Journal Issue
Share
Article

Import-Reducing Effect of Trade Barriers

Author(s):
Qing Wang
Published Date:
December 2001
Share
  • ShareShare
Show Summary Details

I. Introduction

Governments of almost all the countries in the world routinely intervene in trade across borders, through the use of tariffs, quotas, and other non-tariff barriers (NTBs). Reductions in these trade restrictions have been regularly achieved through multilateral trade negotiations or preferential trade arrangements. Measuring the effects of general trade policy and the economic implications of particular trade reforms is an intriguing task facing both policy-makers and economists. Tariffs have relatively clear primary effects on product prices, but their secondary effects on employment, earnings, profits and consumer welfare are far from straightforward. Non-tariff barriers, on the other hand, have quite unclear effects on product prices, and largely unknown secondary effects. As tariff levels have fallen over the years, non-tariff barriers (NTBs) may have become the instruments of choice for protection.

We carry out a comprehensive empirical investigation to ascertain specifically the import-reducing effect of trade protection barriers. We use a monopolistic competition trade model and 1994 cross-country data on trade barriers, trade flows, and production to estimate the import- reducing effect of trade barriers including both tariffs and NTBs. We used the disaggregated cross-country, cross-industry data on manufactured goods and, unlike previous studies, our sample covers a broad range of countries—more than 70 in total—including countries from the most developed ones like those in the G-7 group to the least developed one like Bangladesh. In estimating the equation, we specify an empirical model that captures the stylized facts well and helps generate more sensible and efficient estimates. Finally, our econometric framework is designed to control for the simultaneous determination of flows, trade barriers and production. The effect of trade barriers on trade flows is the focus of this study. We obtain reasonable and robust results. We find that both tariff and NTBs are quite significant in restricting imports.

The paper has six sections. The next section we present some general evidence on global trade protection. Section III is the literature review. In section IV, we specify an empirical model based on the monopolistic competition model of international trade. In section V, we discuss the estimation and the results. The last section summarizes the paper.

II. Global Trade Protections: An Overview

In this section, we present an overview of global trade protections based on a Bilateral Trade Protection Database (BTPD)2 from two perspectives: country and industry

A. Global Trade Protections: A Country Perspective

The bilateral trade protection matrix is the core of the BTPD; and it contains very rich information for comparing trade protection regimes of different countries.3 In the matrix for a specific commodity group, the elements down a column are the import tariff rates or NTB coverage ratios imposed by a home country against its trading partners, and the elements across a row are the tariff rates or NTB coverage ratios imposed upon the home country′s exports by each of its trading partners. Taking the average of each column, we can get, for a specific category of commodity, the tariff rates or NTB coverage ratios imposed by each home country. Similarly, taking the average of each row, we can get the average tariff rates or NTBs coverage ratios faced by each of the home countries. Therefore, we can figure out which country is protected or being protected against by what kind of trade protection measures. Furthermore, by comparing the variation of numbers across each column or row, we can get an idea of how discriminatory a country′s import regime is or whether a country is being treated equally by its trading partners.

Table 1 presents the column means and other statistical indicators for the bilateral trade protection matrix at the highest aggregation level, i.e. covering all commodity categories. For each country, we list the average tariff rate, NTBs coverage ratio, and their corresponding coefficient of variation (COV), which we use as an indicator of the degree of discrimination of that country′s import regime. The table is sorted by the average tariff rate. A country is more protected, the higher its tariff rate or NTB coverage ratio; and a country′s import regime is deemed more discriminatory, the larger its COV. We computed the Spearman rank correlation of per capita GNP of each country with its average tariff rate, NTB coverage ratio and the coefficient of variation, respectively.4 The results are listed in Table 2.5

Table 1.Trade Protection Imposed by Each Importing Country in 1994
CountryTariff

AVE (%)
COVNTBs

AVE (%)
COV
Bangladesh45.100.902.872.86
Algeria21.850.8415.601.99
Tunisia21.720.576.842.66
India19.090.7210.581.23
Philippines18.720.630.000.00
Kenya18.650.890.000.00
Egypt16.590.860.000.00
Jamaica14.191.0128.301.02
Mauritius13.251.010.000.00
Sri Lanka12.630.870.024.66
Poland12.610.530.000.00
Madagascar12.330.850.000.00
Hungary12.091.130.000.00
China12.000.812.212.49
Cameroon11.500.790.000.00
Cote d’lvoire11.320.870.000.00
Mexico11.260.6717.111.39
Peru11.160.535.882.88
Argentina10.510.575.491.89
Congo10.481.120.000.00
Ecuador10.110.730.000.00
Venezuela10.090.6111.791.87
Gabon9.790.600.000.00
Malawi9.781.260.000.00
Nicaragua9.520.954.454.36
Bolivia9.400.970.000.00
Thailand9.140.6417.221.70
Chile9.010.433.473.96
Costa Rica8.871.200.000.00
Brazil8.721.0811.731.63
Dominican Republic8.421.120.000.00
Central African Republic8.311.230.276.52
Chad8.251.300.000.00
Trinidad & Tobago8.150.820.000.00
Uruguay7.900.782.012.53
Saudi Arabia7.710.720.046.50
Korea, Republic of7.480.690.164.07
Guatemala7.210.790.000.00
Turkey7.131.000.622.08
Honduras6.770.850.000.00
Morocco6.191.592.562.48
Indonesia6.040.730.000.00
South Africa5.810.780.000.00
Paraguay5.741.000.000.00
El Salvador5.671.0813.211.35
Colombia5.231.080.000.00
Malaysia5.191.365.462.58
Canada5.160.9313.161.56
Czechoslovakia4.790.630.365.72
United States4.671.3619.761.01
European Union4.450.6022.161.07
Norway3.871.196.552.04
Iceland3.791.330.713.76
New Zealand3.620.910.894.85
Australia3.530.770.903.15
Oman3.431.432.852.62
Israel3.281.270.000.00
Japan2.810.712.711.42
Singapore0.000.003.163.83
Switzerland0.000.000.000.00
Hong Kong0.000.000.000.00
Table 2:Rank Correlation Between Per Capita GNP and Import Regime Indicators Number of Observations: 61
Rate/RatioDegree of Discrimination (COV)
Tariff-0.71

(0.00)
-0.24

(0.06)
NTBs0.27

(0.04)
0.27

(0.03)

The correlation coefficients show that, in relative terms, the richer a country, the lower is its average tariff rate, and the less discriminatory is its tariff structure. However, when it comes to the NTBs, the situation is exactly the opposite: the richer a country, the higher is its average NTB coverage ratio, and the more discriminatory is its NTB structure. Thus efforts to discover the relationship between the level of a country′s per capita income and its import protection regime yield mixed results. Two explanations can be put forward: first, implementing NTBs usually involves high administrative cost and thus poor countries tend to resort to tariffs both as means of protections and as source of fiscal revenues in government finance; second, countries, especially developed ones, use NTBs to offset the reduced tariffs negotiated in the various GATT rounds.

In Table 3, we show the same statistics as in Table 1 except that this time they are computed for the rows so that we can compare protections faced by exports out of different countries. Again, the table is sorted by the average of tariff rates, with the countries at the top being subject to the most protections from the rest of world. The rank correlation coefficients between per capita GNP and trade protection indicators are as follows:

Table 3.Trade Protection Faced by Each Exporting Country in 1994
CountryTariff

AVE(%)
COVNTBs

AVE (%)
COV
Mauritius10.671.499.112.45
China10.311.213.173.75
Madagascar10.201.681.653.78
Bolivia10.101.476.573.12
Ecuador9.881.356.593.13
Hungary9.841.386.712.11
Sri Lanka9.711.455.242.85
Morocco9.691.524.542.17
Guatemala9.401.707.233.03
Iceland9.221.501.023.71
Hong Kong9.191.565.521.93
Cameroon9.101.744.784.06
Argentina9.031.456.532.45
New Zealand8.701.395.912.92
Turkey8.661.625.522.24
Colombia8.561.294.963.75
Trinidad & Tobago8.481.712.943.87
Korea RP8.461.605.231.85
Singapore8.411.623.192.04
Czechoslovakia8.311.253.052.32
Oman8.211.292.672.83
Jamaica8.181.533.941.93
South Africa8.161.562.442.41
Bangladesh8.141.258.242.69
Chile8.141.293.054.27
Egypt8.021.375.702.51
Japan7.961.544.102.29
Malawi7.931.500.974.96
Uruguay7.901.455.702.92
Tunisia7.861.265.312.86
Malawi7.601.593.812.49
Peru7.571.402.692.69
Thailand7.271.644.072.46
Gabon7.251.345.222.56
El Salvador7.131.369.592.58
Mexico7.061.302.862.97
Australia7.061.542.452.57
Cote d’lvoire6.941.471.354.80
Poland6.931.522.932.68
Philippines6.841.605.652.62
Canada6.791.533.842.90
Brazil6.781.342.332.46
Switzerland6.711.692.083.18
European Union6.691.604.151.91
Venezuela6.681.373.974.23
Nicaragua6.521.272.444.33
Honduras6.501.363.634.03
Norway6.411.682.122.51
Kenya6.391.543.773.83
Israel6.301.823.743.00
United States6.081.462.782.34
India5.791.225.202.30
Dominican Republic5.691.353.133.38
Paraguay5.371.284.593.28
Saudi Arabia5.241.222.425.37
Algeria5.041.842.215.93
Indonesia5.011.273.943.66
Chad3.951.821.107.24
Congo3.721.700.237.65
Central African Republic3.692.340.535.31
Costa Rica3.381.424.114.43

Table 4 indicates that, in relative terms and in terms of tariff rates, exports coming out of richer countries tend to be more protected but less discriminated against by the rest of the world. While, in terms of NTBs, the exact opposite happens: it is the poor countries that are being targeted by both protection and discrimination. Still, there is no clear-cut answer on the relationship between countries’ income levels and the degree of protection and discrimination. It all depends on which of the two indicators is being used - tariff rate or NTB coverage ratio.

Table 4:Rank Correlation Between Per Capita GNP and Protection Faced by Exporting Countries Number of Observations: 61
Rate/RatioDegree of Being Discriminated (COV)
Tariff0.17

(0.20)
-0.49

(0.00)
NTBs-0.32

(0.01)
-0.41

(0.00)

B. Global Trade Protections: An Industry Perspective

Bilateral trade protection matrices similar to those shown in appendix II exist for individual commodity groups. The number in the lower right corner of each matrix can be interpreted as the average tariff rate or NTB coverage ratio imposed on that specific commodity group on the world market. In the current version of BTPD, the commodities have been aggregated according to a 120-trade-sector scheme.6 Taking the number in the lower right corner of each of the 120 matrices, we get the average tariff rates and NTB coverage ratios imposed upon the commodities from those 120 sectors, which help to examine the global trade protections from an industry perspective. Tables 5 and 6 list the twenty most and least protected trade sectors by tariffs and NTBs, respectively. Four rank correlation coefficients are shown in Tables 7 and 8: the first is between the tariff rates and NTBs coverage ratios, the second between their respective coefficients of variation, and the last two are those between industries’ world trade shares and tariff rates/NTBs coverage ratios. The findings are: in relative terms, if an industry has high protection and discrimination by one type of trade barrier, it tends to face high protection and discrimination by the other type as well. Also a high level of protections appears to be given to large trade sectors.

