I. Introduction

The negative effect of trade costs on the trade performance of countries has long been a subject of trade research. An important aspect of trade costs specific to the trading environment would be the level of trade facilitation, or the trade costs associated with transport, regulatory and other logistics infrastructure of a country. According to De (2007), trade costs consist of two aspects: trader-specific trade costs which depend on operational efficiency, and trade costs specific to the trading environment incurred by in-built inefficiencies, such as institutional bottlenecks, information asymmetry, and administrative power that give rise to rent-seeking activities at various stages of transaction. With tariff rates no longer serving as effective trade barriers due to increasing WTO membership, trade facilitation has become an important trade-impeding factor and has received increasing attention as a substantive non-tariff barrier to trade, especially for developing countries.1 In fact, depending on the level of efficiency in delivering exported products across the border, traded volumes of products can be significantly increased. On the other hand, poor international or domestic logistical performance of a country can hamper trade. However, trade liberalization and growth has been empirically proven to lead to economic growth and development. Especially with the ongoing negotiations of the Doha Development Agenda on trade facilitation - the only surviving issue among the original four Singapore issues2 - numerous studies on trade facilitation have been conducted, many focusing on the implications of policy reform.

Recent existing studies on the negative impact of trade costs on trade have measured trade costs in terms of export cost (Dennis and Shepherd 2011), number of signatures (Sadikov 2007), costs and time involved in exporting and importing (Martinez-Zarzoso and Marquez-Ramos 2008), number of days needed to export or import (Persson 2008), or transport cost (Clark et al. 2004). There have also been efforts to construct a more composite “trade facilitation index”, consisting of measurements of port efficiency, customs, regulatory environment, and e-business infrastructure (Wilson et al. 2003, 2005).

On the other hand, recent literature on trade has analyzed export growth by decomposing it into the extensive and intensive margins - measuring the amount of impact in terms of the range of goods being traded (growth of exports in new categories of goods) as opposed to the volume of each traded good (growth of exports in goods that are already being exported). Examples of research that have used disaggregated trade data to measure the extensive (and also sometimes the intensive) margin include Kehoe and Ruhl (2002), Bergin and Glick (2005), Hummels and Klenow (2005), Brenton and Newfarmer (2007), Persson (2008), Dennis and Shepherd (2011) and Besedes and Prusa (2011). However, while a bulk of the previous literature have focused on the extensive margin of trade to find implications for export diversification, recently, an increasing number of work have dealt with the importance of the intensive margin of trade.

This paper aims to contribute by bringing these two lines of research together: by focusing on the different impact of trade facilitation on the extensive and intensive margins of trade. There are surprisingly very few, if any, previous literature examining this blended area of research, which in fact may have important trade facilitation-related policy implications, especially for developing countries. We can suggest that if trade grows at the extensive margin through improved trade facilitation, it implies a wider range of goods exported - export diversification in terms of product variety - which has been theoretically and empirically proven to lead to economic growth.3 On the other hand, if trade grows at the intensive margin, it is suggestive of larger market shares in the destined import markets due to increased volumes of existing exported goods, which is, in many cases, evidence of upgrade in a country’s development process to a more advanced level.4 Accordingly, depending on where a country stands in terms of its level of economic development, trade facilitation reforms will bring different implications, as the results of this research suggest. Furthermore, a deeper understanding of the role of trade facilitation on trade may help promote further integration of developing countries into the international trade arena.

This paper pursues an additional question of whether the extensive and intensive margins of trade in primary and manufactured goods are affected the same way by trade facilitation levels among developing countries with different income levels. Considering that the main export sector of a country has important implications for economic development, these additional elements have been incorporated in our research to present a more detailed analysis of this subject.

