MNGARM BLZ TGO ZMB KWT CMR JAM SEN AZE NGA GHA PRY NIC SLE BOL KHM TZA BHR FJI ISL TTO OMN ALB GEOKGZ MLT MDG LBN NPL KAZ MUS CYP AFG JOR SAU SLV VEN ECU HND DOM SWZ BIH MDA BGD URY GTM ZWE CRI MKD KEN PAN GRC LVA IRN PER HRV LKA SRB CHL MAR SYR LTU TUN ARE BLR EST NOR EGY COL NZL VNM ARG PRT ISR IRL PHL ROM AUS PAK UKR FIN MEX SVK SVN CAN BGR HUN DNK POL RUS MYS TUR AUT CHE SWE ESP CZE BRA BEL ZAF GBR IDN THA KOR NLD FRA USA IND JPN ITA GER CHN

2

4 2

Export Concentration

COM

IRN ARE NGA OMN SAU YEM SYR KWT VEN NOR QAT ECU MLI AZE RUS GAB COG BWA COL ZMB BFA DZA JAM TCD SLE KAZ SDN MWI CRI TTO GIN ZWE CHL PER CMR NER CUB EGY GHA PRY KEN AUS MRT MOZ IRLMYS BEN VNM BHR MLT PHL ETH UGA MEX PNG MUS ARG ISL BRA ISRZAF BDI GUY CAF SWZ CAN BOL MNG BHS LVA BMU KGZ RWA GNB WSM TGO KORIND JOR HND NPL GTM ARM TZA LTU IDN GEO FJISLV BGD BLR GMB ATG MDG NZL PAK FIN THAJPN HUN FRA UKR BLZ SVK KHM NIC SEN TUN URY MDA LBN CYP SWE DOM ESP MAR STPBIH GBR PRT VUT NLD BRB EST CHE DNK TUR CHN MKD LKA BGR GRC USA CZE BEL ALB POL GER ROM COM SVN SRB KIR HRV AFG DJI AUT ERI CPV ITA

0

Export Concentration

6

Import Concentration

0

2

4

6

0

Import Concentration

2

4

6

Import Concentration

(c) Intensive margin

(d) Extensive margin

Figure 1: Average export versus import concentration for the period 1992 to 2009 for 151 countries

27

8 6 4

STPKIR

VUT BMU AFG PAN ATG GNB ERI BHS COMWSM ARM DJI CAF SLE TCD IND NER MRT TGO BEN BFA KHM GEO KGZKHM KGZ GMB BDI MLI RWA GIN MNG PHL MWI MLT MDA UKR NPL JPN CPV ZWE YEM BLR KOR FJI COG USA SEN MOZ AZE BHR UGA CYP GUY ZMB MDG ETH LTU ZAF BLZ ITAGER CHN TZA QAT THA PRY SWZ PNG SDNCUB LBN JAM KEN MYS BRBNIC OMN PAK GHA ALB CMR BIH JOR BGR BGD TTO HND DOM ISR SVK NLD GAB SYR URY LKA HUN BRA KAZ KWT VNM IRL MUS SRB MAR SLV FIN TUR CHE FRA EST DZA IDNESP BWA SWE EGY GRC BEL BOL CHL ISLMKD GTM CRI CZE IRN NGA LVA ROM PER NZL ARE ECU PRT TUN SVN AUT HRV GBR SAU POL RUS AUSCAN MEX COL NOR DNK VEN ARG

0

Import Concentration

NGA TCD YEM OMN GAB COGMLI IRN GNB QAT KWT ARE SAU COM WSM BDI BFA BWA AZE RWA BMU BEN SDN DZA GIN MWI MRTNER CAF ZMB VUT GMB CUB VEN JAM ECU MOZ PNG UGA SYR ETH BHS SLE ATG CMR CPV GUY ERI MNG GHA PRY NOR TTO TGO ARM KAZ DJIBLZ ISL BHRBOL MLT BRB KGZ TZA MUS CRI COL ZWE FJISWZ SEN NIC GEO KHM RUS PER PAN NPL KEN CHL EGY MDG JOR HND SLV LBNGTM BGD LVA CYP ALB VNM DOM IRL MDA ARGAUSMEX LTU BLR ISRPHL BIHURY AFG CAN NZL MYS MKD TUN MAR ZAF GRC LKA EST PAK BRA FIN UKR SRB HRV SVK HUN PRT KOR IDN ROM BGR DNKCHE SWE THA SVN IND TUR ESPFRA POL GBR CZE BEL JPN NLD AUT USA GER CHN ITA

2

8 6 4 2

Total Theil index − HS 6 digit Mean of the log index from 1992−2009

0

Export Concentration

STPKIR

Total Theil index − HS 6 digit Mean of the log index from 1992−2009

−10

−5

0

−10

Log of GDP relative to the US R2=0.39

0

R2=0.38

(a) Overall concentration of exports

(b) Overall concentration of imports Extensive Theil index − HS 6 digit

Mean of the log index from 1992−2009

Mean of the log index from 1992−2009 6

Extensive Theil index − HS 6 digit 6

TCD BDI ERI RWA BMU GMB BENGAB DJI CAF COG MRT ATG BFA GIN MWI SDN BWA BHS NER PNG MLI MOZ DZA GUY BRB UGAETH QAT CUB YEM MNG BLZ TGO ARM ZMB JAM AZE KWT CMR SEN NGA GHA PRY NIC SLE BOL KHM TZA BHR FJI ISL KGZ TTO OMN GEO ALB MLTMDG NPL LBN KAZ MUS CYP AFG JOR SLV VEN SAU ECU HND DOM BGD SWZ MDA BIH URY GTM ZWE CRI MKDPAN GRC IRN LVAKEN HRV LKA SRB CHL MARPER SYR TUN BLR EST LTU NOR EGY COL NZLARE VNM ARGAUSMEX PRT ISR PHL ROM PAK UKR FIN SVKIRL SVN CAN BGR HUN DNK POL RUS MYS TUR ESP AUT CHE SWE CZE BRA BEL ZAF GBR IDN THA KOR FRA NLD USA ITAIND GERJPN CHN