Table 5.Trade Protection Across Industries by Tariffs in 1994
SectorShare of World Trade

Twenty Most Protected

Trade Sectors
AVE (%)COV
36 Wearing apparel3.4061.11.45
35 Floor coverings0.2058.41.58
33 Cotton fabric0.4248.81.45
34 Other textile products1.6440.71.50
32 Yarns and threads0.9033.81.58
20 Preserved fruits, vegetables0.6727.42.58
27 Food products n.e.c.0.8724.61.74
6 Cotton0.1824.51.88
7 Wool0.1023.02.46
12 Coal0.3521.13.12
25 Sugar0.2720.91.34
107 Motorcycles and bicycles0.1518.01.50
29 Alcoholic beverage0.6015.41.71
62 Cement0.0915.42.11
1 Unmilled cereals0.6515.11.86
39 Footwear0.8215.11.69
67 Aluminum0.8415.02.06
22 Vegetable, animal oils, fats0.7014.01.85
23 Grain mill products0.2313.91.59
48 Fertilizers0.5213.01.96
Twenty Least Protected

Sectors
113 Watches and clocks0.420.43.97
46 Printing, publishing0.510.32.41
110 Other transport equipment0.190.32.53
115 Musical instruments0.110.33.50
8 Other natural fibers0.010.22.86
42 Furniture and fixtures0.910.23.60
44 Newsprint0.220.24.03
50 Paints, varnishes, lacquers0.240.23.53
79 Other power machinery0.060.24.03
13 Non-ferrous metal ore0.400.14.03
41 Other wood products0.410.12.59
76 Boilers and turbines0.480.14.03
77 Aircraft engines0.300.14.03
11 Iron ore0.150.00.00
17 Electrical energy0.160.00.00
57 Product of coal0.050.00.00
68 Nickel0.080.04.03
104 Ships for military purpose0.000.00.00
109 Aircraft1.560.04.03
118 Works of art0.140.00.00
Table 6.Trade Protection Across Industries by NTBs in 1994
SectorShare of World Trade

Twenty Most Protected

Trade Sectors
AVE (%)COV
29 Alcoholic beverage0.6033.21.26
31 Tobacco products0.4830.11.10
36 Wearing apparel3.4024.50.66
106 Motor vehicles6.2124.50.88
35 Floor coverings0.2024.50.77
39 Footwear0.8224.30.68
30 Non-alcoholic beverage0.0924.20.77
24 Bakery products0.2023.70.75
20 Preserved fruits, vegetables0.6723.70.86
38 Leather products0.4023.20.73
18 Meat1.0022.11.47
42 Furniture and fixtures0.9121.90.73
2 Fresh fruits, vegetables0.6521.61.83
52 Soap, other toilet preparations0.6021.40.80
95 Household electrical appliances0.6721.00.80
33 Cotton fabric0.4220.20.88
25 Sugar0.2720.01.07
119 Manufactured goods nee1.0219.90.79
26 Cocoa, chocolate, etc.0.3919.90.74
63 Ceramics0.3319.80.77
Twenty Least Protected Trade Sectors
16 Non-metallic ore0.267.41.28
78 Internal combustion engines2.096.91.13
51 Drugs and medicines1.356.71.24
8 Other natural fibers0.016.51.57
44 Newsprint0.226.41.16
55 Fuel oils0.955.91.02
109 Aircraft1.565.81.19
120 Scraps, used, unclassified3.755.81.46
43Pulp and waste paper0.545.51.78
6 Cotton0.185.41.20
5 Silk0.015.12.98
13 Non-ferrous metal ore0.404.91.25
57 Product of coal0.054.61.58
12 Coal0.354.41.49
15 Natural gas0.614.41.19
14 Crude petroleum3.623.82.02
7 Wool0.103.81.48
11 Iron ore0.151.82.16
17 Electrical energy0.160.24.39
104 Ships for military purpose0.000.00.00
Table 7:Rank Correlation between Tariff Rate and NTBs Coverage Ratio Across Industries Number of Observations: 120
Rate/RatioDegree of Discrimination Faced
0.53

(0.00)
0.19

(0.04)
Table 8:Rank Correlation between Industry′s Trade Share and Tariff Rate/NTB Coverage Ratio Across Industries Number of Observations: 120
TariffsNTBs
0.23

(0.01)
0.11

(0.24)

C. Summary

Despite many years of multilateral trade negotiations and unilateral cuts in trade protection measures, the level of overall trade protections was still high as of 1994. It also varies considerably across countries and industries. Without delving into commodity details, we examined the overall trade protection regime from two perspectives: country and industry. The main findings are:

  • Countries of lower per capita income tend to impose higher tariffs and more discriminatory tariff structures against imports from their trading partners.
  • Counties of higher income tend to impose higher NTBs and more discriminatory NTB structures against imports from their trading partners.
  • Exports coming out of higher income countries tend to face higher protection but less discrimination in the form of tariffs by the rest of the world.
  • Exports coming out of lower income countries tend to face higher protection and higher discrimination in the form of NTBs by the rest of the world.
  • If an industry has a high level protection and discrimination by one type of trade barrier, it tends to face a high level of protection and discrimination by the other type as well.
  • A high level of protection tends to be given to large trade sectors.

It should be pointed out, however, that the statistical results should be interpreted with some caution in view of the potential data problems, especially those related to NTBs, as mentioned above.

III. An Overview of the Literature

This study is built upon a relatively small literature. The topic has long intrigued the profession. However, the substantial data requirement usually involved in this kind of study is so demanding that comprehensive studies did not appear until late 1980′s, when progress in information technology made the task less onerous. In Table 9, we briefly summarize the key features of several representative studies closest to this one.

Table 9.Summary of Representative Studies
Author (date)Theoretical ModelData Sample CoverageEmpirical ModelMajor Findings
Learner

(1988a)
Generalized factor proportions modelFourteen OECD countries; nineteen commodity categories in 1978.Regress domestic output on capital, labor, tariffs and commodity dummies; also do the same regression for each industry separately.Measured trade barriers have a clear effect on the composition of output; Credible estimates of trade barriers may require simultaneous treatment for the trade barriers.
Learner

(1990)
Generalized factor proportions modelFourteen OECD countries; ten commodity categories in 1983.Regress imports on tariffs and NTB′s with importer and commodity dummies; Baysian method is used to help to generate sensible coefficient estimates which vary across importers and commodities.The model that accounts for differences in commodities and countries shows that NTB′s effects are difficult to detect, whereas the effects of tariffs are more substantial.
Lawrence

(1987)
Monopolistic Competition ModelThirteen OECD countries for the period of 1970-1983; Twenty-one manufacturing sectors.Regress import penetration ratio on world production share, distance and country dummy.Japan has an unusually low volume of imports in manufactures.
Harrigan

(1993)
Monopolistic Competition ModelThirteen OECD countries; Twenty-eight manufacturing sectors in 1983.Regress the share of bilateral imports in aggregate spending on exporting countries’ output, bilateral trade barriers and transport cost.In 1983, imports were not reduced very as much by NTBs; transport cost and tariffs had large negative effects on imports.
Ray

(1981b)
Generalized factor proportions modelThe U.S. data in 1970 at 4-digit SIC levelImports and trade barriers equation are simultaneously estimated; factor intensity and NTBs are included in the imports equation.The NTBs have no apparent concurrent impact in the structure of imports across the manufacturing sectors.
Trefler

(1993)
Generalized factor proportions modelThe U.S. data in 1983 at BEA Input-Output classification level.Import penetration ratio equation and NTBs equation are simultaneously estimated in a Tobit model; factor shares and NTBs are included in the import penetration ratio equation.When NTB are modeled endogenously, their restrictive impact on imports is much larger than otherwise.
Lee and

Swagel

(1995)
Monopolistic Competition ModelForty countries; twenty-eight manufacturing sector in 1988.Import penetration ratio and NTBs are simultaneously determined; output share, distance and trade barriers are included in import equation.The structure of NTBs across countries and industries can be explained by sectoral conditions, which is consistent with the political-economy explanations of trade protection.
Harrigan

(1996)
Monopolistic Competition ModelTwenty OECD countries; Twenty-eight manufacturing sectors in 1985.Regress bilateral imports on exporting countries’ output and commodity, exporter, and importer dummies.The openness between Japan, the US and the EU countries differs significantly.

A. The Theoretical Foundation

When studying the effects of trade barriers on trade flows, a natural starting point is a theoretical model describing what trade patterns would be in absence of trade barriers. The usual practice in the literature is then to modify the original model by adding variables related to trade barriers. Therefore, how one models the effects of trade barriers on trade flows, to a large extent, depends on one′s choice of trade determination model.

The Ricardian model is the cornerstone of international trade theory, and it attributes comparative advantage entirely to differences in labor requirements of production. Its most important implication is that there is complete specialization in equilibrium under free trade. If relative labor costs of production could be observed, a simple regression of trade on these labor costs would suffice to test the theory and then make inferences on the trade pattern. However, observing relative labor requirements has at least two almost insurmountable obstacles: first, relative labor requirements are just as difficult to observe as relative autarky prices; second, comparing labor requirements in all countries of the world poses enormous practical difficulties. It is no surprise that we do not find, in the literature, any study of the trade barriers effects being made within the Ricardian framework.

The Heckscher-Ohlin (H-O) model has occupied a central place in trade theory for much of the post-war period. It says that countries will tend to export those goods which use relatively intensively the relatively abundant factors of production. Put differently, countries will tend to export the services of their abundant factors, embodied as factor content in the goods they trade. The H-O model is generally regarded as superior to the Ricardian model because it offers an intellectually more sophisticated explanation of trade. Leontief′s (1954) seminal application of the H-O model of factor proportions stimulated a large body of research that continues today in an effort to more rigorously test the theory. Nonetheless, it remains true that no unambiguously correct and conclusive test has been formulated and applied. Realizing that a full understanding of trade pattern seems to require some departure from the H-O assumptions, economists resort to a “generalized factor proportions model” as a theoretical basis for empirical work and adopt a new strategy which is, as characterized by Learner and Levinsohn (1994), “estimate, don’t test!”. The generalized factor proportion model allows for factors beyond just capital and labor. As a general approach to understanding trade, the factor proportions theory has stood remarkably well to the empirical scrutiny of commodity composition of trade.

Within the framework of the generalized factor proportions model, Edward E. Learner is the leading figure in exploring the empirical issues of the effect of trade barriers. In a series of studies which started with Learner (1974) and culminated in Learner (1990), Learner gives this issue a more persistent and comprehensive treatment than anyone else does in the literature. His contributions range from building the theoretical foundation for empirical models suitable for cross-section estimation, and discussing the data problems caused by the dimension of the data sets to applying particular econometric techniques to the estimation of the effects of trade barriers. Learner (1990) estimates the effects of trade barriers based on cross-country as well as cross-commodity variability of barriers and imports. Although it has been hailed as the best attempt at the relevant issues, Learner′s model is not free from the common weakness of the factor proportion model, namely its inability to address the bilateral patterns and gross volume of trade. In a strict sense, the model derived by Learner is for the determination of net trade flows. However, when it comes to the empirical study, the trade barriers’ impact on gross import is what should be under investigation. In a word, Learner′s empirical model specification is supported only by a loose theoretical justification.