To examine these questions, we use highly disaggregated data on imports to 26 EU countries from 150 developing and developed countries in 2007 to decompose these countries’ exports into the extensive and intensive margins, using the method proposed by Hummels and Klenow (2005).5 To measure trade costs, we used the Logistics Performance Index (LPI) of the World Bank (2007)6 as proxy for each country’s trade facilitation level. While various types of trade cost measures have been used for different purposes, the use of measures that better reflect the entire process of the supply chain in trade may be warranted. Furthermore, despite the advantages presented by these measurements, there have been few work using the LPI to date.7

This paper proceeds as follows: Section II provides an overview of the previous literature on the research areas that this work attempts to bring together - the role of the extensive and intensive margins on trade, and the relationship between trade facilitation and trade. Section III explains the data and the empirical approach used for the empirical analysis, and Section IV reports the results. Finally, a brief conclusion and implications of the results are offered in Section V.

1)World Bank reports that the effective rate of protection provided by the transport costs is higher than that provided by tariffs. A more recent research (De 2007) has shown that a reduction in tariffs and transport costs by 10 percent would increase bilateral trade by about 2 and 6 percents, respectively.

2)Trade facilitation was formally raised at the 1996 ministerial meeting in Singapore. The three other Singapore issues are transparency in government procurement, trade and investment, and competition policy. While the latter three issues have been dropped from the WTO negotiation agenda, trade facilitation remains a strong agenda item within the current DDA negotiations and also in the ongoing aid-for-trade and trade capacity building programs.

3)There have been a number of empirical work on the link between export diversification and economic growth. Imbs and Wacziarg (2003) investigate the relationship between domestic sectoral concentration and per capita income patterns across countries, and find a U-shaped pattern whereby countries in their early stages of development diversify production, but specialize at higher income levels. Agosin (2007) finds that export diversification has a stronger effect on per capita income growth when a country’s exports grow faster, while Hesse (2009) finds a positive, but non-linear, effect of export diversification on per capita income growth, with developing countries benefiting more from diversifying their exports. Hausmann and Rodrik (2003), Hausmann, Hwang and Rodrik (2006) present a theoretical framework analyzing the benefits of export diversification for economic growth, whereby economic growth is driven by countries’ diversification of their investments into new activities generating a spill-over effect.

4)Recent empirical literature on export growth find that trade growth at the intensive margin is more important to export growth. Brenton and Newfarmer (2007), using panel data for 1995-2004, suggests that exporting larger quantities of existing products matters more than exporting a wider set of products for explaining export growth. Besedes and Prusa (2011) finds that the intensive margin of exported goods is more important for long-term export survival. Although in the earlier stages of development the exported goods show increasing patterns at the extensive margin, such export relationships are highly likely to be short-lived.

5)While there are many possible ways to define the extensive and intensive margins, Hummels and Klenow (2005) measured these margins by weighing categories of goods by their overall importance in exports to a given country.

6)The World Bank has reported the Logistics Performance Index (LPI) in 2007 and 2010. The LPI is comprised of seven sub-indicators: customs, infrastructure, international shipments, logistics competence, tracking and tracing, domestic logistics costs, and timeliness.

7)Hoekman and Nicita (2008) and Bang (2009).

II. Literature Review

There have been two strands of research relevant to the current study. The first strand of literature is the work on the role of the extensive and intensive margins in the growth of trade. There have been two opposing conclusions, with a number of papers having found that the extensive margin is more important in trade growth. Most prominent among these is Hummels and Klenow (2005) which find that the extensive margin accounts for 60% of the exports of larger economies. Kehoe and Ruhl (2002) and Bergin and Glick (2005) find that in the earlier stages of trade liberalization, the increased variety in traded goods plays a more important role in explaining trade growth. On the other hand, a large body of work has found the intensive margin to be more influential. Amurgo-Pacheco and Pierola (2008) find that exports at the intensive margin account for the most important share of overall trade growth, while at the extensive margin, geographic diversification is more important than product diversification, especially for developing countries. Besedes and Prusa (2011) find that developing countries would experience significantly higher export growth if they were able to improve their performance at the intensive margin. Some other papers expand this strand of research to connect them to the type of goods traded (homogeneous and differentiated goods).8