Import Concentration

0

4

KIR STP

VUT GNB ERI TCD AFG COM WSM SLE DJI CAF ATG BMU ARM GMB MRT RWA KGZ COG NER CPV BDI BEN MWI MNG BFAGEO TGO MOZ MLI KHMNPL FJI GIN GUY PNGBIH AZE ZWE BLZ YEM MDA BRB ZMB ALB SWZ UGA CMR GAB SEN SDNCUB QAT MDG SYR KAZ BHR ETH BHS CHN TZA GHA BLR NIC EST GERJPN KEN PAK LVA MKD DOM HND KWTBGD MLT LTU OMN ITAIND VNM JAM MUS PRY BGR TTO UKR PAN BOL LKA URY SLV JOR USA IRN NGA LBN ISL CYP THA SVN IDN NLD FRA KOR SVK HUN BRA GTM ZAF CZE TUR PMYS PHL HL PER TUN SWE GBR POL ECU RUS ISR ROM BWA CRI BEL MAR CHE ARE EGY AUT ESP ARG DNK SRB FIN COL HRV DZA NZL IRL CHL MEX PRT VEN NOR GRC SAU AUSCAN

0

2

4

COM GNB WSM CPV VUT

STP

2

KIR

Export Concentration

−5

Log of GDP relative to the US

−10

−5

0

−10

Log of GDP relative to the US R2=0.80

(d) Extensive margin of imports

Mean of the log index from 1992−2009 6

Intensive Theil index − HS 6 digit

Mean of the log index from 1992−2009

PAN BHS

BMU IND PHL KOR UKR MLT CYP JPN USA ZAF MYS THA BLRIRL LBN LTU ISR ITA GRC PRY JAM ATG SVK BHR HUN SRB SEN FIN EGY MDA DZA OMN MAR JOR AFG CHL ARM ESP BRAGER CHN CHE KEN KHM TUR BGD BWA PAK BGR SWE ETH BEL TZA MDG TTO SAUNLD URY NZL PRT GEO YEM TGO UGA LKA HND BFA ROM IDNMEXFRA GTM MLI DOM CZE ARE GIN NIC SLV QAT BEN PER HRV GHA AUS CRI ECU CUB VNM ISL ZMB NER MUS AUTPOL SDN KWT TUN NOR MNG MKD NGA BOL AZE VUT ZWE IRN GBR COL VEN RUS SWZ MRT SVN FJI MWI KGZ DNK NPL CMR CAN BRB SYR EST BLZ LVA ALB ARG KAZ DJI GUY GAB MOZ BDI GMB PNG BIH RWA CPV CAF COG STP COMWSM SLE ERI TCD GNB KIR

0

MLI RUS COG GAB BWA COL ZMB BFA JAM SDN DZA TCD SLE MWI KAZ CRI TTO GIN ZWE NER CHL PER CMR EGY AUS PAN GHA PRY KEN CUB PHL MRT MLT MOZBHR IRL MYS BEN VNM ZAF ETH UGA MEX PNG LVA MUS ARG ISL ISR BDI SWZ GUY CAF CAN BOL MNG BHS BRA KGZ RWA GNB KOR WSM BMUTGO JOR HND NPL ARM TZA IND LTUGTM IDN GEO BGD GMBBLZFJI BLR ATG MDG NZLHUN PAK SLV FIN UKR THA JPN SVK KHM NIC SEN TUN URY MDA LBN SWE NLD ESPFRA DOM MAR CYP STP GBR CHN PRT VUT BRB EST CHE DNK BIH TUR MKD LKA BGR USA GRC CZE BEL ALB POL GER ROM SVN SRB KIR COM DJI AUT HRV AFG ERI ITA CPV

4

Import Concentration

4

IRN ARE NGA OMN SAU YEM SYR KWT VEN NOR QAT AZE ECU

2

6

Intensive Theil index − HS 6 digit

0

2

0

R2=0.56

(c) Extensive margin of exports

Export Concentration

−5

Log of GDP relative to the US

−10

−5

0

−10

Log of GDP relative to the US

−5

0

Log of GDP relative to the US

R2=0.01

R2=0.14

(e) Intensive margin of exports

(f) Intensive margin of imports

Figure 2: Average export and import concentration versus the log of average relative GDP with respect to the United States (log( GDPUS ) = 0) for the period 1992 to 2009 for 151 countries.

28

Total Theil Index

Total Theil Index

Data versus simulation

Data versus simulation

12

8 Red − Simulated Data Blue − Empirical Data

11

Import Concentration

Export Concentration

9 8 PLW PLW

7 6 5 4 3 2

NGA TCD YEM GAB OMNKWT COG IRN MLI GNB SAU QAT BDI COM ARE BFABWAAZE RWA MLI BMU BEN SDN DZA GIN MWI MRT CAF COM GNB ZAR NER VEN CPV ZMB DJI VUT BDI GMBBMU CUB JAM ECU MOZ PNG SYR MRT NER ETH UZB ATG BHS ERI UGA CMR CPV GUY ERI CAF GHA PRY NOR MNG COG TTO MLT BLZ VUT TGO GMB FJI MWI ARM MOZ PNG BHR GUY KAZ PAN BRA DJI BOL ISL MNG SEN MLT BLZ BRB ZMB GEO KGZ JOR LVA GIN LAO TCD TZA BHS BFA NIC ARM YEM MUS POL CRI TZA COL ZWE UKR ZAF FJI TUN SEN LBN TGO MDA NIC ZAR GEO EST SWZ ALB KHM RUS GER URY UZB PAN NPL RWA KEN LTU SWZ CHLCOL BGR DOM PER MDG JOR BWA BRB LAO JAM HND BEN ISL MUS GBR ETH NLD THA GAB EGY KGZ ARG VEN MDG CRI OMN SLV MKD HUN KWT BHR AFG IRN KHM TUR LBN LVA NOR GHA CYP BIH GTM BLR PER HRV BGD UGA ECU IDNESP AUT CMR CYP ALB PRY KAZ TTO VNM BGD AUS PAK DOM PHL GTM BOL IRL MDA QAT CHE SVN CHL SLV ARE AZE IND ARG ISR SWE BEL LKA CAN LTU EGY URY SDN GRC BIH ROM SAU ISR SVK CUB MEX SRB SYR ITA BLR NZLVNM AFG MAR MYS FRA CZE CAN PRT NGA MEX NZL DNK MKD MAR FIN IRL KOR MYS RUS DZA ZAF GRC LKA EST TUN PAK BRA FIN UKR SRB HRV CHN SVKHUN PRT KOR JPN IDN BGR ROM DNK SWE THA SVN IND TURESPFRA POL CHE GBR CZE BELNLD JPN AUT USA GERCHN USA ITA