Beginning in the late 1970′s, an initially small group of theorists began to develop a different approach to international trade, which later became known as the New Trade Theory. This line of work was, in part, motivated by the observation that much international trade appears to be in goods that are quite similar. The core of the new trade theory is the so-called monopolistic competition trade model as summarized in Helpman and Krugman (1985). Two major assumptions distinguish the monopolistic competition trade model from various traditional factor proportions models. The first is that there are internal economies of scale at the level of the firm; the second is that there is an aggregate demand for variety in goods. This demand for variety can come from variety-loving and/or heterogeneous consumers, or from final goods production processes that make use of differentiated intermediate products (Ethier, 1988). Because of the interaction between scale economies and demand for variety, in equilibrium each firm in an industry produces a single differentiated product.7

One of the major implications of the model is that the volume of trade is much larger than it would be if differences in international factor endowments were the only cause of trade. The model gives predictions about the equilibrium volume of trade:8

where Mjn is the total gross import of good n by country j; Mijn is the gross import of good n by country j from country i, Sj is country j′s share of total world spending; Yjn is output of industry n in country j: yn is the total world production of good n. Equation (1) and (2) provide a basic framework to estimate trade pattern and gross volume of trade. This frictionless model predicts that a country′s import of good n is proportional to the amount of good n produced outside that country.

Lawrence (1987) was the first to use the monopolistic competition model to specify predicted volumes of trade and to use disaggregated data on production and trade flows to determine which countries and industries differ significantly from the model prediction. Lawrence′s conclusion that Japan has an unusually low volume of imports attributable to the existence of trade barriers has attracted considerable attention in the literature. Harrigan (1993) investigates import-reducing effects of trade barriers in OECD countries for the year of 1983. His model is based on Equation (2). Rather than attributing any deviation of actual imports from predicted imports to the effects of protections as did in Lawrence (1987), Harrigan explicitly adds measures of trade barriers to the original model so that he can directly examine the impact of trade barriers on trade flows. His finding is that in 1983, tariffs and transport costs were a more substantial barrier to trade in manufactures between developed countries than were NTBs. In a follow-up study, Harrigan (1996) addressed a similar issue using a slightly different version of his 1993 model. Lee and Swagel (1995) was another recent study of trade barriers within monopolistic competition framework. Theirs has so far been the most comprehensive study, in which they investigate the trade flows at 3-digit ISIC just as Harrigan (1993) did, but the number of countries that they covered was almost triple that of Harrigan (1993). However, their focus is more on the political economy determinants of NTBs than on the impact of protections (both tariff and non-tariff measures) on disaggregated trade flows.

B. Data Support and Empirical Model Specification

Strong data support is critical to reliable empirical estimation. It is more so for the estimation of the effect of trade barriers, which usually involves data from multiple sources including trade flows, trade protection measures, production, factor endowments and so on. All the investigations in the literature face the same fundamental problem, which is caused by the dimension of those data sets. In terms of time dimension, it is usually easy to obtain time series data on trade flows, but, for trade protection data, it is extremely difficult. In terms of the number of countries, for a long time, only OECD countries published reliable trade protection data. It is not until recently that a more comprehensive data set has been made available to the public, namely the Trade Analysis Information System (TRAINS). In terms of the number of commodities, the situation is much better --- we have many commodities. It is, accordingly, essential to pool across countries and/or commodities to estimate the effects of trade barriers.

Learner (1990) estimated a model similar to:

where i is importer, n is commodity, M is imports, TAR is the tariff rate and NTB is NTB coverage ratio, a and b are constants. He used one-year trade protection and import data from 14 OECD countries with commodities disaggregated into 10 categories. In order to capture trade barriers’ import-reducing effects that vary by importers and commodities, he pooled across both countries and commodities and controls for their difference by dummy variables. One contribution of his paper was that he used a Baysian estimation method to overcome the lack of degrees of freedom usually required by such kind of dummy variable model. However, Learner did not hesitate to mention his discomforting with the fact that he had to resort to cross-commodity variation to carry out the estimation, because it was like estimating a demand equation by comparing demand for different commodities.

Lawrence (1987), for the first time in the literature, used a monopolistic competition model to investigate the issue of openness. His model was a variant of Equation (1),

where DU is domestic use (production + imports - exports), T is transaction cost. Instead of modeling trade barriers explicitly, he attributed deviations from model predictions to trade protection measures. Free from the dimensionality constraint imposed by the paucity of trade protection data, he was able to run regressions by industry based on a panel of 13 OECD countries for the period of 1970-83. Lawrence avoided Learner′s dilemma of pooling across commodities, but he ignored the simultaneity problem between trade and production, which is a potentially important issue since the monopolistic model suggests that imports and production are jointly determined. Also, there may be many other sources of errors in the model, not just trade barriers.

Harrigan′s (1993) data set was the same as Learner′s (1990), but since he adopted a distinct theoretical framework and thus could specify an empirical model which allowed him to take advantage of another dimension of the original data set. He slightly transformed equation (2):

where, j is importing country, i is exporting country, n is commodity, π is aggregate spending. Equation (5) says that bilateral imports are proportional to each partner country′s output. This implication allows him to exploit the bilateral variation in the trade pattern within a certain category of commodity and, therefore, nearly increase the number of observations by a factor of its original size. (i.e. from the number of importing countries to the product of the number of importing countries and trading partners.) For each of the 28 sectors at 3-digit ISIC level, he estimated an equation as follows:

Harrigan also addressed the simultaneity problem between imports and output. In his study, economy-wide factor endowments were used as an instrument for production. He found that in 1983, gross imports were not reduced much by NTBs, and although their levels are generally low, average tariffs had large negative effect on imports.

Puzzled over the “small” estimates of the impact of NTBs, some trade theorists look for answers from the endogenous protection literature. The theory of endogenous protection predicts that, in response to increased import competition, domestic interests will intensify their lobbying activity for protection, which implies that higher levels of import penetration will lead to greater protection.9Ray (1981b) was among the first to test the prediction. He estimated trade and protection equations simultaneously for both United States and an aggregate of seven other industrialized countries, and he found no empirical evidence supporting the notion that trade protection and imports are concurrently determined. In a more recent attempt at this issue, Trefler (1993) found that, just on the contrary, taking into account the simultaneous determination of imports and trade protection results in a significantly larger estimate of the effect of protections on imports. Since both studies were based on U.S. data, the mixed results call for testing in a broader context.

With their focus on the trade protection determination, Lee and Swagel (1994) also estimated trade and protection equations simultaneously. They included in their sample both developed and developing countries, and their results turned out to be similar to Trefler′s. However, due to the model specification in the paper, their work seemed to bring more debates than closure to this issue. When specifying the determinants of trade protection, previous researchers chose import competition and variables such as industry concentration, economies of scale, labor structure, occupation, foreign protection level etc., which are presumably close proxies of political-economy factors and, in this context, reasonably exogenous. However, when the sample is expanded to include many other countries, the same kinds of variables are extremely difficult to obtain. Lee and Swagel therefore had to use some readily available data such as real wage change, export share in gross output, labor productivity, sectoral share of value added. From the point of view of estimating the trade barriers’ effect, we think that introducing those remote proxies of political-economy factors into the protection equation opens the door to a wide variety of endogeneity problems, precisely what their use was supposed to overcome.

IV. The Empirical Model

The monopolistic competition model suggests that correlating the difference between the actual trade flows and the flows predicted by the model with information on trade barriers can give an indication of the trade-reducing effects of trade barriers. One prediction from Equation (1) is that the import share of a good in a particular country is inversely related to that country′s share of world output of that good,

Three assumptions are, however, crucial for this result: similarity in tastes, absence of trade barriers, and zero transaction costs. If countries have a preference for goods made at home, shares of home goods in domestic consumption will exceed those of home goods in world production. Import barriers such as tariffs and NTBs will raise the share of home goods in home consumption relative to their share in world production. If there are international transaction costs, home goods will be relatively cheaper in the domestic market and their share in domestic consumption could deviate from that in world production.

In specifying the equation to be estimated, we add to the theoretical model those real world complexities that have been originally assumed away. We follow Lawrence (1987), Harrigan (1993), and Lee and Swagel (1994) among many others in the literature and adopt a log-linear functional form:

Where

min= total value of imports of commodity n by country i.
duin= domestic demand (production + imports – exports) of commodity n by country i.
dun= world total domestic demand of commodity n.
yin= output of commodity n in country i.
yn= world total output of commodity n
DISTANCEDi= the trade-weighted average of the distance between country i and all its trading partners.
TARin= ad valorem tariff rate imposed on commodity n by country i.
NTBin= NTB coverage ratio imposed on commodity n by country i.10
a, b, c, d, and e are parameters to be estimated.

The larger the share of a country′s output in the world, the larger is portion of its domestic demand that will be met by its own production, and thus the smaller is the import penetration ratio (min/duin). Distance is used as an indicator of international transaction costs. Higher transaction costs prevent a country from importing more. On top of that, the presence of tariffs and NTBs will further reduce the volume of imports. Since the production of each good is determined simultaneously with trade flow, we follow Harrigan (1992) and use factor endowments as instrumental variables for the sectoral production share. Specifically, we regress production share on factor endowments such as skilled labor, unskilled labor, capital stock and land. We then use the fitted values of production shares in estimating equation (8). However, it should be pointed out that there is a tradeoff associated with introducing instrumental variables. On the one hand, instrumental variables may help to get asymptotically consistent estimates; on the other hand, they can compromise the efficiency of the estimates.

As for another simultaneity problem caused by the political economy factors leading to import barriers often being erected in response to large volumes of imports, we also control for it by a set of instrumental variables. Since tariff rates in most of the countries are under WTO strictures, they can be more comfortably taken as exogenous than NTBs. Moreover, because our main concern is to control for the simultaneity problem rather than to specify a structural model of trade protection, in the instrumental variables set we include all the predetermined variables as well as variables such as tariffs and NTBs faced by a country′s exports, which can be justifiably treated as exogenous. The econometric strategy amounts to a two-stage estimation.

V. Model Estimation and Results

As pointed out by Learner and Bowen (1981) and Learner (1988), the response of imports to tariffs and NTBs is likely to vary across industries, since it depends on the elasticities of supply and demand, which might differ widely across industries. In contrast, there is likely to be less variation in these elasticities within a given industry across countries. We choose to pool across both countries and industries and use industry dummy to control for the industry-specific effect.

Our estimation procedure involves two stages. The first stage includes two steps. In step one, for each industry, we regress production shares on factor endowments. The results are presented in Table 10. The estimated coefficients, which are comparable across the industries, indicate that capital stock and skilled labor are important to all the industries, whereas unskilled labor has a negative effect on output share for a majority of the industries. In step two, aiming at controlling for the endogeneity problem of NTBs, we regress the import NTB coverage ratio of a country on a set of instrumental variables which includes the tariff rates and NTB coverage ratios faced by the country′s exports plus other exogenous variables including the country′s import tariff rates, distance and factor endowments such as the areas of different kinds of land, skilled and unskilled labor, and capital stock. Our argument is that, when it comes to a country′s decision on import NTBs, the tariffs and NTBs faced by that specific country′s exports can be justifiably treated as given. The regression results are presented in Table 11. A point worth mentioning is that one needs to be very careful in selecting instruments, because the potential exogeneity of the instruments is just as important as their relevance. At the second stage, we estimate the model as specified in equation (8) by using the corresponding fitted values of the output share and the NTBs coverage ratio from the first stage regression. As mentioned in previous section, there is a tradeoff between efficiency and consistency associated with introducing instrumental variables. For the sake of comparison, we report the estimation results with and without instrumental variables for either NTBs or output share.