The second strand of research has been the work on the trade effects of trade facilitation. Clark et al. (2004) estimated the effect of transport cost on bilateral trade using the traditional gravity model and found that transport costs are an important determinant of bilateral trade flows with a significant negative impact. Lee et al. (2004) provides an empirical assessment of the economic impacts of trade facilitation on developed and developing countries’ trade flows, and finds that the benefits of trade facilitation are larger for the developing countries. Wilson et al. (2003, 2005) has significantly contributed to this area of research by constructing a composite trade facilitation index, measuring various areas of trade facilitation including port efficiency, customs, regulatory environment and e-business infrastructure. Their work has shown that while regulatory barriers deter trade, improvements in customs and e-business significantly expand trade, but to a lesser degree than improvements in ports or regulations. Hoekman and Nicita (2008) found that measures to improve logistics performance and facilitate trade are likely to have the greatest positive effects in expanding developing country trade. Lee and Bang (2009) has shown that at least for the East Asian countries, logistics infrastructure directly affects bilateral trade flows, and their improvement can dramatically increase trade performances.

However, all this work has been conducted with aggregate trade data. There has been surprisingly very few research on the trade effects of trade facilitation using decomposed trade data into the extensive and intensive margins. Such work will show the specific mechanisms of how trade facilitation contributes to trade growth - whether advanced trade facilitation levels are associated with exports of goods that are already being exported, or with the growth of exports in new categories. It will also contribute to understanding whether there are different implications for respective country groups by income levels. In a more closely related work to the present paper, Persson (2008) has explored the link between trade facilitation and the extensive and intensive margins of trade, additionally controlling for the type of goods being traded. Using import data of EU from developing countries in 2005, the author decomposed these countries’ exports into the extensive and intensive margins. As a proxy for trade transaction costs, the author used the number of days needed to export a standardized good, and found that there is a statistically significant and negative association between export transaction costs and the extensive margin for differentiated goods, but no such negative effect on the extensive margin for homogeneous goods. Comparing the effects on the extensive and intensive margins, empirical evidence is found that for differentiated goods, the extensive margin is more negatively affected by export transaction costs than the intensive margin, and that the negative trade effect on homogeneous goods stem from effects on the intensive margin.

Hence, this paper is mostly related to this last group of study in the sense that we use highly disaggregated data at the 5-digit level to construct the measurements for the extensive and intensive margins of trade. Unlike Persson (2008), this paper uses a different measurement method for capturing the extensive and intensive margins, using weighted trade shares as constructed by Hummels and Klenow (2005) instead of count data (counting export categories). Also, while Persson (2008) was not able to present meaningful results on the relationship between trade facilitation and the intensive margin of trade, we further break down the extensive and intensive margins of trade into primary and manufactured goods trade, since trade infrastructure may have differing impact depending on whether the export category is a primary or manufactured product. In addition, we took into account the impact of different income level of countries on trade facilitation levels, in order to explore whether trade facilitation levels may have different effects depending on the type of good exported and income level of the exporting country. As a result, we were able to reveal meaningful implications for trade both at the extensive and intensive margins. The construction of data and empirical method used for this analysis are explained in the following section.

8)Chaney (2008) shows that for homogeneous goods, the intensive margin is fairly sensitive to changes in trade barriers while the extensive margin is not, while Persson (2008) finds that differentiated products are more sensitive than homogeneous products in terms of the trade facilitation effects on the extensive margin of trade.

III. Empirical Analysis

1. Data

Our trade data comes from the United Nations Commodity Trade Statistics Database. We use the export data of 150 developing and developed countries to 26 EU countries9 in 2007, recorded using the SITC Revision 3 dataset. We use 5-digit level data10 to obtain disaggregated data, which is needed for meaningful construction of the extensive and intensive margins.

A key step in our analysis involves converting the annual trade data into the extensive and intensive margins using the Hummels and Klenow (2005) method. While there are a number of papers measuring the extensive and intensive margins of trade, many of the work on export diversification have measured the extensive margin by counting the number of products that a country exports11 (either by simple counting, counting over a certain size, or weighting categories in various ways) or using the ‘Herfindahl Hirschman Index’ which measures the degree of a country’s export concentration. However, the Hummels and Klenow method is also a useful way which appropriately weights categories of goods by their overall importance in exports to a given country so that it prevents a category from appearing important solely because an exporter exports a large quantity in a certain category.