1 −6 10

−4

−2

10

6 VUT

5

PLW

4

PLW

3 2

10

−4

10

0

10

Log of GDP relative to the US

(a) Overall concentration of exports

(b) Overall concentration of imports

Extensive Theil Index

Extensive Theil Index

Data versus simulation

Data versus simulation

6

6 Red − Simulated Data Blue − Empirical Data

PLW

COM

PLW

TCD GNB CPV BDI VUT ERI RWA COM BMU GNB GMB VUT DJICAF BEN GAB COG MRT ATG BFA GIN ZAR ATG MWI BWA SDN BHS NER PNG CPV MLI MOZ UGA ETH DZA GUY BRB QAT CUB YEM GMB MNG BLZ TGO ARM ZMB AZE KWT LAO DJI JAM BMU CAF BDI FJI CMR ERI SEN MRT NGA GUY ZWE BLZ GHA NIC PRY TGO BOL KHM SWZ TZA BHR FJI MWI KGZ ISLGEO COG TTO TCD RWA OMN GIN MNG MLT BRB BEN NER LAO ARM ALB BFA BHS UZB MOZ MDA MLT KGZ MDG MLI NIC PNG ISL MUS LBN KAZ ALB AFG SEN ZAR MDG ZMB GAB NPL GEO EST MKD MUS CYP AFG KHM BIH BHR JOR CYP SLV DOM ECU VEN SAU HND UGA GHA CMR NPL HND LVA JOR PAN BIH TZA JAM TTO SWZMDA BWA PRY YEM BOL UZB URY BGD QAT OMN SLV ETH AZE KEN LTU SVN URY LBN GTM ZWE CRI TUN SYR HRV SDN KWT LKA DOM MKD BLR BGR SRB SVK KEN PAN CUB GTM ECU GRC IRN LVA MAR PER HRV LKA SRB NZL CHL BGD KAZ VNM MAR SYR PRT NGA CZE TUN DNK PHL ROM ARE FIN BLR VEN EST LTU NOR CHL UKR ARG AUT IRL PER MYS THA SWE PAK BEL DZA ZAF CHE EGY POL COL ISR SAU HUN NZL ITA GRC TUR GBR IRN IDNMEX NLD AUS FRA VNM ARG PRT GER ESP ISR IRL PHL ROM RUS AUS IND KOR CAN PAK UKR FIN BRA MEX SVKHUN SVN CAN BGR DNK POL RUS JPN MYS TURESP AUT CHN CHE SWE CZEBEL BRA ZAF GBR IDN THA KOR FRA NLD IND JPN USA ITA GERCHN USA

4

3

2

1

0 −6 10

−4

−2

10

Red − Simulated Data Blue − Empirical Data 5

Import Concentration

5

Export Concentration

−2

10

Log of GDP relative to the US

VUT GNB

3 PLW

2

0 −6 10

10

ERI TCDAFG

COM

1

0

10

PLW

4

DJICAF COM VUT GNB ATG ATG LAO BMU CPV GMB DJI BLZ ZAR ZWE RWA ARM GMB GUY CAF BDI MRT KGZ COG NER CPV BMU ERI BDI NER COG TGO BEN FJI BRB MWI MNG MRT SWZ BFA NPL RWA TGO GEO GIN BHS MOZ MLI GIN KGZ TCD LAO MWI KHM UZB BEN MDA FJI MLT MLI MUS GUY MOZ BFA PNG ALB GAB MDG ARM ISL AZE ZMB MKD PNG ZWE AFG NIC ZAR SEN BLZ YEM GEO BIH KHM MDA BHR BIH EST TTO PRY BRB BWA NPL HND JOR ZMB ALB SWZ CYP UGA GHA AZE PAN UGA LVA ETH CMR JAM BOL KEN GAB TZA SEN SDNKAZ QAT SLV URY CRI YEM MDG LBN LTU UZB SYR BHR ETH BHS CHN CUB QAT OMN SDN GHA LKA BLR BGR SVN NICESTTZA TUN DOM HRV GTM GER BLR ECU KEN PAK KAZ LVA SRB IND MKD DOM ISR HND KWT HUN MLT LTU OMN PER ITA CUB JAM VNM MUS PRY BGR MAR JPN USA TTO UKR KWT SVK IRL PAN BOL DZA LKA URY SLV BGD NZL JOR IRN CHL NGA VNM LBN ROM USA ISLCYP DNK FIN ARE UKR CZE THA CRI PHL SVN MYS IDN VEN NLD FRA COL NOR KOR PAK PRT EGY SVKHUN BRA CHE GRC GTM ZAF CZE AUT SWE TUR PER PHL GBRJPN POL ECU RUS ISR ROM BWA TUN BEL MAR MYS CHE ARE EGY AUT THA ESP DNK ARG SRB KOR FIN IRN COL SAU POL HRV DZA IDN CAN CHN NLD ESP NZL IRL TUR CHL MEX PRT VEN CAN NOR AUS BRA GER IND GRC SAU RUS FRA GBR ITA −4

10

0

10

Log of GDP relative to the US

(c) Extensive margin of exports

(d) Extensive margin of imports

Intensive Theil Index

Intensive Theil Index

Data versus simulation

Data versus simulation

8

8 Red − Simulated Data Blue − Empirical Data

7

OMN

4 3 PLW PLW

Import Concentration

5

ARE NGA

Red − Simulated Data Blue − Empirical Data

7

6

2

−2

10

Log of GDP relative to the US

Export Concentration

AFG

BMU

PAN ATG GNB ERI BHS COM ARM TCD VUT COM DJI GNB ATG CPV GMBCAF IND NER MRT DJI BLZ LAO TGO BEN BFA KHM GEO KGZ CAF ZWE GUY GMB MLI RWA GIN BDI MNG FJI ERI PHL BMU BDI ZARNPL MWI MLT MDA UKR BHS ALB RWA JPN MRT ARM MDG PNG TCD KGZ BFA BRB MNG CPV ZWE YEM MWI KOR TGO NER COG UZBBLR EST AZE GIN BEN FJI USA SEN LAO MLI MDA NIC MOZ COG BHR BWA MUS MLT BIH GAB UGA CYP GUYSWZ ISL GEO KHM NPL MKD ZMB MDG ETH AFG LTU ZAF BLZ ITA TZA TTO QAT GERCHN THA CMR PRY SWZ ZAR PNG SDN LBN BOL JAM KEN UGA MYS BRB SEN UZB QAT ZMB PAK PAN LVA SVN CUB OMN GHA HND NIC ALB ETH CMR BIH SLV PRY JOR AZE BGR LTU LBN CRI TTO BGD HND DOM ISR SVK JOR NLD BRA GAB SYR TZA URY LKA YEM HUN KAZ KWT DOM VNM SYR ECU IRL MKD BLR MUS SRB MAR SLV FIN TUR CHE FRA ESP EST TUN DZA IDN BGR BWA SWE LKA EGY HRV SDN GRC BOL BELIRN CHL ISL LVA PHL GTM CRIGTM MYS CZE NGA ROM NZL PER ARE BGD ECU DZA DNK KAZ HUN PRT TUN CHL SVN ISR CZE NGA AUT HRV COL MAR PER GBRJPN SAU IRL GRC POL RUS VNM FIN ROM AUS SVK CUB MEX UKR COL KWT NOR EGY DNK ARE NZL GERCHN USA VEN IND CHE PAK ZAF ARG BEL SWE THA AUT SAU CAN ITA IRN FRA GBR POL KOR TUR IDN NLD MEX AUS ESP BRA CAN RUS