Table 10:Results from Regressing Output Share on A Set of Instrumental Variables by Industry
Dependent variable: Output shareIndependent Variables
SectorsNo. of

Observations
R2ConstantCapital

Stock
Unskilled

Labor
Skilled

Labor
Crop LandPasture

Land
Other Land
311 Food products580.89-13.38*0.74*-0.62*0.67*0.42*-0.0020.30*
313 Beverages620.82-12.37*0.78*-0.47*0.54*0.080.19*-0.27*
314 Tobaccos600.83-15.77*0.86*0.180.020.080.02-0.16*
321 Textiles610.86-18.59*0.70*-0.050.73*0.25*0.03-0.44
322 Apparel, except footwear520.82-15.22*0.98*0.040.010.60*-0.01-0.57*
323 Leather products540.82-18.28*0.76*0.150.100.35*0.18*-0.38*
324 Footwear560.76-18.22*0.62*0.100.63*-0.030.16-0.33*
331 Wood products580.80-17.07*1.20*-0.65*0.230.53*-0.28*0.16
332 Furniture, except metal600.86-10.91*1.52*-81*0.030.41*0.03-0.25*
341 Paper and products590.88-16.09*1.00*-0.70*0.86*0.24-0.09-0.16
342 Printing and publishing610.84-12.91*1.08*-0.93*0.87*0.33*-0.09-0.28*
351 Industrial chemicals610.85-17.32*1.12*-0.44*0.530.55*-0.07-0.43*
352 Other chemicals610.85-17.85*0.74*-0.40*0.80*0.40*0.07-0.43*
353 Petroleum refineries550.64-16.95*0.62*-0.91*1.13*0.56*0.03-0.31
354 Miscellaneous petroleum and coal540.55-14.18*0.97*-0.120.100.45-0.11-0.33
355 Rubber products600.85-18.08*1.05*-0.160.030.67*-0.06-0.32*
356 Plastic products570.87-16.69*0.94*-0.78*1.00*0.330.06-0.36*
361 Pottery, china, earthenware590.67-15.46*1.17*-0.100.120.220.15-0.44*
362 Glass and products600.78-15.75*1.13*-0.44*0.370.140.120.19
369 Nonmetal mineral products610.77-14.60*0.89*-0.300.400.29*0.02-0.32*
371 Iron and steel580.86-20.32*1.14*-0.300.390.49*-0.20*-0.12
372 Nonferrous metals580.81-18.41*1.54*-0.50*0.060.23-0.090.14
381 Fabricated metal products600.89-15.55*1.27*-0.68*0.62*0.30*-0.01-0.32*
382 Non-electrical machinery580.90-19.62*1.52*-0.97*1.12*0.36-0.07-0.41*
383 Electric machinery590.91-17.78*1.36*-0.49*0.59*0.40*-0.14-0.39*
384 Transport equipment610.88-21.75*1.44*-0.370.310.42*-0.04-0.19
385 Professional and scientific equipment540.84-14.16*1.55*-0.78*0.660.000.04-0.32*
390 Other manufactures540.73-15.85*1.11*-0.310.270.41-0.18-0.21
Note: * significant at 0.05
Note: * significant at 0.05
Table 11.Results from Regressing Import NTB Coverage Ratio on a Set of Instrumental Variables with Industry Fixed Effect
Independent Variable: NTBsParametert-ratio
NTBs faced by exports-0.14-5.72
Tariff faced by exports0.051.92
Import Tariff0.072.97
Distance-25.2-5.06
Capital4.151.44
Skilled labor-0.54-0.11
Unskilled labor6.201.70
Crop land0.090.97
Pasture land2.251.48
Other land-3.05-1.62
Adjusted R20.16
Number of Observations1652

Table 12 shows that the estimation results are in general in line with the predictions of our model. Output share, distance and tariffs are correctly signed and statistically significant. Trade flow-weighted distance as a proxy of transaction cost does effectively impede the potential trade flows between countries. The presence of tariffs significantly reduces a country′s imports. Our estimation shows that, assuming that domestic demand does not change, a one percent increase in the tariff rate will lead to a two percent decrease in a country′s imports. Introducing instrumental variables for the output share and NTBs, while having little impact on the estimates of the tariff effect, does make a difference for the estimate of the effect of NTBs. Using instrumental variables for NTBs gives the correct sign for the estimated effect of NTBs, and the estimation is also statistically significant, whereas using instrumental variables for the output share leads to an enhanced effect of NTBs.

Table 12.Estimation of Trade Protection Effects on Trade Flows with Industry Fixed Effects
Independent Variable: Import Penetration Ratio(1)(2)(3)(4)
Output Share-0.14

(-17.6)
-0.13

(-15.0)
-0.10

(-11.5)
-0.08

(-8.5)
Distance-0.13

(-3.0)
-0.20

(-3.6)
-0.18

(-4.2)
-0.31

(-5.4)
Tariff-2.30

(-14.3)
-2.23

(-13.3)
-2.21

(-12.9)
-2.03

-(11.5)
NTBs0.06

(0.30)
-0.66

(-2.0)
0.002

(0.01)
-1.22

(-3.4)
Adjusted R20.530.530.490.49
Number of Observations1675167516751675
Note:In column 1, no instrumental variables are used for either NTBs or output share.In column 2, instrumental variables are used for NTBs, but not for output share.In column 3, instrumental variables are used for output share, but not for NTBs.In column 4, instrumental variables are used for both NTBs and output share.In parentheses are the corresponding t-ratios
Note:In column 1, no instrumental variables are used for either NTBs or output share.In column 2, instrumental variables are used for NTBs, but not for output share.In column 3, instrumental variables are used for output share, but not for NTBs.In column 4, instrumental variables are used for both NTBs and output share.In parentheses are the corresponding t-ratios

The functional features of various NTBs are not homogenous. As a matter of fact, different categories of NTBs may play quite different roles in restricting imports. We categorize the NTBs into five types based on UNCTAD′s trade barriers classification scheme and calculate a coverage ratio for each of them.11 Then, in place of a single overall NTB coverage ratio, we include in the equation the coverage ratios for all the five different types of NTBs. In carrying out this estimation, we decided not to tackle the simultaneity problem for NTBs, because it is very difficult to get a different set of instrumental variables for each different type of NTBs. In column 1 and 2 of Table 13, we present the estimation without and with instrumental variables for the output share, respectively. After breaking down NTBs by categories, the coefficients on output share, relative demand, distance and tariff are still signed correctly as well as statistically significant, and their magnitudes do not differ much from those from the previous estimation. As far as NTBs are concerned, three of the five types of NTBs, including quantity control measures, monopolistic measures and technical measures, are estimated to negatively influence the import penetration ratio, and the magnitudes of their effects all fall into a range between -0.4 and -0.6. We found positive coefficients on tariff measures12 and price control measures regardless of whether we used instrumental variables for output share or not. To the extent that the estimated positive coefficients of NTBs can be explained by the theory of endogenous protection, the results in Table 13 appear to indicate that, when facing high import penetration ratios, a country is more likely to have recourse to tariff measures and price control measures than to quantity control, monopolistic measures and technical measures.

Table 13.Estimation of the Trade Protection Effects on Trade Flows with Industry Fixed Effects and Categorized NTBs
Independent Variable: Import Penetration Ratio(1)(2)(3)(4)
Output Share-0.14

(-18.1)
-0.10

(-12.0)
-0.13

(-16.2)
-0.09

(-10.2)
Distance-0.12

(-3.0)
-0.18

(-11.1)
-0.13

(-3.1)
-0.18

(-4.2)
Tariff-2.04

(-12.5)
-1.90

(-11.1)
-2.22

(-13.4)
-2.08

(-11.9)
Tariff measures1.39

(4.9)
1.28

(4.4)
-0.50

(…)
-0.50

(…)
Price control measures2.83

(5.5)
3.06

(5.8)
-0.50

(…)
-0.50

(…)
Quantity control measures-0.42

(-2.4)
-0.46

(-2.6)
-0.43

(-2.4)
-0.49

(-2.6)
Monopolistic measures-0.61

(-1.4)
-0.53

(-1.2)
-0.51

(-0.7)
-0.42

(-0.5)
Technical measures-0.44

(-2.5)
-0.58

(-3.4)
-0.54

(-3.1)
-0.62

(-3.9)
Adjusted R20.550.520.530.50
Number of Observations1675167516751675
Note:In column 1, no instrumental variables are used for output share;In column 2, instrumental variables are used for output share;In column 3, no instrumental variables are used for output share with the coefficient for tariff measures and price control measures being constrained to be –0.50;In column 4, instrumental variables for output share, with tariff measures and price control measures being constrained to be –0.50.The corresponding t-ratios are shown in parentheses.
Note:In column 1, no instrumental variables are used for output share;In column 2, instrumental variables are used for output share;In column 3, no instrumental variables are used for output share with the coefficient for tariff measures and price control measures being constrained to be –0.50;In column 4, instrumental variables for output share, with tariff measures and price control measures being constrained to be –0.50.The corresponding t-ratios are shown in parentheses.

It follows that the estimated coefficients for quantity control measures, monopolistic measures and technical measures may be, to some extent, less affected by the endogeneity problem. Therefore, we re-estimated the equation, constraining the coefficients of both tariff measures and price control measures to be -0.5, the magnitude with which we feel more comfortable.

As shown in columns 3 and 4 of Table 13, the constrained estimation results in lower R2, the other coefficients, however, are robust to the constraints. According to the estimation, assuming that domestic demand does not change, a one percent increase in the coverage ratio of one category will lead roughly to a 0.5 percent decrease in a country′s imports. It should be pointed out, however, that while the model used in this paper controls for differences across industries, it assumed away the differences across countries, which may call for caution in interpreting its results.

VI. Summary

The bilateral trade flow data from WTDB and trade barriers data from TRAINS were combined, to produce a bilateral trade protection database. One of the important features of bilateral trade protection database is that it reveals the effective bilateral discrimination in a country′s trade protection regime. The analysis based on this bilateral trade protection database shows that despite many years of multilateral trade negotiations and unilateral cuts in trade barriers, the current level of overall trade barriers was still high as of 1994. Trade barriers also vary considerably across countries and industries. Moreover, we also showed that the bilateral effective discrimination of trade barriers attributable to the difference of a country′s trade structure in terms of commodity difference was quite significant. We presented our analysis on the global trade protection regime from two perspectives: country and industry.