Hummels and Klenow (2005) constructed the following measurement method to analyze trade growth in terms of the extensive and intensive margins. For the case when exporting country j's shipments to importing country m is a subset of k (rest-of-the-world)’s shipments to m, the extensive margin is defined as:

EM_jm equals k's exports to m in products I_jm relative to k's exports to m in all I categories. Therefore, the extensive margin can be thought of as a weighted count of exporting country j's categories relative to rest-of-the-world k's categories.

The corresponding intensive margin in given by:

IM_jm equals j's exports relative to k's exports in those categories in which j exports to m (I_jm). Thus, the intensive margin can be referred to as the proportion of exporting country j's exports relative to rest-of-the-world k's exports in importing country m's market.

In cross-section data, the extensive margin more likely refers to the proportion of the goods that a country exports as compared to the total export volume of the rest of the world. In other words, the extensive margin accounts for the variety of a country’s export goods relative to the world’s exported products. On the other hand, the intensive margin corresponds to the volume of a country’s export goods to a given market as compared to the traded volume of the same products exported by the rest of the world. Thus, the intensive margin would account for the market share that an exporting country has obtained in its destined import markets.

In Tables 1 through 3, we present a partial view of the extensive and intensive margins dataset per income group. The reported values are the highest in the respective income groups, ranked in descending order for each bilateral trade relationship (exporter-importer). As shown in Table 1, exporters in the higher income groups are found to have higher extensive margins than exporters in the lowest income group, although the difference is quite marginal within the highest range. It is also noteworthy that high-performing exporters in the lower middle-income country (LMIC) group are recording relatively higher extensive margins of trade than those in the upper middle-income country (UMIC) group. Likewise, as shown in the first two columns of Table 1, developing countries are similarly engaged in large volumes of exports with wider varieties as much as higher income countries. Conversely, in the case of the intensive margin, bilateral trade relationships with the highest intensive margins are mostly those among the lower income groups, as shown in Table 2. In very exceptional cases, intensive margin of 1 (maximum value) is measured for the bilateral trade relationships between Rwanda-Cyprus, Rwanda-Finland, and Uganda- Estonia. Bilateral trade with high intensive margins appear to be the case for importers in the lower income groups, implying that import markets of the developing countries are predominantly concentrated in terms of import varieties. Also, as shown in Table 3, exporters with lower levels of trade facilitation seem to be engaged in higher intensive margin trade, suggesting that advanced trade facilitation is not necessarily conducive to increasing trade volumes of existing export goods. Furthermore, when comparing the corresponding LPI levels of the high-performing exporters at the extensive and intensive margins, it reveals that exporters recording high extensive margin trade have relatively higher LPI scores, suggesting that trade facilitation may be more helpful in increasing exports of a wider variety of products.

To measure the level of trade facilitation, we use the World Bank’s (2007) Logistics Performance Index (LPI). This measurement of worldwide logistics performance levels is an in-depth cross-country assessment of the logistics gap among countries and provides a comprehensive picture of supply chain performance. It is comprised of seven areas of logistics performance: efficiency of customs clearance process; quality of transport and information technology infrastructure; ease and affordability of arranging international shipments; competence of the local logistics industry; ability to track and trace international shipments; domestic logistics costs; and timeliness of shipments in reaching destinations. The composite LPI index has been aggregated as a weighted average of the seven sub-indicators. We use the overall LPI score for analysis in this paper for purposes stated above.

Hoekman and Nicita (2008) used the LPI data for the first time as one of its variables to proxy trade facilitation levels to estimate its effect on trade. However, while using the overall LPI score for their analysis, the sub-indicators were not used for further analysis. There have been a few research using the LPI dataset to compare its implications on different regions. For example, Bang (2009) has studied the trade effects of the logistics efficiency in the East Asian region with detailed analysis using the LPI sub-indicators. While this paper does not provide detailed analysis using the sub-indicators, the calculated magnitude of coefficients for each income group on the extensive and intensive margins are reported in Appendix Table 9.