1 −6 10

0

10

Red − Simulated Data Blue − Empirical Data

7

10

IRN SAU

YEM SYR KWT VEN MLI BRA NOR QAT MLI AZE ECU RUS COG GAB UZB BWA NER POL GER DJI BDI COL ZAF ZMB UKR BFA COL MRT DZA JAM PAN BMU TCD SDN ERICOG GBR MWI NLD KAZ MLT CRI CPV ZWE TTO TUN GIN CHL PER NER CMR CUB PNG JOR IRN LVA THA HUN EGY PAN MOZ GHA ESP TUR ARG LBN YEM PRY VEN KEN CAF AUS SEN PHL BGR ZAR MRT MOZ TZA IDN MYS DOMIRL BLZ MWI BEN VNM GEO NOR KEN ZMB PER BHR ZAF CAN MLT ETH UGA LTU MEX IND PNG MNG URY UZB MUS AUT ARGAUS NPL FJI ISL ISR BDI GUY CAF SWZ PAK CAN BOL CHE MNG BHS LVA PHL NIC BRA CRI HND BWA BFA KWT JAM KAZ GIN ETH ISR GRC EST ARM TCD KGZ CHL RWA ECU ZAR EGY GNB BLR BGD ATG SAU IDN TGO SWE BEL KOR OMN HRV ALB JOR HND NPL ARE ITA GTM ARM GMBBMU MDA TZA IND FRA MEX KOR LTU ROM CHN RUS GNB GEO FJI BGD BLR GMB ATG MDG NZL PAK SLV ZWETGO FIN GHA GAB ISL THA MUS BHR MKD HUN COM NZL MDG RWA UKR BRB LKA JPN BLZ CZE CUB PRY CMR CYP SVN SWZ UGA LAO NGA QAT SDN PRT SVK TTO VNM KHM BEN DNK BIH BOL NIC SRB IRL SEN AFG MYS TUN SLV KGZ DZA MAR URY AZE FIN MDA SWENLDESPFRA LBN SYR DOM MAR CYP GBR CHN USA PRT VUT BRB EST CHE DNK BIH TUR MKD LKA BGR ROM USA GRC CZE BEL ALB POL GER VUT COM SRB AUT HRV AFG SVN DJI ERI ITA CPV

1

6 5 PAN

4 BMU

3 2

PLW

1

BHS

IND PHL KOR UKR MLT CYP JPN USA ZAF MYS THA BLR LBN EST LTU ISR ITA ARM GRC IRL CYP BWA ALB PRY MDG SVN QAT PNG JAM ATG BHR SVK BFA PHL MYS SEN SRB HUN NPL FIN MDA UZB DZA OMN MAR NLD GER AFG JOR CHL ARM IND NIC ESP BRA ITA CHE CMR BIH KEN BOL JAM CHN JPN KHM TUR BGD FRA GBR BWA ECU EGY MOZ PAK BHS SRB BGR ATG TCD BLR SWE ETH FJI BEL MWI LTU MDG TZA KGZ LBN GRC TTO DOM SAU URY COL CHN NZL CPV BGD PRT MLI LVA SLV DNK GEO CZE YEM FRA TGO UGA ISL RWA LKA PAN GAB CRI BEN ETH HND BFA GTM ROM IDNMEX VUT MKD BDI ERI CHL NGA DOM MDA LAO URY DZA MLT ARG GMB MRT ARE GNB AFG TUN FIN SAU GIN COM HND NIC SLV BEN QAT SYR PER HRV GHA YEM AUS CAF AZE ROM EGY CRI BGR RUS ECU CUB BRB UKR THA VNM GUY ZAF ISL TZA USA ZMB LKA IRN BMU VEN NER MUS AUT CHE SDN ZAR ZWE KWT BEL BLZ IRL HUN TUR BRA TGO TUN DJI PRY MAR NOR POL SWE MNG ISR ARE PAK MKD SEN NLD PER IDN NGA KAZ KOR BOL AZE VUT IRN JOR NER GBR SVK COL ESP VEN ZMB POL RUS SWZ GTM CAN MRT KWT CUB NZL SVN FJI MWI KGZ NPL DNK CMR CAN BRB SYR KAZ BLZ EST LVAUZB ALB ARG GUY DJI GAB MOZ BDI COG GMB PNG BIH RWA CPV CAF LAO COG ZAR COM ERI TCD GNB

PLW

0 −6 10

−4

10

−2

10

0 −6 10

0

10

Log of GDP relative to the US

−4

10

−2

10

0

10

Log of GDP relative to the US

(e) Intensive margin of exports

(f) Intensive margin of imports

Figure 3: Simulated (in red) and empirical observed (in blue) export and import concentration versus GDP across 151 countries. The simulation uses parameterized trade costs to match the data using a country specific export cost.