An econometric estimation helped to determine the marginal import-reducing impact of trade barriers. We used a model that is based on the monopolistic competition trade model and 1994 cross-country and cross-industry data on trade barriers, trade flows, and production to estimate the import-reducing effect of trade barriers. The estimation results are in general consistent with the predictions of the model. Output share, distance and tariffs are correctly signed and statistically significant. The more a country produces a kind of good domestically, the less it will import from abroad. Trade flow-weighted distance as a proxy of transaction cost does effectively impede the potential trade flows between countries. The presence of tariffs significantly reduces a country′s imports. More specifically, our estimation shows that, assuming that domestic demand does not change, a one percent increase in the tariff rate will lead to a two percent decrease in a country′s imports. Introducing instrumental variables for output share and NTBs, while having little impact on the estimate of the tariff effect, does make a difference for the estimation of the effects of NTBs. Using instrumental variables for NTBs helps to get the correct sign for the estimated effect of NTBs, and the estimation is also statistically significant, whereas using instrumental variables for output share leads to an enhanced effect of NTBs. The different impact of various types of NTBs was also explored.

The results were mixed however: among the five types of NTBs, quantity control measures, monopolistic measures and technical measures reduce a country′s imports and the magnitudes of their estimated effects all fall into a range between –0.4 and –0.6. Not controlling for the NTBs’ simultaneity problem may explain the fact that we obtained positive estimates for the effect of tariff measures and price control measures on imports.

References

    Armington, P. S., 1969, “A Theory of Demand for Products Distinguished by Place of Production,” IMF Staff Papers, 16, pp. 159-176.

    Baldwin, Robert E., 1985, The Political Economy of U.S. Import Policy, Cambridge, Mass., MIT Press.

    Baltagi, B. H., 1995, Econometric Analysis of Panel Data, John Wiley & Sons, New York.

    Bergstrand, J. H., 1985, “The Gravity Equation in International Trade: Some Micro-economic Foundations and Empirical Evidence,” Review of Economics and Statistics, 67, August, pp. 474-481.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Bhagwati, J., Panagariya, A. and Srinivasan, T.N., Lectures in the Trade Theory, 1998, second edition, MIT Press.

    Bowen, Hollander and Viaene, Applied International Trade Analysis, 1998, The University of Michigan Press.

    Brock, William A., and StephenMagee, 1978, “The Economics of Special Interest Politics: The Case of the Tariff,” AER Papers and Proceedings, 68.

    • Search Google Scholar
    • Export Citation

    Feenstra, Robert C., 1995, “Estimating the Effects of Trade Policy,” NBER Working Paper No. 5051.

    Feenstra, Robert C., editor, 1988, Empirical Methods for International Trade, The MIT Press, Cambridge, Mass.

    Harrigan, James, 1996, “Openness to Trade in Manufactures in the OECD,” Journal of International Economics 40, pp. 23-39.

    Harrigan, James, 1995, “Factor Endowments and the International Location of Production: Econometric Evidence for the OECD, 1970-1985, Journal of International Economics, 39, pp. 123-144.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Harrigan, James, 1993, “OECD Imports and Trade Barriers in 1983,” Journal of International Economics35, pp. 95-111.

    Helpman, E., and P.Krugman, 1985, Market Structure and Foreign Trade: Increasing Returns, Imperfect Competition, and the International Economy (MIT Press, Cambridge, MA).

    • Search Google Scholar
    • Export Citation

    Hertel, T. W., 1997, Global Trade Analysis: Modeling and Applications, Cambridge University Press.

    Hillman, Arye L., 1977, “Declining Industries and Political-Support Projectionist Motives,” American Economic Review, 72, pp1182-87.

    • Search Google Scholar
    • Export Citation

    Hummels, D., and Levinsohn, J., “Monopolistic Competition and International Trade: Reconsidering the Evidence,” Quarterly Journal of Economics, 110(3), pp. 799-836.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Kendall, Maurice, Jean DickinsonGibbons, (1990), Rank Correlation Methods, London, Edward Arnold, a division of Hodder & Stoughton.

    Krugman, P., 1990, Rethinking International Trade, The MIT Press, Cambridge, Mass.

    Laird, S, and AlexanderYeats, 1990, Quantitative Methods for Trade-Barrier Analysis, New York University Press.

    Laird, S., 1996, “Quantifying Commercial Policies,” WTO Staff Working Paper TPRD-96-001.

    Lawrence, Robert, 1987, “Imports in Japan: Closed Markets or Minds?,” Brookings Papers on Economic Activity, no. 2.

    Learner, Edward E., 1974, “The Commodity Composition of International Trade in Manufactures: An Empirical Analysis,” Oxford Economics Papers, 26, 350-74.

    • Search Google Scholar
    • Export Citation

    Learner, Edward E., 1988a, “Cross-section Estimation of the Effects of Trade Barriers,” in RobertFeenstra, ed., Empirical Methods for International Trade, The MIT Press.

    • Search Google Scholar
    • Export Citation

    Learner, Edward E., 1988b, “Measures of Openness,” in Robert E.Baldwin, ed., Trade Policy Issues and Empirical Analysis (University of Chicago Press, Chicago, IL) 147-200.

    • Search Google Scholar
    • Export Citation

    Learner, Edward E., 1990, “Latin America As A Target of Trade Barriers Erected By The Major Developed Countries in 1983,” Journal of Developmental Economics 32, pp. 337-368.

    • Search Google Scholar
    • Export Citation

    Learner, Edward E. and Levinsohan, J., 1994, “International Trade Theory: The Evidence,” NBER working paper No. 4940.

    Lee, Jong-Wha, and PhillipSwagel, 1994, “Trade Barriers and Trade Flows Across Countries and Industries,” NBER Working Paper No. 4799.

    • Search Google Scholar
    • Export Citation

    Leontief, W.1954, “Domestic Production and Foreign Trade: the American Capital Position Re-examined,” Economia Internazionale.

    Maskus, K. E., 1991, “Comparing International Trade Data and Product and National Characteristics Data for the Analysis of Trade Models,” in Hopper, P. and J.DavidRichardsonedited, International Economic Transactions.

    • Search Google Scholar
    • Export Citation

    Magee, Stephen, WilliamBrock and LeslieYoung, 1989, Black Hole Tariffs and Endogenous Policy Theory: Political Economy in General Equilibrium, New York, Cambridge University Press.

    • Search Google Scholar
    • Export Citation

    Ray, Edward J., 1981a, “The Determinants of Tariff and Nontariff Trade Restrictions in the United States,” Journal of Political Economy, vol. 89, no.1.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Ray, Edward J, 1981b, “Tariff And Nontariff Barriers to Trade in the United States and Abroad,” The Review of Economics and Statistics, vol LXIII, no. 2.

    • Search Google Scholar
    • Export Citation

    Trefler, Daniel, 1993, “Trade Liberalization and the Theory of Endogenous Protection: An Econometric Study of U.S. Import Policy,” Journal of Political Economy, vol 101, no.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
APPENDIX I

Data in this Study

Data requirement for this study is large. It involves four data sets, the scopes of which are all across countries and industries: trade flows, trade protection measures, production and factor endowments. Ideally, in order to be consistent with the theoretical model, we should include in our sample as many countries and commodities as possible. However, at the more disaggregated level of commodity, we end up with fewer countries and hence a smaller number of observations for each commodity category.13 In other words, there is a trade-off between the sample size in terms of countries and commodities. As pointed out by Learner (1988), in this context, pooling across countries for each industry is definitely a more legitimate practice than pooling across industries for each country. After balancing between the losses and gains, we decided to settle for a disaggregation level of 28 industries, i.e. 3- digit ISIC (Rev. 2) so that our sample covers as many as 74 countries.

1. Trade Flows

The source of our trade flow data is the World Trade Database (WTD) released by Statistics Canada. The WTD is a complete matrix of international trade flows, created from data reported by member countries to the United Nations (U.N.) Statistical Office and broken down by partner country and commodity. In constructing the WTD, Statistics Canada has performed a number of adjustments to minimize inconsistency in the data as reported to the United Nations. Relying on the principle that import statistics are generally more accurate than export statistics, the WTD uses imports as the basis for allocating international trade flows. In cases where reporting countries group individual partner countries differently, geographic groupings have been created for which trade is comparable. The trade of non- reporting and late reporting countries is imputed using the trade data reported by their trading partners. The commodity data reported by the U.N. are adjusted to conform to the Canadian version of the SITC at the 4-digit level. The value of trade is measured consistently in thousands of U.S. dollars and valuation adjustments are performed to ensure that the dollar value of exports will equal the dollar value of import for all trade flows.14

The WTD covers trade flow data for some 160 countries for the period of 1980-96. We get the unilateral and bilateral imports data in this study by aggregating the data from 4-digit SITC to 3-digit ISIC level according to a classification conversion table between the two. The trade data from WTD is also used in computing tariffs and NTBs indicators.

2. Trade Protections

2.1 The Measurement of Trade Protections

This sub-section discusses measurement of trade protections for use in the trade policy modeling. Trade protection measures consist of two broad categories: custom tariffs and NTBs.

A. Tariffs

Customs tariffs are usually published in book form indicating the percentage of customs duty to be charged on commodities being imported, and they are typically classified according to national tariff classifications, which, based on the 6-digit Harmonized Commodity Coding and Classification System (HS), may contain as many as 13 digits. Since most of the tariffs are currently specified in ad valorem form, measurement of tariffs entails taking the average of tariff rates at the tariff line level in order to obtain the ad valorem tariffs at a more aggregated level. If it is desired to use some kind of weighting, an ideal but unavailable set of weights for these averages would be the level of imports that would have occurred if there were no barriers.

Three alternative sets of weights are suggested instead of the ideal weights: home imports, global imports and equal weights. However, each of these sets of weights is likely to depart from the ideal. Weighting by home imports understates the ideal rates if barriers are effective in reducing imports. Weighting by total global imports can also suffer from downward bias especially when the commodity structure of barriers is similar in most countries. Moreover, since the same set of weights is used for every country, country-specific import structures are not taken into account. Unweighted averages seem likely to be even worse approximation to the ideal average since barriers against commodities with negligible trade are treated the same as those against the imports of major commodities. In addition, like world import-weighted averages, unweighted averages also ignore different countries’ characteristics that would cause differences in their free-trade import levels. It has also been suggested that domestic output (or demand) should be used as weights instead of trade flows. However, output (demand) weights may introduce upward bias, since high trade barriers could lead to larger domestic output than would occur if there were no barriers. Moreover, since domestic output (or demand) data are usually not available at the tariff line level, it is still necessary to make simple averages of tariff lines to the point where the lowest level of output classification starts.15

The average tariff rates generated by using different weighting methods can be utilized as indicators of the home country′s tariff barriers against all of its trading partners. However, tariff rates faced by different exporters may vary for a variety of reasons. Discriminatory rates could result from regional trade agreements and an importer′s refusal to grant Most Favored Nation (MFN) status to a particular exporter. It has also been argued that, in reality, implicit discrimination, operating through the composition of trade in country-specific tariff schedules, far outweighs in importance the explicit discriminatory tariff practices like preferential trading arrangements.16 So if one chooses to focus on the issue of bilateral tariffs, then, when the relevant average is taken, the bilateral trade flows should be employed as weights to account for the compositional differences in trade.