Table 4 reports the basic statistics of the LPI dataset per income group and geographical regions. As also shown in Figure 1, the LPI levels correspond positively with the income levels of countries, with average LPI levels increasing as countries get richer. As for the geographical regions, countries in Africa and the Latin America and the Carribean (LAC) regions fare lowest in terms of LPI performance, while countries in South East Asia and EU have higher LPI scores.

The regressors used in this study are listed in Table 5 below. We selected the dataset for 2007, since the LPI dataset is available only for 2007 and 2010, while the data for trade and other variables is considered more reliable for 2007. Data for GDP for exporter and importer countries in 2007 have been obtained by the World Development Indicators (World Bank 2010), and the gravity data from the CEPII database. Country classifications in terms of income groups and geographical regions are based on the World Bank’s country classification standards and the United Nations Statistics Division.

2. Empirical Model

The empirical model used in this paper is the traditional cross-section gravity model, which is generally used to capture the effect of trade costs on trade by including several time invariant trade-impeding variables such as distance, common language, colonial history, and landlockedness. Regional dummies for the EU, South East Asia, Africa, and the Latin America and Caribbean regions have been added to control for regional differences in export levels and to take into account the reduction of trade costs in intra-EU trade. As explained in the previous section, trade costs are captured by the trade facilitation indicators which are proxied by the LPI scores provided by the World Bank (2007). The basic estimation model incorporating the above is as follows:

where Trade_ij is decomposed into the extensive and intensive margins of bilateral trade between exporting country i and importing country j. Each margin is then further broken up into primary and manufactured goods trade. This is based on the understanding that trade infrastructure and logistics efficiency may have a more differentiating impact depending on whether the exported category is a primary or manufactured product. The rest of the variables are explained in Table 5.

In order to analyze the effects of the exporters’ trade facilitation level, additional models have been estimated, incorporating the quadratic function of the exporter’s LPI in order to capture the decreasing or increasing marginal effects on trade (4), and using interaction terms of exporter’s LPI with different income country groups (5).12

In an attempt to minimize the possibility of endogeneity bias, which may be suspected to exist in cross-sectional data, we also estimate the equations using the Two Stage Least Squares (TSLS) estimation method. We use regulatory quality and telephone infrastructure in bilateral pairs as instruments, mainly because regulatory quality and telephone infrastructures are important services aspect of trade facilitation that may affect trade, but are unlikely to be affected by the total volume of exports in a given country pair. The results are reported in Appendix Table 8. As shown, the significance and size of the coefficients are not largely different from the OLS estimation results, except that the TSLS estimation produces somewhat more conservative magnitudes of impact.

9)Poland has been excluded from EU-27 due to unavailability of data for key variables.

10)Using the 5-digit level data produces 2791 product categories.

11)Dennis and Shepherd (2011) and Persson (2008).

12)Note that the interaction term with the high income country (HIC) group has not been included in the model since it is used as the base group.

IV. Results

1. Trade Facilitation and Extensive and Intensive Margins of Trade

The results of the pooled OLS estimations are presented in Table 6. The baseline estimation in Model (1) shows that the exporting country’s trade facilitation level is, while important for the variety of the goods exported (extensive margin), not significantly associated with the market share obtained in the import markets (intensive margin). The coefficient can be interpreted as an elasticity, so improving trade facilitation levels by 1 percent will increase the extensive margin of primary exports by approximately 1.3 percent and that of manufactured exports by 2.4 percent. This finding is broadly consistent with previous findings that lower trade costs are associated with greater export diversification. Dennis and Shepherd (2011) found that export costs negatively affect (by 0.3 percent) developing country export diversification. Persson (2008) showed that decreasing border delays by 1 percent would increase the number of exported products by about 0.61 percent.