29

Total Theil Index

Total Theil Index

Data versus simulation

Data versus simulation

16

8 Red − Simulated Data Blue − Empirical Data

12 10 8 6 4 2 −6

−4

10

−2

10

6

BWA CAF CMR RWA DJI BDI VUT ZMB SDN BGD BMU AFG NGA VEN NPL GAB IRN TZA AFG UGA NER BFA KEN YEM BRA PER GIN GNB MWI ETH MOZ MDG PAN BEN OMN ZWE DZA BOL FJI AUS JPN STP COL COGPNGDOMLKA TCD KGZ KIR ATG IND PAK GNB SLV ARG JAM MLI ALB KHM ARM AZE COM USA KAZ TUN ERI CPV MAR VUT BHS TGO MEX SEN COM CHL PHL ARM TCD SYR MRT WSM GMBBRB ISR NIC TTOGTM DJI ECUQAT CUB KWT SLE WSM ATG EGY CRI NZL ZAFSAU CAF BLZ MUS GEO SLE IND ROM GHA BHS LBN MKD BIH NER MRT MDA FIN TGO SRB URY BFA BEN KHM GUY GEO KGZ STP PRY ERI GMB BDI MLT MLI RWA GIN MNG BMU BLR HND NOR JOR TUR GRC PHL IDN RUS CHN JPN MWI MLT LVA CYP MDA HRV UKR NPL LTU BGR SWZ KIRCPV YEM KOR HUN COG ISL FJI SEN BLR USA MNG ZWE SVKPRT MOZ AZE SVN EST BHR CYP GUY MYS VNM ZMB MDG ETH LTU ZAF BLZ UGA ITA TZA CHN QAT GER THA PRY SWZ PNG SDN LBN JAM KENCMR MYS BRB OMN PAK CUB GHA NIC ALB BIH JOR BGR GER TTO BGD HND DOM UKR ISR IRL SVK NLD GAB SYR URY KOR LKA HUN BRA KAZ ARE KWT VNM MKD MUS SRB MAR SLV FIN TUR CHE FRA ESP EST THA DZA IDN BWA SWE EGY GRC BOL BELESP ITA CHL ISL GTM CRIECU CZE NGA IRN POL LVA PAN AUT PER NZL ARE PRT NLDCAN GBR SWE CHE TUN SVN ROM AUT HRV IRL BEL CZEPOL GBR SAU DNK RUS AUS MEX COL NOR DNK VEN ARG CAN

5 4 3 2 1 −6 10

0

10

Red − Simulated Data Blue − Empirical Data

7

COM TCD ERI KIRCPV DJI STP BRB SDN WSM GHA GMB VUT BDI NERSEN ZWE TGO SLE BEN SLV CAF FJI MUS AFG MRT NPL BFA GIN GAB TZA BMU COG KEN KGZ MDG LKA GUY ARM ALB ISL MDA MWI PNG DOM MNG BWA GTM VEN GEO ATG MLI BHS ETHTTO GNB UGA SRB YEM CMR AZE CUB PER ZMB BLZ NIC HND MKD BLR BOL RWA NGA IRN MOZ CYP JOR BIH LBN MLT HRV ECU PRY PAN JAM SYR EST SWZ TUN MAR LVA DZAZAF BGD NZL PRT POL LTU CRI KAZ ROM ISR COL VNM NGA DNK TCD GRC PHL TUR EGY YEM URYSVK PAK SWE MEX OMN AUS KHMBGR CHLIRNAUT UKR KWT MLICOG GAB SVNOMN INDESP IRL SAU ITA QAT BDI COM NLD ARG WSM GNB HUN ARE BFA BWA BRA CZE IDN RWA KIR AZE SDN KWT BMU BENBHR RUS GIN STP MWI THA MRT CAF DZAVEN NER ZMB KOR GBR GMBVUT CUB JAM ECU MOZ PNG CMR SYR ETH BHS UGA CHE CHN SLE ATG GUY ERI CPV SAU GHA PRY QAT NOR MNG TTO CAN TGO BEL ARM BHRBOL KAZ DJI ISL MLT BLZ BRB KGZ NOR TZA FIN COL ZWE FJI MUS SEN NIC GEO KHM RUS PER NPL PAN KEN CRI EGY SWZ CHL MDG JOR HND SLV LBN LVA GTM BGD VNM CYP ALB AUS DOM PHL MDA ARG LTU URY BIHBLR ISR MEX MYS IRL AFG CAN JPN NZL MKDEST TUN MAR ZAF GRC LKA BRA FIN UKR SRB HRV SVKPAK PRT HUN KOR FRA IDN BGR DNK SWE THA ESP FRA GERUSA SVN ROM CZE IND TUR POL CHE BEL NLD GBR JPN AUT CHN GERUSA ITA

Import Concentration

Export Concentration

14

10

−4

0

10

(b) Overall concentration of imports

Extensive Theil Index

Extensive Theil Index

Data versus simulation

Data versus simulation

14

6

10 8 6 4 2 0 −6 10

Red − Simulated Data Blue − Empirical Data

Red − Simulated Data Blue − Empirical Data 5

VUT

Import Concentration

STP CPV KIR WSM ERI COM

12

Export Concentration

10

Log of GDP relative to the US

(a) Overall concentration of exports

CAF NER SLE BMU MNG GMB TCD BDI BFANPL GNB BRB FJI GUY MUS ATG ALB DJI COG ARM AFG ISL MLI RWA KGZ UGA BOL MDG BHS BEN BLZ BIH TGO PRY GEO MWI ZMB YEM BWA PNG TTO ETH JAM SWZ MRTMKD ZWE MDA DOMLKA MLT BGD AZE CUB SEN TUN TZA LBN GHA GIN SRB KEN GAB MAR NIC JOR CYP CMR PAN DNK SLVHRV LVASDN DZA HND LTU MOZ KIR COM PHL URY PER KHM NZL EST SYR KAZ TCD GNB VNM IRL EGY ECU STP SVN BLR CZE WSM VUT BDI VEN ERI CPV RWA ROM GTM SVK PAK CHL BMU IRN POL BGR GMB DJI KWT CAFMRT UKR AUTSWEKOR BEN OMN GRC NGA GAB COGBHR ATG BFA GIN ISR NLD MWI SDN HUN BWA BHS NER PNG MLI MEX MOZ ETH DZA CHE GUY BRB UGA CUB QAT YEM TGO MNG CRI ESP CAN BLZ ARM ZMBAZE KWT COL JAM QAT CMR SEN NGA GHA PRY NIC SLE IND BOL AUS ITA KHM TZA FJI KGZ BHR ISL TTO OMN ARE TUR GEO ALB MLT MDG LBN NPL KAZ CYP AFG PRT VEN JOR SAU THA SLV ECU HND DOM BIH BRAGBR MDA SWZ MUS BEL BGD URY GTM ZWE CRI ARGRUS MKD SAU KEN PAN GRC LVA PER IRN NOR HRV LKA SRB CHL ZAF FIN MAR SYR LTU TUN ARE BLR NOR EST EGY COL NZL VNM ARG PRT ISR FIN IRL PHL ROM AUS PAK IDN UKR MEX SVK SVN CAN BGR HUN DNK POL RUS MYS TUR AUT CHE SWE ESP BRA CZE BEL ZAF GBR IDN THA KOR FRA USA MYS IND NLD ITA GERJPN JPN CHN FRA GERUSA CHN −4