B. Non-tariff Trade Barriers (NTBs)

In the literature and practice, two different approaches have been adopted to quantify the NTBs. These alternative approaches differ considerably in their methodology and in the nature of their empirical results. The first approach attempts to quantify trade and other economic effects of NTBs, often through the estimation of their ad valorem equivalents. The second, often referred to as the “inventory” approach, has been used primarily to produce descriptive statistics on the kinds, pattern and frequency of use of NTBs.

For empirical modeling, an important input is the price effect or “price wedge” associated with each NTB - often called the tariff equivalent of the NTBs. This is the difference between the world price of a product and the domestic price which is protected by NTBs. If world prices are genuinely free, they can be obtained from world commodity markets. They can then be compared directly with the domestic prices of identical products. It might also be possible to use an economic model of an industry, together with relevant supply and demand elasticities, to compute the price wedge based on observed changes in the volume of production and trade. Since no central records exist for non-tariff nominal equivalents, they must be independently estimated. As far as modeling disaggregated trade flows on a global scale is concerned, this approach is deemed impractical because it usually involves collecting data on prices from various sources which are not readily available.

The inventory approach is to record the number, form, and trade coverage of non-tariff trade policies as determined through surveys, frequency of complaints by trading partners and government reports. For empirical analysis involving NTBs inventories, two indices have been designed. One measure is a frequency index (Fj) showing the percentage of tariff lines covered by some pre-selected group of non-tariff measures,

where Ni is i‘th tariff line, Di is a dummy variable that takes the value of unity if one or more NTBs are applied to the ith item, or zero otherwise. The above summation is made over all countries exporting to the importing country j. Given that matched tariff-line-level import statistics are available, in which individual countries of origin for shipments are identified, a second index showing the share of total imports subject to NTBs can be computed. This trade coverage ratio (Cj) is defined as,

where Vi, t-n represents the value of imports if tariff-line item i in year (t-n), and Di, t is a dummy variable that takes the value of unity if an NTB is applied to the item, and zero otherwise. If n and m are zero, the index is based on current trade values, otherwise it is expressed in a base year′s trade weights. Holding n constant and varying m will measure the effects of changes in effective protections with constant trade weights, whereas holding m constant and varying n will measure the effects of changes in effective protections caused by changes in the structure of trade.

Since both the frequency index and the coverage ratio are numbers falling into the range between 0 and 1, in empirical analysis, they are typically treated in the same way as are the average ad valorem tariff rates. In the literature, almost all the empirical studies which involve NTBs in a multi-countries and multi-sectors context have used either of those two ratios as NTBs indicators.

A point to note, however, is that the inventory data are compiled mainly from official publications such as national customs schedules or GATT notifications. The reliance on official sources may cause understating the importance of some NTBs. Furthermore, the import coverage ratio and frequency ratio measure the extent to which imports are subject to NTBs and not the degree to which they are restricted. Finally, since the coverage ratio involves the value of imports, those drawbacks associated with tariffs when the import-weighted average of ad valorem tariff rates is being computed will nevertheless exist in the case of NTBs.

2.2 A Bilateral Trade Protection Database

As noted above, the measurement of trade protections across industries and countries involves computing the weighted average ad valorem tariff rates and NTB coverage ratios based on very detailed data of tariffs, NTBs and corresponding trade flows.

The United Nations Conference on Trade and Development (UNCTAD) has been tracking and compiling the information on worldwide trade protection measures for decades. The Trade Analysis and Information System (TRAINS), published by UNCTAD, is acclaimed by a leading figure in this field as “the most comprehensive collection of publicly available information” (Laird, 1996) on tariffs and NTB′s. It contains, inter alia, information on tariffs, NTBs and trade flows for most OECD countries and some 80 developing countries. For each basic Harmonized System item (6-digit level) as well as for any aggregate thereof, TRAINS allows for a cross-country comparison of indicators on import regime, such as tariff average and NTB′s frequency ratio; likewise, it allows the same comparison to be made of import values. Tariff schedules for most countries contain between 5,000 and 10,000 tariff lines and product differentiation. Empirical trade policy models necessarily incorporate trade policies based on aggregations of these tariff lines. However, when TRAINS carries out the aggregation, it only computes the unweighted average of tariff rates and NTB frequency ratios which are, as pointed out before, probably the worst approximation to the ideal average. Moreover, since TRAINS does not have data on bilateral trade flows, it cannot provide information on the bilateral trade protection, which is particularly important and interesting as far as the trade relationship among a group of countries is concerned.

By combining the WTD and TRAINS, we can create a Bilateral Trade Protection Database (BTPD) which constitutes the data source of bilateral and unilateral tariffs and NTBs in this study. From TRAINS, we can extract the ad valorem nominal rates for total import charges which include all duties and customs fees collected at the national border, NTB frequency ratios and the corresponding value of imports for each item of 6 digit HS. For the sake of simplicity, we call the total import charges “tariff rates” below. We then compute, with the help of a concordance table between 6 digit HS and 4 digit SITC, the import-weighted tariff rates, NTB coverage ratios and the corresponding trade value of imports at 4-digit SITC for each country.17 It is from this point that the combination of TRAINS and WTDB begins. Bilateral tariff rates and NTB coverage ratios at higher aggregation levels than 4-digit SITC are constructed by aggregating average tariff rates and NTB coverage ratios to the desired level using bilateral import values as weights. As pointed out above, this can result in significant differences in aggregated tariff rates or NTB coverage ratios for the same commodity imported from different sources. Given the Armington framework (product differentiation by region of origin), the incorporation of bilateral average tariff rates means that each trade flow can be subjected to a unique tariff rate or NTB coverage ratio.

When taking the import-weighted average of tariff rates and NTB coverage ratios, we use trade flow data in 1994 from the WTDB. The resulting BTPD consists of three kinds of matrices: bilateral trade flows, bilateral tariff rates and bilateral NTB coverage ratios, each for one category of commodity at the desired level of aggregation based on 4 digit SITC. The dimension of the matrix varies according to how the trading partner countries are grouped.

In this study, all the information on trade protections is from the BTPD. The information includes unilateral tariff rates and NTB coverage ratios, bilateral tariffs and coverage ratios, and export tariff rates and NTB coverage ratios faced by each country. An overview of the trade protection data from the BTPD is provided in the next section.

3. Production

As specified in the model, the production data at industrial level for all countries involved is required to calculate the import penetration ratios as well as production share. The United Nations Industrial Development Organization (UNIDO) makes available to users its database of industrial statistics at 3-digit ISIC level. The database covers 89 countries and regions. The data are generally for the period 1981-96 and arranged according to ISIC Revision 2. Information is presented by country, year and industry.

We take from the UNIDO database the production data for one single year, 1994. Since the industrial output is reported in domestic currency, we convert it into U.S. dollar based on exchange rates provided by the World Development Indicator (WDI) Database of The World Bank. The industrial production data are then used to calculate the output ratio and, together with import and export data from the WTD, the import penetration ratio.

4. Factor Endowment and Distance

The factor endowments include skilled labor, unskilled labor, capital stock and land. The data on cropland, pasture and other land areas can be taken directly from the WDI database. However, we have to compute the capital stock based on time series data on gross fixed investment, which can also be found in the WDI. In deriving the capital stock, we use an overall depreciation rate of 13.3%, the same rate as used by Summers and Heston (1990) in their construction of the Penn World Table. The International Labor Organization (ILO) publishes labor force data according to seven occupational categories. Following Maskus and Penubarti (1995), we define skilled labor as occupational category 0/1 and 2 and unskilled labor as total labor force minus skilled labor.

The distance data is downloaded from the web-site maintained by Dr. Jon Haveman.18 In the original data set, he provides the distance between economic centers of any two countries for some 100 countries. Following the usual practice in the Gravity Model literature, we use a trade-weighted measure of distance between a country and its trading partners as proxy for its transportation cost.