Model (2), which estimates the impact of the exporters’ trade facilitation levels using the quadratic form, shows that, at least for trade at the extensive margin, there exists a nonlinear relationship. Calculation of the turning point where the increasing trend starts to decline for the extensive margin reveals that high-income countries (HIC) with higher-than-average LPI scores no longer continue increasing their export varieties in both export sectors. This implies that trade growth at the extensive margin is predominantly the case for developing countries rather than for developed countries. This result is consistent with the previous work of Kehoe and Ruhl (2002) and Bergin and Glick (2005) which found that in the earlier stages of trade liberalization (mostly, the case of developing countries), the extensive margin plays a more important role in trade growth.

In Model (3), we have used interaction terms for the exporter’s LPI variable in order to take into account whether different income levels may affect the impact of trade facilitation on trade. The results show that in the primary sector, the low-income countries (LIC) and the lower middle-income countries (LMIC) with better logistics efficiency trade more varieties of primary products, while in the manufacturing sector, higher logistics capability had the highest impact for upper middle-income countries (UMIC), both in terms of product variety and market share. On the other hand, when countries at all income levels are pooled together (Model (1)), the trade facilitation level appears to have an insignificant impact on the intensive margin of trade. However, taking into account the different income levels reveal that the lower middle-income countries (LMIC) with better logistics efficiency trade most at the intensive margin in primary goods (with impact magnitude of 1.479 = 0.570 + 0.909). This implies that LMIC have the largest market share in primary goods trade, while the upper middle- income (UMIC) and high-income countries (HIC) have the largest market share in the manufactured exports in the destined import markets (with impact magnitudes of 0.842 for UMIC and 0.326 for HIC). Consequently, it appears that more advanced trade facilitation levels have different magnitudes of impact depending on the export sector and different income level of the exporting countries. On the whole, more efficient trade procedures have more impact on manufactured exports for the upper middle- income countries (UMIC). Whereas for the lower income countries (LIC), logistics efficiency have more impact on primary exports. However, the impact of better logistics procedures for the lower middle-income country (LMIC) group in exporting more varieties of manufactured goods is almost 3.5% (3.106 + 0.353), which is not a small amount.

To sum up the regression results of our key variables of interest, the benefits of trade facilitation for the developing countries lie in exporting primary goods at both the extensive and intensive margins. The lower middle-income country (LMIC) group, however, is benefitting the most from better trade facilitation procedures in primary goods trade at the intensive margin. Conversely, less developed countries were not able to reap the benefits of trade facilitation for exports of manufactured goods. Instead, the upper middle-income country (UMIC) group was gaining most from advanced trade facilitation for engaging in manufactured goods trade at both the extensive and intensive margins. These results further substantiate previous work by Lee et al. (2004) which conclude simply that trade facilitation benefits are larger for the developing countries. In other words, our results suggest that trade facilitation benefits are larger for the developing countries in terms of the extensive margin than the intensive margin, with slightly larger benefits for trade in primary goods than the manufactured goods at the extensive margin.

TSLS estimation results are reported in Appendix Table 8. As we are conducting a cross-section analysis, we need to deal with the possibility that measures of trade facilitation might be endogenous, thus we re-estimate the baseline model and use regulatory quality and telephone infrastructure in bilateral pairs as key instrument variables. F-tests of the null hypothesis that the instruments are exogenous conclude positively at the 1% significance level. The results of the TSLS estimation are more conservative, presenting smaller coefficients and t-values than the OLS results, but do not differ significantly for the main variables of interest. Also, to control for heteroskedasticity in large sample sizes, we report heteroskedasticity robust t-values in all the models.

The other explanatory variables show interesting results. Distance has expected negative and significant coefficients, while sharing a joint colonial history has a significant positive effect on both margins of trade. However, sharing a common language and being landlocked is significant only at the extensive margin and not at the intensive margin. This may be explainable by the fact that EU imports from developing countries are somewhat largely affected by past colonial relationships. The regional dummies show that exporters from Europe and Southeast Asia trade significantly in almost all sectors and at both margins of trade. However, exporters from Latin America and Africa trade less at both the extensive and intensive margins for manufactured exports than those from Southeast Asia and the rest of the world.