−2

10

3

2

0 −6 10

10

BWA CMR CAF RWA BDI SDN VEN ZMB VUT DJI IRN NPL GAB GNB NER UGA BGDNGA BFA PER TZA AFG BRA MWI YEM KEN BEN MDG ETH BOL GIN GNBTCD OMN FJI AFG COL COM TCD ZWE MOZ DOMLKA DZA COG AUS ARG WSM SLEMLIALB JAMSLV COM AZE USA DJI MEX CAF ARM SEN TUN NICPNG VUT ATG KHM CPV ISR SAU SYR ECUQAT BRB KWT TTOGTM CHL JPN CUB TGO GMBATG BMU CRI MRT EGY PHL KGZ SLE MAR RWA ARM GMB BDI ROM NZL ZAF IND PAK BLZ MRT BHS GEO GHA KGZ BIH LBN COG MKD MDA NER WSM KAZ MUS SRB ERI CPV FIN BEN MWI GUY MNG PRY BMU BFA NPL NOR RUS CHN TGO TUR GEO GIN JOR URY HND MOZ STP BLR KHM CYP LVA FJIMLI MLT GRC GUY ISL HUN LTU HRV IDN AZE BGR PNG ZWE BHR BLZ SVKPRT MDA BIHEST YEM SVN KIR BRB ALB SWZ UGAZMB CMR GAB SDN KAZ SWZ SEN QAT MDG SYR BHR ETH BHS CHN CUB TZA GHA BLR NIC KEN ESTLVA GER PAK IND MKD DOM HND KWT MLT LTU OMN ITA JPN JAM VNM MUS PRY BGR TTO UKR PAN BOL LKA URY BGDNGA UKR SLV AREIRN JOR LBN ISL THA CRI SVN IDNSWE NLD GERUSA FRA KOR SVK BRA MYS GTM CYP ZAF CZE TUR PER PHL TUN GBR ECU POL HUN RUS ISR ROM BWA BEL MAR MYS CHE ARE AUT EGY POL ESP ARG DNK SRB FIN COL FRA HRV DZA NZL IRL CHL MEX PRT VEN CAN NOR AUS AUT SWE PAN GRC CZE SAU ESP ITA MNG VNM THA CHE IRL KOR BEL GBR DNKNLDCAN KIR STP ERI

1

0

10

4

Log of GDP relative to the US

−4

Data versus simulation

10

6 Red − Simulated Data Blue − Empirical Data

9

Red − Simulated Data Blue − Empirical Data 5

SDN

7

3 2 1 0 −6 10

Import Concentration

8

4

0

10

Intensive Theil Index

Data versus simulation

5

10

(d) Extensive margin of imports

Intensive Theil Index

6

−2

10

Log of GDP relative to the US

(c) Extensive margin of exports

Export Concentration

−2

10

Log of GDP relative to the US

ZAF GHA GTM SLV SEN GAB ZWE GIN KEN TZA BLR CRI

PRT VEN NGA CHN IRN IDN TUR AREIRN COL NGA ARGRUS OMN SAU DJI TGO AUSBRA PER MRT LKA ISR ECU ARE BEN IND BRB ITA MDA YEM SYR MOZ THA KWT VEN MEX POL DOM CMR HND GBR NORSWE ROM SRB EST SYR GRC QAT PNG MLINIC SAU ESP AZE HRV ECU MDG RUS TCD MWI NZL CYP AFG GAB COG JOR CUB KAZ BGR AZE KGZ BWA MUS AUT NLD MYS PAK COL UKR GMB BDIFJIBFA TTO HUN BEL ETH ZMB NOR VNM PAN FIN CHL SVK COG JPN DZA OMN JAM LTU TCD GEO SLE BHR BHS EGY LVA SDN MWI KAZ CRI DZA LBN TTO GERUSA ZWE GIN NER PER CMR CHL CUB FRA EGY PAN GHA ISL PRY KEN AUS PHL MRT MOZ ARM IRL MYS BEN MARQAT VNM BHRBOL ZAF ARG MLT ETH PHL UGA ALB MEX KWT PNG CHE MUS ISL ISR BFA BDI GUY CAF MKD SWZ CAN MNG URY BHS LVA YEM BRA NPL NER BMU IRLIND KOR ZMB KGZ RWA SVN TGO TUN SLE WSM GNB HND JOR KHM NPL GTM ARM UGA TZA CZE LTU KORCAN IDN GEO FJI BGD GMBATG BLR MDG PAK NZL SLV BLZ FIN MLISEN THA JAM HUN BGD FRA UKR BLZ COM ATG SVK KHM NIC DNK TUN URY SWE MDA LBN DOM ESP MAR PRT TUR STP CYP GBR JPN VUT NLD CHN BRB EST SWZ CHE DNK BIH BMUALB MKD LKA BGR BOL GRC CZEPOL PRY BEL GERUSA CAF RWA ERI KIR COM AUT HRVSVN ROM AFG SRB DJI ERI CPV ITA WSM GNB VUT MNG STP CPV

−4

10

−2

10

3

2

1

0 −6 10

0

10

Log of GDP relative to the US

PAN

4

BHS BMU MNG

IND PHL VNM KOR UKR MLT JPN CYP USA MYS PAK ZAF IND KOR KAZ MYS KGZ JPN THA BLRLTUMAR LBN NLDCAN GBR ISR FIN ITA SWZ SEN GRC IRL BEL ESP PRY DNK ITA CHE GER JAM BHR SVK SRB URY HUN AFG MDA DZA MAR OMN PHL NLD GER JOR CHL PAN FRA ARM KIR WSM ATG ESP CZETUR BRA CHE KEN PNG AUT KHM BGD BWA UKR EGY PAK BGR ETH MLT BEL STP TZA MDG TTO SAU URY SVN SWE CHN NZL POL KHM PRT GEO YEM FRA TGO UGA BGR BGD LKA MRT HND GRC LTU BFA ROM IDN MLI GTM MUS MOZ DOM EST BLR CZE DZA ARE GEO CHL GUY TUN GIN NIC SLV VUT DJI FIN HND BEN QAT LVA PER HRV PRY RUS CHN USA GHA AUS CAF CYP BLZ NZL ZAF CRI SVK CUB ECU MKD JOR VNM ISL ZMB LBN AZE NGA NER ARM PRT MUS AUT SDN ZWE KWT BIH TZA GNB BHR MEX TUR HUN ERI CPV BMU TUN BHS BWA SRB COM GMB NOR GIN MNG MKD TTO NGA MDA KEN BOL AZE SYR EGY VUT IRN ROM ATG SEN OMN GBR SLE ETH COL DOM CRI CUB GTM ISRVEN POL RUS COG SWZ LKA MDG MEX BRB MRT ALB ZMB SLV GAB ECU SVN FJI QAT MWI KGZ NPL DNK BFA BRA CMR MLI BEN NIC CAN SDN JAM BOL YEM BRB RWA KWT NER SYR COL MWI EST BLZ TCD PER IRNSAU UGA LVA ALB BDI ARG KAZ GUY DJI MOZ VEN BIHGAB RWA CPV GMB BDI CAF COGPNG WSM STP COM SLE ERI GNBTCD KIR

−4

10

−2

10

0

10

Log of GDP relative to the US

(e) Intensive margin of exports

(f) Intensive margin of imports

Figure 4: Simulated (in red) and empirical observed (in blue) export and import concentration versus GDP across 151 countries. The simulation is based on estimated trade costs form bilateral trade shares including an exported fixed effect.