APPENDIX II
Bilateral Tariff Rate Matrix in 1994 Upper Left Panel(In percentage)
USASINSWIHKJAPNORICECANAUSEUISRNEWKORCHICZEMALSAUARGMAUOMAVENURUMEXSOUHUNCOLTHACOSBRAGABTRI
USA00002.72.96.1034.14.3334.310.44.92.99.711.126.7610.49.604.911.310.513.211.610.110.36.1
SIN20001.64.402.61.94.42.53.311.58.11.96.811.815.519.958.69.910.25.511.94.813.414.38.97.313.5
SWI3.60001.74.15.14.22.543.73.88.610.95.4310.410.919.65.28.210.68.82.6135.11117.88.26.95.9
HK6.4000612.410.48.98.27.19.37.310.310.55.77.71218.811.94.816.715.617.512.317.713.719.627.410.115.411.8
JAP2.400002.60.6464.84.29.47.810.38.18.311.814.933.6522.8129.911.818.92.619.619.411.31216.5
NOR1.50003.3020.522.31.747.29.43.64.310.813.124.62.95.45.410.34.892.16.210.36.96.52.2
ICE0.80003.91.701.1210.50.347.510.81.17.612.119.517.7013.7012.17.440.303.4024.514.40
CAN00001.70.63.502.82.32.43.18.410.85.64.26.511.623.36.813.97.203.211.86.58.513.927.37.87.6
AUS2.80005.60.60.12.401.54.35.15.810.91.83.7106.19.83.68.25.320.96.79.33.63.710.410.27.95.1
EU30002.94.34.14.34.505.95.48.810.56.47.31113.721.55.51111.510.97.114.45.519.816.110.39.56.1
ISR2.70000.63.94.63.43.2503.57.96.46.20.6010.918.207.68.29.86.25.95.18.412.612.37.39.3
NEW4.50006.11.60.86.74.52.74.9011.39.21.32.46.913.752.519.86.942.26.514.31.215.7016.7134.8
KOR3.60003.73.31.95.37.55.76.1609.711.14.612.11625.9523.213.911.817.216.111.622.616.115.918.916.2
CHI2.30001.80.80.61.91.81.96.13.94.404.71.21211.33.859.412.913.71.37.613.67.7035.913.521.9
CZE4.30003.84.71.76.85.34.64.34.57.510.90512.313.736.85.412.25.510.95.78.93.912.35.90.922.77.8
MAL2.30001.38.75.15.33.65.503.26.210.33.3012.313.321.52.711.411.112.14.58.27.913.322.58.77.615.2
SAU0.80000.46.800.10.20.900.35.1115.77.203.724.44.91204.90.36.505.70014.86.8
ARG5.10004.30.503.83.23.41.5119.610.91.60.95.509.90.115.510.19.52.374.416.122.212.312.524
MAU00003.216.114.9221.98.30.45.77.21112.617.21.819.800022.6355.47.80193015.100
OMA5.500002.1012.90.24.10050013.49.200010.1010.70.10.6000000
VEN1.20000.33.501.852.30.603.11100.115.26.70003.47.88.512.22.960015.86.7
URU6.10004.40.2066.13.20.27.94.210.51.708.116.20018.3016.75.414.21213.7206.310.79.8
MEX00001.40.614.403.73.57.54.88.510.15.624.612.412.73851110.608.813.88.821.308.212.16.5
SOU2.100020.223.43.62.73.44.112.910.92.42.711.911.626.84.54.99.3808.28.49.618.417.97.65.6
HUN50003.17.95.96.956.64.3513.59.695.211.516.134.14.312.211.613.46.2010.915.716.725.313.312.2
COL3.60004.60.50.62.60.37.68.44.25.710.93.18.514.313.504.910.8012.7036.8013.3028.713.49.8
THA4.90002.51.40.95.14.75.16.36.61310.92.96.415.613.816.9212.511.69.77.28.26.1020.5711.58.2
COS0.40000.200700.52.204.900004000010.13.6105.50050
BRA5.10004.45.47.95.65.95.85.35.47.3118.66.619.211.419.33.19.213.99.69.917.1012.10089.3
GAB3.30002.40.102.53.24.31.92.33.910.94.57.911.611.753.1015.514.48.83.186.29.122.725.2011.1
TRI1.80008005.54.55.100.38119.6012.46.9009.5109.36.8602.48008.90
Bilateral Tariff Rate Matrix in 1994 Lower Left Panel(In percentage)
USASINSWIHKJAPNORICECANAUSEUISRNEWKORCHICZEMALSAUARGMAUOMAVENURUMEXSOUHUNCOLTHACOSBRAGABTRI
TUR11.900048.48.58.64.48.86.83.514.710.76.810.810.113.65.8610.515.515.721.48.59.812.42513.211.516.7
POL5.10004.440.94.45.85.52.34.37.1114.36.212.314.633.259.511.71.59.35.412.42.126.88.36.6
ECU1.50006.20.2022.29.20.40.35.6112.44.6010.532.2015.96.512.18.536.309.90018.717.1
ALG10003.4000.100.8006.807.621.17.4004.40000005.80000
TUN5.10002.86.24.89.30.58.905.43.706.33.612.114.54056.4214.93.3004.326.811.3150
PER4.50001.81.114.92.11.62.91.73.34.210.97.70128.43.808.515.210.31.17.29.35.603.711.512.6
DOM10.10003.31.511.57.24.567.92.25.41130012.90010025.410.80012.700710.1
GUA3.30000.50.701.71.54.217.513.310.94.906.214.90019.714.76.10.110.813.89.807.36.812.2
PHI30002.48.514.38.754.79.46.314.310.94.72.213.314.33.43.516.615.711.29.45.789.207.812.918.7
PAR6.80001.30010.64.30.809.53.5114.508.840012.78.111.73.6505.70010.416
JAM5.80006.110.78.99.613.16.110.29.116.610.47.910.111.41716.5512.915.711.413.112.913.81324.210.211.68.9
CHN4.90004.40.8103.30.98.521.403.2115.40.712.16.83.9516.63.25.212.658206.126.123.715.60
MOR7.10002.49.614.99.74.37.503.98.110.75.315.912.115184.97.616.612.210.616.39.614.628.210.29.68.4
INE3.60002.7000.860.6017.78.3114.401.43.3059.97.89.81.61.201303.513.10
BOL10.70003.706.11.74.54.402.84.4118.341.2021.8058.319.716.29.958.402.3007.914.4
EGY5.50002.48.41514.17.25.51.80.84.9113.61111.513.213.24.91012.220.712.612.4204.829.46.15.317.2
SAL11.70003.30.103.91.93.60011.710.651.7018.50019.810.312.61.210.605.20012.911.2
SRI12.20003.711.59.416.687.52.44.87.71146.39.28.631.435.601177.63.512.310.80349.97.8
HON3.20000.6000.303.7003.3115000001409.90014.10020.25.42.5
CAM0.40000.59.10052.303.64.900012.100.6011.106.95.64.10322.7000
NIC40001.5000.25.63.90058.15.80014.100200901.111.511.10015.30
COT2.20008.20.203.81.952.74.65.7119.61.300.15.30009.53.44.609.80023.60
INA5.40002.713.36.512.310.36.55.896.310.864.8014.110.34.99.515.815.48.210013.309.511.410.3
CER1.900003.9000100000000000000.10000000
CON0.70000.1001.650.24.40500.30025.208.5013.60.20.1000000
KEN5.60002.94.803.60.26.90.52.65.1117.11.209.63.65007.91.320.203.8002023.2
BAN11.5000617.514.420.34.511.607.77.1118.84017.921.52.4013.427.1926010.30018.62.5
MAD2.70004.1004.31.84.92.109.5114.405.7016.1010.304.86.10.208.20000
CHA0.1000015.204.600.1003000.400000010.77.55000000
MAL47.80000008.60101.70119.700145.3509.204.900000013.6
ROW3.10001.12.11.63.32.73.84.33.26.2115.54.410.25.69.711.6132.611.81.96.29.66.516.89.21015.5
AVE3.300034.23.84.54.71.55.35.314.610.56.26.510.812.919.56.81210.811.77.114.1915.216.610.311.27.3
Bilateral Tariff Rate Matrix in 1994 Upper Right Panel(In percentage)
TURPOLECUALGTUNPERDOMGUAPHIPARJAMCHNMORINEBOLEGYSALSRIHONCAMNICCOTINACERCONKENBANMADCHAMALROWAVE
USA6.318.611.510.621.612.414.49.322.68.924.911.56.78.815.412.29.417.31211.815.215.524.110.712.621.250.422.415.811.918.68
SIN7.41416.550.323.71211.99.3246.613.96.87.308.921.13.319.46.612.4019.631.10031.887.3210020.36.8
SWI6.713.98.717.129.712.213.9622.69.313.315.46.49.119.216.72.813.33.71212.916.128.49.613.122.381.718.71918.614.45
HK10.818.120.731.233.412.613.414.633.910.417.121.518.49.110.251.81524.914.324.327.126.222.724.728.54397.723.921.734.621.512
JAP6.812.625.623.725.111.618.78.621.911.214.61911.28.924.825.210.221.19.518.913.315.129.21422.625.385.919.616.924.616.16.5
NOR4.77.86.524.820.8126.95.6250.89.56.748.913.715.210.311.51218.2010.7161017.632.773.19.310.210.314.72.8
ICE21.414.4000120019023.106007.50000012.621.90031850033.610.57.7
CAN4.5159.7318.313.69.96.121.36.2488.95910.5132.317.98.21011.79.618.811.610.626.958.320.74.27.519.53.7
AUS2.116.28.93.13.514.848.926.55.115.510.6686.39.18.920.85.919.2023.116.87.222.233.570.518.21227.418.66.4
EU7.21411.418.229.312.412.97.623.812.718.620.89.79.113.121.76163.314.614.914.320.714.818.425.476.22016.221.4183.6
ISR7.814.512.6019.4129.16.620.27.59.612.87.59.411.921.12.419.35.917.311.717105024.784.619.910.21017.14.8
NEW0.726.514.15.227.524.75.318.2263.613.814.17.9107.613.315.624.91.2026.71215.30031.1690013.418.47.1
KOR10.617.527.430.828.213.310.715.825.38.716.822.911.28.917.340.414.425.816.52116.714.230.717.325.331.594.423.718.828.819.48.3
CHI0.46.312.632.137.3148.312.719.712.11.618.51.39.314.7910.810.110.125.515.925.932.70220025.404016.24.9
CZE8.712.233.61423.111.829.79.522.2033.1346.37.86.632.37.113.312.520.18.420.429.5020.823.231.21029.90186.9
MAL4.89.910.925.320.31218.49.625.88.119.910.657.79.517.210.427.14.912.118.117.729.51017.833.683.615.114.722.8194.5
SAU8.911.27.516.419.412014.61909.61.93.70020.4014.38.113.7500000010.1001.42.4
ARG3.612.712.68.516.814.128.16.928.89.764.58.91.79.611.19.65.429.65.520.912.115.629.3028.29.764.13010.2024.611.4
MAU12.326.60027.9000199.819.3022.60019.3021.700012.630.5002010525010.817.59.2
OMA3.727.403418.400019003.95.30038.302.702307.636.3026.527.826.200022.66.3
VEN51516.730.1012.27.27.531.310.8036.9893.804.5012.508.628.6350000000124.6
URU3.99.59.34029.6220.610.119.212.4187.83.29.12030.59.709.7010013.2022065.700022.110.6
MEX9.911.516.27.717.411.67.39.833.412.315.215.928.6316.37.724.58.617.618.415.821.10030.935.2017.823.114.94.3
SOU42.8919.922.211.98.56.318.615.223.73.61.38.412.122.66.616.65.612.4015.926.11819.223.687.715.723.81928.512.9
HUN6.818.115.73233.921.410.521.536.48.94017.25.59.713.941.39.219.913.61821.622.528.128.323.538.281.425.827.128.418.67.6
COL29.921.213.140.1431213.812.22410.929.829.309.716010.112.910.116.51928.5180000000176.9
THA5.116.416.811.520.31214.96.824.412.765.412.74.79.213.213.26.434.18.314.515.213.917.910.621.115.866.121.110.220.921.39.7
COS05.907170051401.13.70.1007.400012.300024.50000009.60.9
BRA6.316.68.34.225.7120031.3015.416.23.4102516.1015.45.110.50030.40044.7016.70014.76.8
GAB11.614.515.119.215.612.216.38.326.51124.75.99.3921.116.4812.86.820.415.210.127.72026.727.101000146.6
TRI2.510.98.6035.6125609008.4101502.205.9011.20350045.778.92002012.66.8
Bilateral Tariff Rate Matrix in 1994 Lower Right Panel(In percentage)
TURPOLECUALGTUNPERDOMGUAPHIPARJAMCHNMORINEBOLEGYSALSRIHONCAMNICCOTINACERCONKENBANMADCHAMALROWAVE
TUR023.59.122.430.91213.911.230.67.611.7127.89.716.536.14.833.510.119.2025.229.712.527.735.297.716.4017.719.112
POL7.905.713.120.411.44.58.620.115.612.74.610.49.814.2205.211.66.911.76.116.924.722.923.119.36828.20015.57
ECU37.819.6041.340.413.709.217.9026.842.5108.5056.70013.1017.900000000010.27.3
ALG5.407.50212000004.30.20016.900000015.5000651010.2013.31.8