An unexpected outcome of the estimations on the intensive margin is that the coefficients of the importing country’s GDP and LPI present negative results (significant at 5% levels), suggesting that exporting to less developed importers produce higher volumes of trade at the intensive margin. However, such results are understandable if we take into account the possibility that import markets in developing countries may be subject to more concentrated imports, since their demand for imported goods may be less sophisticated than that in the markets of developed countries.

2. Policy Simulation

In examining the relationship between trade facilitation through a country’s logistics capability and its impact on trade, an important question raised by policy-makers may be: How much more volume of trade can be realized by engaging in trade facilitation reforms? More specifically, will trade facilitation reforms bring larger magnitudes of change, especially for the developing countries?

To address this question, we have considered two scenarios of trade facilitation. In the scenarios, the initial level of trade facilitation is the mean LPI level for countries in each income group. As a result of the trade policy reforms, the logistics performance levels are assumed to increase by 0.5 standard deviation from the mean LPI (scenario 1) and 1 standard deviation from the mean (scenario 2). Both scenarios are meaningful, especially for the less developed countries, in that the results show the expected amount of export growth when trade facilitation levels are further improved to reach the level of more developed countries. Table 7 reports the expected percentage changes in trade at both the extensive and intensive margins, further decomposed into the primary and manufactured export sectors. The percentage changes have been calculated using the estimated parameters of the OLS baseline regression results in Model (3). This is because the model takes into account the different impact of the exporter’s trade facilitation level per income group, and thus using these coefficients is expected to maximize the distinct impact on trade volumes. Policy simulation results using the TSLS regression estimates also produce similar results in terms of magnitude of impact per income group, albeit with slightly smaller size of coefficients.

In both scenarios, improving trade facilitation produces results that have important implications for the developing countries. Under scenario 1, for developing countries as a whole, there would be a 32.4 percent increase in the extensive margin of trade in primary goods, and almost 21 percent increase in the extensive margin of manufactured trade. On the other hand, primary trade at the intensive margin is expected to grow by 17.6 percent and manufactured trade at the intensive margin by 2.9 percent.

When comparing the magnitude of change among the different income groups, respective income groups are expected to benefit the most in different export sectors from trade policy reforms. It should also be noted that the simulation results are largely consistent with the regression results in Model (3) where the income levels have been taken into account. More specifically, the highest increase in the extensive margin of primary goods trade is expected for the lowest income countries (LIC). In other words, if Gambia (LPI score: 2.52) reforms its overall logistics efficiency to reach the level of Pakistan (2.62), and further to the level of Romania (2.91), its exports volume of primary goods, in terms of its exports variety, will increase by almost 15 percent, and further to 31 percent. On the other hand, the highest increase expected for the extensive margin of manufactured goods trade is in the upper middle- income countries (UMIC). In fact, upper middle-income countries (UMIC) are expected to see the highest increase in its manufactured exports at both the extensive and intensive margins (17 percent and 3.4 percent, respectively) due to improved trade facilitation levels, although the expected growth is far larger at the extensive margin.

As for trade at the intensive margin for the primary sector, the lower middle-income countries (LMIC) are expected to see the highest growth with 6.1 percent increase under scenario 1, and 12.4 percent under scenario 2. This means that when such countries as Bolivia (2.31) reforms its trade facilitation procedures to reach the level of Colombia (2.50) and further to the level of Croatia (2.71), its market share of primary exports may be expected to increase by the cited amounts.

In conclusion, our results provide empirical evidence that, for the developing countries, trade in primary goods will grow quite significantly in terms of both variety and market share after trade-facilitation related policy reforms have taken place. Since low-income countries (LIC) start with comparatively the lowest levels of trade infrastructure and procedures, the percentage changes in export growth due to trade policy reforms may be relatively quite high. Least developed countries may first need to secure a stable trading position in what they are best at producing and exporting (primary goods), prior to following the path of economic development through increased exports of manufactured products.