30

Share of products per number of exporters

Share of products per number of importers

Data versus simulation

Data versus simulation

0.2

0.06 Red − Simulated Data Blue − Empirical Data

0.18

Red − Simulated Data Blue − Empirical Data 0.05

Share of products

Share of products

0.16 0.14 0.12 0.1 0.08 0.06 0.04

0.04

0.03

0.02

0.01

0.02 0 0

10

20

30

40

0 0

50

Number of exporting countries

50

100

150

Number of importing countries

(a) Share of products per exporting country

(b) Share of products per importing country

Figure 5: The simulated (in red) and empirical observed (in blue) share of the number of products traded against the number of trading countries.

31

9

Tables Table 6: Country-Specific Technology and Trade Costs estimates Country USA AFG AGO ALB ARE ARG ARM ATG AUS AUT AZE BDI BEL BEN BFA BGD BGR BHR BHS BIH BLR BLZ BMU BOL BRA BRB BWA CAF CAN CHE CHL CHN CMR COG COL COM CPV CRI CUB CYP CZE DJI DNK DOM DZA ECU EGY ERI ESP EST ETH FIN FJI FRA GAB GBR GEO

Exporter FE

Standard error

Precent cost

Si

Standard error

(λUS /λi )θ

6.59 -2.81 -0.67 -4.47 3.54 2.58 -3.63 -1.80 3.79 1.92 -1.28 -3.73 5.05 -3.17 -3.63 0.99 0.89 0.71 -0.54 -4.41 -0.26 -0.76 -2.67 -2.19 3.97 -2.01 -3.49 -4.15 4.46 3.72 2.35 6.48 -1.24 0.61 0.94 -4.58 -4.43 0.35 -1.59 0.07 1.91 -3.81 2.75 -1.38 -0.21 0.80 0.84 -4.67 4.28 0.20 -2.17 2.81 -2.66 4.73 -1.36 4.89 -1.92

0.2 0.23 0.23 0.22 0.2 0.2 0.22 0.23 0.2 0.2 0.22 0.23 0.2 0.22 0.22 0.2 0.2 0.21 0.23 0.22 0.21 0.23 0.25 0.22 0.2 0.22 0.24 0.23 0.2 0.2 0.2 0.2 0.21 0.22 0.2 0.27 0.24 0.2 0.21 0.2 0.2 0.25 0.2 0.21 0.2 0.2 0.2 0.27 0.2 0.2 0.2 0.2 0.24 0.2 0.21 0.2 0.21

-54.7 40.0 8.4 71.0 -34.6 -26.7 54.6 24.1 -36.5 -20.6 16.6 56.5 -45.4 46.3 54.6 -11.2 -10.1 -8.2 6.6 69.8 3.1 9.5 37.8 30.1 -37.9 27.3 52.0 64.6 -41.5 -36.0 -24.6 -54.0 16.1 -7.1 -10.6 73.2 70.2 -4.1 21.0 -0.8 -20.5 57.9 -28.1 18.1 2.5 -9.1 -9.6 75.1 -40.1 -2.3 29.8 -28.6 37.6 -43.3 17.8 -44.4 25.9

0.13 0.02 -1.07 1.17 -1.44 0.99 0.44 -1.35 0.26 0.66 0.21 0.44 -1.11 0.54 0.74 -0.2 -0.14 -1.15 -0.48 1.5 0.36 -1.4 -0.06 0.77 0.72 -0.67 1 0.55 -0.15 -0.12 -0.06 -0.3 0.33 -1.9 0.53 -0.3 -0.13 0.05 1.21 -0.35 0.04 -0.5 0.43 0.44 0.51 0.19 0.28 -0.33 -0.19 -0.54 0.47 0.24 -0.02 0.26 -0.04 -0.04 0.23

0.14 0.17 0.16 0.15 0.14 0.14 0.15 0.17 0.14 0.14 0.15 0.16 0.14 0.15 0.14 0.14 0.14 0.15 0.17 0.15 0.15 0.16 0.18 0.15 0.14 0.15 0.17 0.16 0.14 0.14 0.14 0.14 0.15 0.16 0.14 0.19 0.16 0.14 0.14 0.14 0.14 0.18 0.14 0.14 0.14 0.14 0.14 0.2 0.14 0.14 0.14 0.14 0.16 0.14 0.15 0.14 0.15

1 39.69 4.75 6.39 1.33 1.6 7.7 3.62 0.88 0.92 4.41 15.72 1.09 10.78 12.61 9.3 4.19 2.48 2.1 5.52 3.59 3.96 1.39 4.47 1.83 4.35 2.11 10.39 0.91 0.71 1.98 3.96 5.25 5.77 3.11 14.29 13.89 2.45 3.53 2.91 1.64 12.47 0.77 4.28 2.41 3.36 6.53 44.91 1.46 2.26 19.86 0.77 4.16 0.9 1.81 1.12 7.36

32

Table 7: Country-Specific Technology and Trade Costs estimates - cont. Country GER GHA GIN GMB GNB GRC GTM GUY HND HRV HTI HUN IDN IND IRL IRN ISL ISR ITA JAM JOR JPN KAZ KEN KGZ KHM KOR KWT LAO LBN LBY LKA LTU LVA MAR MDA MDG MEX MKD MLI MLT MNG MOZ MRT MUS MWI MYS NER NGA NIC NLD NOR NPL NZL OMN