TUN7.65.817.825000000719.15.302515.900012.4014.9353019.8002910.6017.410.4
PER2.48.412.53.833.605.97.312.113.43.43.30.89.622.606.310.49.627.117.821.26.703249.839.923.929.9084.3
DOM9.411.58.504314.109.120.909.42.916.71000008.8510.90000000000.69
GUA8.99.88.63014.318.10003014.14.4016.99.311.83511.4019.7035005010500016.38
PHI7.31411.313.34312.39.88.409.310.14.87.30.224.46.310.411.150022.934.411.102566.910015.119.74
PAR08.46.5027120519.203.72.50.11015.601.101.201022.600000100027.65.5
JAM9.616.316.730.827.112.55.99.231.211.9021.39.78.89.221.815.410.59.222.418.723.228.927.428.340.185.223.220.238.319.76.7
CHN6.63.91.518.423.816.637.2021.403.500.610032.22002022.1021.5332622.538.7022.430.2017.210.5
MOR71113.252.829.613.3012.623.412.610.920.309.2029.22516.713.623.235.213.72125.93037.894.624.527.931.819.57.1
INE0.19.814.427.1012.911.7507.700002000000901.8000000021.75.1
BOL19.61714.240.142.316.508.90017073.1000029.750120000000001.96.4
EGY7.15.4021.820.71215022.821.14.416.36.300001.4025.1012.913.5012.126.364.13030.220157.5
SAL012.710.754.9013.8013.327.5030001000013.211.9019.90000000000.48.4
SRI10.514.713.14015.212.22310.120.610.921.124.95.510033.3901001727.630.10037.684.10022.922.511.3
HON010.47.5001213.712.900025.10025.8012.100024.30450000000214.7
CAM09.8052.517.30001402.212.7000001000018.90.721.829.819.41051019.9017.94.1
NIC0013.90023.216.911.100025.108.330015.4013.100033.9000000014.65.8
COT5.415.524.258.825.800018.506.412.40.20024.700016.600015.600023.7000.13.9
INA6.814.810.118.62612.408.226.509.317.64.29.5019.1019.3618.613.20015.400020.2001.35
CER0006000000000000000010.900007.700027.2017.22
CON0.210.80022.20000007.5000000029.300029.80000000.10.5
KEN5.510.810.66015.100017.5011.83.555031.4016.8010.5000000014.4003.96.3
BAN614.4040.534.316.3024.611.709.773.410029.7027.52510.1100000000001.610.4
MAD14.19.3059.817.612400004.513.70008.1025.3000043.300311050019.2228
CHA0.10.70327.80000012.500040.80008.600023.80000000.10.7
MAL253101539.11200590018.510000.200010000000012.3008.17.9
ROW7.510.69.918.918.213.66.7920.811.7137.73.69.418.418.616.523.713.512.819.419.325.11717.925.476.213.826.214.404.2
AVE7.21414.117.827.412.713.79.524.310.818.517.49.39.114.720.99.420.110.214.715.61523.815.718.726.481.719.616.619.517.35.4
Bilateral NTB Coverage Ratio in 1994 Upper Left Panel
USASINSWIHKJAPNORICECANAUSEUISRNEWKORCHICZEMALSAUARGMAUOMAVENURUMEXSOUHUNCOLTHACOSBRAGABTRL
USA00.7001.70.32.500.52.900.201.303.104.403.69.13.800009.509.200
SIN21.20001.21050.43.400.500.104.908.5023.42.3011.300020012.500
SWI6.40.2001.80.80.520.24.600.200.10.15.400.902.60.10.13.10000.5012.800
HK32.60.4000.825.2119.80.622.602.11.3004404.305.74.10.351.90001.5017.900
JAP14.50.700002.60.70.21.400.100.1010022.801.954.521.99.80007.2014.400
NOR40008.7010.10.65.800.10003.400.30141.102.40000027.700
ICE5.7000151.600.40.313.9000000000000000000000
CAN00.7001.90.16.401.24.500.10.35.11.13.603.2018.322.60.6000012.606.300
AUS9.20.5001.4002.4010.802.20.50012.60.20.101.30.40.89.70000.4018.400
EU11.30.9003.32.52.23.60.6000.30.10.10.59.60.211.4015.157.6100009.909.900
ISR12.10000.11.60.94.1211.200.3000000.9001.509.500019.6012.500
NEW6.90.4002.60.119.912.31.210.2000001.41.80.7011.50250005.600.100
KOR33.60.3002.81.53.1100.57.901.10003.4015.306.846.115.726.70006.1029.200
CHI5.100011001.30.85.900000007.3000.70.950003.7019.100
CZE16.60.3001.53.70.411.30.910.900.40005.8012000.50.24.90004.203.200
MAL30.50.5000.410.408.40.45.800.500.10004.9011.914.50.131.20009.806.600
SAU0.40000000.10.30.1000004.20000.60.801000099.30000
ARG16.10003.5001.619.58.10009.800000061.45.13.500036.500.200
MAU00005.638.5097.7058.7034.1025.602.8064.100001000000043.700
OMA34.70000.20033.60900000200000030.400000000
VEN1.80.1000.20000700000000.300008.400091.40000
URU32.80007.90017.2016.6000000018002.406.200019.108.800
MEX00.5000.90.6000.66.300.10000.3012.1005.37.6000015022.300
SOU13.60.5003.6001.50.313.800.10.500302.501.91.107.500011.5014.700
HUN30.92.2002.418.80.314.40.815.901.600.2010.803.401.30.71.338.300010.1027.900
COL10.34.7001.200.12.8048.20000000000007.700000000
THA17.32.5001.12.406.1218.402.200.5021.401200.313.28.116000003.200
COS00000.20010.100.30000098.900000000001000000
BRA30.60002.4308117.3000006.5011.100003.20000.20000
GAB17.64.3000.5002.10.260.30000.1031.109.80010.50.315.20002.701.100
TRI9.20001.900402.300000000006.804.700059.80000
Bilateral NTB Coverage Ration in 1994 Lower Left Panel(In percentage)
USASINSW1HKJAPNORICECANAUSEUISRNEWKORCHICZEMALSAUARGMAUOMAVENURUMEXSOUHUNCOLTHACOSBRAGABTRI
TUR54.50.2001181.43510.848.800001.239.602.1000.10.326.10005.3017.800
POL24.60005.25.405.71.419.700001.1301.9000.46.70.90000.105.100
ECU4.179.8003.3002.7025000000000017.9022.900054.90000
ALG00000000.200.200000000000000001000000
TUN35.30005.611.5049.2053.500001.8002.3000084.10000018.200
PER18.700022008.4014.300.10000.400.6002.109.70000.200.200
DOM53.10006.710.9017.74.821.20001.30000.10011.1059.10000.30000
GUA11.60000001.8076.600095.50000.20065.303.80000.30000
PHI20.90.5002.918.60.513.61.19.400.60.1000.901.200.2556.712.50000.3062.200
PAR22.715.2000.4009.701.70000.40008.40053.30000046.50000
JAM26.70.800320.70.2200.618.201.91003.408.301.704.525.50003.6031.400
CHN15.70005.20.604.30.331.30000000000000.30000.2065.100
MOR35.30.2001.333.9026.90.933.901.80004.401.400.702.530.90001.7010.300
INE110000.1002.204.400017.10002.40049.20.94.200010.2010000
BOL59.30002.2004.1016.6000000035.4047.80.305.20000.10000
EGY41.60000.50.2060.60.232.5000016.3000.404.40030.600021.5025.600
SAL76.450.80000.1014.2095.500042.8013.801.40094.7022.700055.20000
SRI78.96.2001.534.6058.90.249.701.80.9000.10001.3003.5000160100
HON5.60000000.1086.7000000000059.70000000000
CAM0.10000000020.600000000000096.800099.10000
NIC3.80000.3000066.400014.900000049.704.800000000
COT0.716.7000.8000.3014.6000000.10000000.400000000
INA35.30.8002.918.60.843.10.236.101.20.1004020.803.20.927.729.400014.105600
CER10.5000000002.8000000000000000000000
CON0000000000.2000000000000000000000
KEN35.22.1004.30.5014039.400.20.2009.80003.30021.900000000
BAN82.40007.254.4089.3079.3004.700002002.50081.600030000
MAD14.70.1004.5002.8042.7000000000000000000000
CHA0000062.304.900.5000000000000000000000
MAL5.900000035.7011.8000000000000000000000
ROW15.82.3001.33.50.410.20.216.501.6000.510.502.901.851.132.600078.9012.400
AVE15.90.8002.12.922.80.63.700.90.11.60.46.80.19.307.412.76.41200019.6010.700
Bilateral NTB Coverage Ratio in 1994 Upper Right Panel(In percentage)
TURPOLECUALGTUNPERDOMGUAPHIPARJAMCHNMORINEBOLEGYSALSRIHONCAMNICCOTINACERCONKENBANMADCHAMALROWAVE
USA0.70016.91.84.7000039.11.73.300023.50002.4013.70000.700010.94.1
SIN0.3000.53.40.1000015.91.20.700022.10.1000022.20004.500011.37.3
SWI0.40014.70.30000152.80.500047.9000009.20001.30003.44
HK1.6008.219.81.9000016.15.67.100013.90.100003600027.90007.817.6
JAP0.6000.16.60000016.933.72.300011.40000080001.100010.66.8
NOR0.50000.40000016.40.63.100019.20000013.40000.30008.55.3
ICE0.100000000000000000000021.400000004.811.5
CAN0.1005.10.48000049.40.41.200067.90000.103.7000000013.17.4
AUS000000.3000030.504.900029.40.100002.50000.4000135.5
EU1.1006.6272.4000020.95.65.500036.80000.6033.90000.9000122.8
ISR200000.1000021.60.53.400061.7000005700000009.39.5
NEW0.2008.592.377.9000060.30210000.6000001.8000000012.96.8
KOR1.3000.116.51000018.522.62.200012.20000013.500015.200010.911.6
CHI00010001.700002.50000015.3000000.200000008.95.5
CZE4.2000.210.200000367.32.600035.8000007.700010007.59.4
MAL0.1000.70.80000056.60.11.100060000012.40003.100019.310.4
SAU0.1005.40.80000024.801.8000000000000000001.12.8
ARG0.40072.7016.1000066.601.300034.60000023.7000000024.118.2
MAU000000000015.60000000000016.10005000012.552.9
OMA00005.20000005.39.100000000012.300012.500011.33.7
VEN0.2000.100.10000001300015.10009601.1000000011.610.5
URU000100072.1000028.8000002.9000000.1000000020.617.5
MEX4.600000.7000054.91000027.60002.504000000010.515.8
SOU0.30010.800000027.2020004.70.300001.80000.300027.815.7
HUN1008.92370.6000060.14.10.400029.50000023.40004.30009.712.5
COL00010000000098.30.5000013.20000.1013.100000007.219.7
THA0.4000.200.5000063.51.30.800029.5000006.9000300019.513.3
COS000000000037.200000000000000000008.32.6
BRA1.5007.54.10.8000018.11.30.100000000019.100000008.915.5
GAB7.30037.202.2000066.600.100033.90000.2012.300000008.223.2
TRI0000000000001.600064.6000005.8000000021.611.1
Bilateral NTB Coverage Ratio in 1994 Lower Roght Panel(In percentage)
TURPOLECUALGTUNPERDOMGUAPHIPARJAMCHNMORINEBOLEGYSALSRIHONCAMNICCOTINACERCONKENBANMADCHAMALROWAVE
TUR0008.59.80000034.22.10.200015.3000002.600000007.432.7
POL0.8000.28.90000011.9122.900048.6000002.800000003.615.4
ECU00093.900.40000001.3000000094.10000000008.58.7
ALG00000.10000000000000000026000000010.52.9
TUN0008.200000038.110.60000000000000000001144.8
PER00036.70000001.40000028.70000010.500000001310.8
DOM000007.3000000.10000000000000000000.245.6
GUA0000045.500001000000010.70001.5030.500000005.122.4
PHI0.60073.20000007.902.30000.10000032.60000.100026.311
PAR000000000093.30000018.2000000000000014.829.1
JAM1.10013.561.5000002.20.500013.3000001300015.30006.38.8
CHN0002.20.53.3000060.7000000000000.600000006.221.7
MOR1.5001.1151.6000022.61.200000000002800012.500011.714.1
INE00023.506.400000000000000003.500000009.46.7
BOL00010097.7000000038.4000000000000000000.529.6
EGY1006.83.60000071.77.10.100000000015.800000001324.2
SAL0000026.400001000000000