It is also noteworthy that improvements in trade facilitation will lead to increased exports of manufactured goods at the extensive margin for low-income countries (LIC) as well. Considering that exports in the manufacturing sector, especially in terms of export diversification, have important implications for economic development,13 our results that demonstrate the positive association between better logistics capacity and manufactured export growth at the extensive margin may be sobering news for developing countries.

Therefore, investing in trade facilitation related infrastructure and improving regulatory procedures for increased efficiency and transparency may prove to be an imperative policy for the developing countries, either through their own resources or through international assistance programs14 in case of the least developed countries.

13)Refer to footnote 3) above for empirical and theoretical evidence linking export diversification and economic growth.

14)For an overall review on the subject of aid-for-trade, refer to OECD and WTO (2009) and Hoekman and Wilson (2010); and Lee et al. (2004) for trade facilitation-related assistance measures for developing countries.

V. Conclusion

This paper has sought to analyze the role of trade facilitation on trade by decomposing trade into the extensive and intensive margins, additionally taking into account the export sectors and income levels of countries, in order to understand the dynamic mechanisms behind trade growth. Trade facilitation has become an important issue in trade cost discussions, serving as a substantive non-tariff barrier in the current trade environment, especially for the developing economies. Previous literature on this subject using such disaggregated data are surprisingly few, and most have used trade cost variables that only represent the time and/or cost aspects of the trade-impeding factors, rather than the overall performance level of trade facilitation. Accordingly, this paper aims to contribute to this line of research by using variables that comprehensively measure the trade facilitation levels of each country since they may better reflect the entire process of trade facilitation.

The analysis we have conducted in this paper provides sufficient evidence that trade facilitation levels have significant influence on trade, albeit with different magnitudes of impact for developing countries at different income levels and for different export sectors. On the whole, trade facilitation appears to have more impact on increasing trade at the extensive margin (increased export varieties) rather than at the intensive margin (increased market share of existing goods). However, more advanced trade facilitation levels, measured through each country’s international and domestic logistics performance levels, have diverse effects depending on export sector and income level of countries.

In conclusion, the results of our regression analysis show that more efficient trade procedures had the largest benefits for low-income countries (LIC) and lower middle-income countries (LMIC) in exporting primary goods at both the extensive and intensive margins, and the largest benefits for upper middle-income countries (UMIC) in exporting manufacturing goods at both the extensive and intensive margins. More specifically, in the primary sector, the impact of higher logistics efficiency was large for all developing countries (LIC > LMIC > UMIC) in trading more varieties (extensive margin) of primary products. However, for securing larger market shares (intensive margin) in primary exports, trade facilitation had the largest impact on LMIC and LIC (LMIC > LIC > UMIC). In the manufacturing sector, the impact of advanced trade facilitation was the largest for upper middle-income countries (UMIC) and high-income countries (UMIC > HIC) at both the extensive and intensive margins. These results are not surprising when considering the findings in the previous literature which suggest that countries in earlier stages of development trade more at the extensive margin than more advanced countries which export more at the intensive margin.15 An additional finding in this paper concerns the export sector, which suggests that developing countries are not only trading more at the extensive margin than their more developed counterparts, but trading more primary than manufactured goods.

The results of the policy simulations substantiate our estimated findings. Our results suggest that LIC and LMIC will significantly benefit from trade facilitation policy reforms, especially in terms of increased varieties of primary exports (almost by 15% and 13%, respectively). LMIC and LIC can expect increased market shares for their primary exports (by 6.1% and 5.3%) through trade facilitation reforms. More importantly, such policy reforms will also enable developing countries as a whole to benefit from increased trade in manufactured goods at the extensive margin (13.3% for LIC, 14.9% for LMIC, and 17% for UMIC). Considering the wealth of academic research that associates manufactured exports with economic growth, the findings of this paper that substantiate the evidence that improvement in trade facilitation help developing countries not only to increase trade in what they are already good at exporting but also to increase product diversification in manufacturing exports may be a sobering reminder for trade policy-makers in the developing countries for validating the need for investment in trade facilitation reforms.

15)Bergin and Glick (2005) and Besedes and Prusa (2011).