Exporter FE

Standard error

Precent cost

Si

Standard error

(λUS /λi )θ

5.46 0.44 -1.81 -2.82 -4.39 1.26 -0.03 -1.11 -0.05 -0.28 -4.61 1.03 4.66 4.58 3.75 0.17 -0.67 0.30 4.34 -1.33 -0.24 5.91 0.46 0.13 -2.99 -2.58 6.03 -0.88 -4.48 -1.01 -1.88 0.89 -0.15 -1.04 1.14 -2.75 -1.78 2.39 -2.53 -2.82 0.54 -3.05 -1.44 -1.71 1.05 -1.59 6.12 -3.76 1.20 -1.58 5.43 1.85 -4.58 3.05 -1.06

0.2 0.2 0.21 0.24 0.29 0.2 0.21 0.22 0.21 0.2 0.25 0.2 0.2 0.2 0.2 0.21 0.21 0.2 0.2 0.21 0.21 0.2 0.21 0.2 0.24 0.22 0.2 0.2 0.28 0.2 0.23 0.2 0.21 0.21 0.2 0.22 0.21 0.2 0.22 0.21 0.2 0.25 0.21 0.22 0.2 0.21 0.2 0.21 0.2 0.22 0.2 0.2 0.23 0.2 0.21

-48.0 -5.1 24.3 40.2 69.4 -14.0 0.4 14.3 0.6 3.4 73.9 -11.6 -42.9 -42.3 -36.2 -2.1 8.4 -3.5 -40.6 17.3 3.0 -50.8 -5.4 -1.6 43.2 36.3 -51.5 11.1 71.2 12.8 25.4 -10.1 1.8 13.3 -12.8 39.1 23.8 -24.9 35.5 40.2 -6.2 44.3 18.8 22.8 -11.9 21.0 -52.0 57.0 -13.4 20.8 -47.9 -19.9 73.2 -30.6 13.6

-0.02 -0.94 -0.11 -0.93 -0.32 0.38 0.28 -1.17 -0.79 0.18 0.37 0.35 -0.6 -0.22 -0.71 1.11 -0.02 1.21 0.34 -0.25 -0.55 0.51 0.54 -0.15 0.29 0.91 -0.66 0.43 1.3 0.08 0.86 -0.03 0.18 0.59 -0.27 0.4 -0.08 0.5 0.41 0.56 -1.1 -0.17 -0.19 -0.85 -1.78 0.38 -2.07 0.38 -0.53 -0.16 -0.99 0.67 1.27 -0.11 0.6

0.14 0.14 0.15 0.17 0.21 0.14 0.14 0.15 0.14 0.14 0.19 0.14 0.14 0.14 0.14 0.15 0.15 0.14 0.14 0.15 0.15 0.14 0.15 0.14 0.17 0.15 0.14 0.14 0.22 0.14 0.16 0.14 0.15 0.15 0.14 0.15 0.14 0.14 0.14 0.15 0.14 0.18 0.14 0.16 0.14 0.15 0.14 0.15 0.14 0.15 0.14 0.14 0.16 0.14 0.15

0.82 10.44 6.55 15.37 13.64 2.16 4.84 5.58 6.25 2.77 15.29 1.76 4.29 7.91 0.88 2.6 0.94 1.18 1.02 4.29 5.52 0.67 2.84 10.87 7.78 6.32 1.51 0.93 8.12 5.14 1.24 5.71 2.58 2.23 5.03 6.25 11.6 2.2 3.69 8.45 2.05 5.87 13.62 6.6 4.48 11.96 2.84 10.93 6.41 7.44 0.99 0.58 13.06 1.21 1.08

33

Table 8: Country-Specific Technology and Trade Costs estimates - cont. Country PAK PAN PER PHL PLW PNG POL PRT PRY QAT ROM RUS RWA SAU SDN SEN SLV SRB SVK SVN SWE SWZ SYR TCD TGO THA TTO TUN TUR TZA UGA UKR URY UZB VEN VNM VUT YEM ZAF ZAR ZMB ZWE

Exporter FE

Standard error

Precent cost

Si

Standard error

(λUS /λi )θ

1.26 1.93 0.77 1.86 -4.21 -1.40 1.59 2.37 -0.43 -1.36 1.20 3.13 -3.54 1.95 -2.47 -1.04 -1.59 -1.49 0.42 -0.06 3.26 1.19 -0.98 -3.83 -1.49 5.71 0.17 0.13 2.35 -0.62 -1.88 2.03 0.81 -2.16 -0.05 3.19 -2.44 -2.58 4.64 -2.04 -1.64 -0.33

0.2 0.21 0.2 0.2 0.4 0.24 0.2 0.2 0.22 0.21 0.2 0.2 0.23 0.2 0.2 0.21 0.21 0.21 0.2 0.2 0.2 0.23 0.21 0.26 0.21 0.2 0.21 0.2 0.2 0.2 0.2 0.2 0.21 0.25 0.21 0.2 0.29 0.22 0.2 0.24 0.21 0.21

-14.1 -20.6 -8.9 -20.0 65.7 18.2 -17.4 -24.7 5.3 17.7 -13.5 -31.3 52.9 -20.9 34.5 13.3 20.9 19.5 -4.9 0.7 -32.3 -13.3 12.5 58.3 19.5 -49.6 -2.0 -1.5 -24.6 7.8 25.3 -21.7 -9.2 29.7 0.6 -31.8 34.0 36.3 -42.7 27.7 21.8 4.1

0.64 -1.31 0.68 -0.13 -1.18 -0.08 0.59 -0.39 -0.17 0.83 0.42 0.51 0.09 0.15 0.72 -0.45 0.28 0.46 0.27 -0.04 0.45 -2.17 0.36 -0.07 -0.45 -1.45 -0.63 -0.21 0.6 -0.23 0.43 0.23 0.08 0.74 1.17 -0.79 -0.74 0.5 -0.86 -0.19 0.18 0.13

0.14 0.15 0.14 0.14 0.33 0.16 0.14 0.14 0.16 0.15 0.14 0.14 0.15 0.14 0.14 0.15 0.15 0.15 0.14 0.14 0.14 0.18 0.15 0.18 0.15 0.14 0.15 0.14 0.14 0.14 0.14 0.14 0.15 0.19 0.14 0.14 0.21 0.15 0.14 0.18 0.15 0.15

7.39 5.27 2.78 4.69 2.99 3.96 2.14 1.75 5.5 0.5 3.14 2.36 17.61 1.27 4.92 11.43 4.95 4.14 1.89 1.56 0.76 6.63 4.77 11.19 11.05 3.84 2.01 3.58 2.51 15 13.31 4.47 2.27 7.43 1.65 7.29 5.14 5.57 2.86 19.49 6.38 6.56

34