Export performance of developing countries: Does ...

DOI: 10.1111/rode.12389

REGULAR ARTICLE

Export performance of developing countries: Does landlockedness matter? Ramesh C. Paudel1

| Arusha Cooray1,2

1 Arndt–Corden Department of Economics, Crawford School of Public Policy, Australian National University, 2601, Canberra, ACT,Australia

Abstract Landlockedness imposes additional costs on trade and reduces international competitiveness. This paper exami-

2

Business School, University of Nottingham, Malaysia Correspondence Ramesh C. Paudel, Arndt–Corden Department of Economics, Crawford School of Public Policy, Australian National University, 2601, Canberra, ACT, Australia Email: [email protected]

nes the determinants of export performance in developing countries, within a comparative perspective of landlocked developing countries (LLDCs) and non-landlocked developing countries, by using a standard gravity modeling framework. The study covers data from 1995 to 2015. The results suggest that despite recent trade policy reforms, the overall export performance of LLDCs is lower than that of non-landlocked developing countries due to the inherent additional trade costs associated with landlockedness. The conventional wisdom that export performance is aided by economic openness also applies to LLDCs, but distance-related trade costs have a greater negative impact on exports from LLDCs than on other developing countries. The immediate trade policy challenge for LLDCs is therefore to create a more tradefriendly environment by lowering tariffs, reforming exchange rates and entering into regional trade agreements.

1

| INTRODUCTION

In 2015, landlocked developing countries (LLDCs) accounted for 0.6 percent of world exports and 7 percent of the world’s population, while non-LLDCs accounted for 34.4 percent of exports and 78 percent of the world’s population. Developed countries account for the rest of exports and 15 percent of the world population. In 1995, exports from LLDCs accounted for 0.3 percent of world

Rev Dev Econ. 2018;1–27.

wileyonlinelibrary.com/journal/rode

© 2018 John Wiley & Sons Ltd

|

1

2

|

PAUDEL

AND

COORAY

exports, compared to 17.6 percent from non-LLDCs. Per-capita non-oil exports of LLDCs in 2015 were a mere US$175, compared to US$843 for non-LLDCs (World Bank, 2016a). These statistics indicate that landlockedness may be an important factor in explaining the low export performance of LLDCs. If landlockedness does matter for export performance, how do landlocked countries compare with those that are not landlocked? This is an issue that warrants empirical investigation. The literature highlights the importance of export-led growth (Weiss, 1999: Greenaway, Morgan, & Wright, 2002; Balassa, 1985; Krueger, 1990; Santos-Paulino & Thirlwall, 2004; Awokuse, 2008; Athukorala, 2011). In particular, the East Asian Miracle led to increased support for this hypothesis (Stiglitz, 1996). Despite the importance of trade for economic growth, other factors such as geography could also play an important role. This was observed by Adam Smith as far back as 1776 (Radelet & Sachs, 1998). In this study we focus on landlockedness. Studies show that modeling exports is difficult in the case of LLDCs due to higher costs of production (Arvis, Raballand, & Marteau, 2007; Radelet & Sachs, 1998). Studies also suggest that landlocked countries face higher trade costs and lower trade volumes than non-landlocked countries (Faye, McArthur, Sachs, & Snow, 2004; Silva & Tenreyro, 2006; Behar & Venables, 2011; Coe & Hoffmaister, 1999; Lim~ao & Venables, 2001). The studies of Arvis et al. (2007), Grigoriou (2007), and Faye et al. (2004) examine the effect of landlockedness as one of many variables on trade. Arvis et al. (2007) use a supply chain approach put forward by Baumol and Vinod (1970). The Faye et al. (2004) study is a descriptive analysis. The emphasis of the Raballand (2003) and Grigoriou (2007) studies is on Central Asia. The existing literature has not attempted to investigate the difference in export performance between non-LLDCs and LLDCs from a comparative perspective. Nor has the emphasis of these studies been on landlockedness. Landlockedness has been included in these empirical studies as one of many variables, particularly to examine its effect on economic growth, not export performance. Following on from this discussion, this paper aims to contribute to the literature by specifically examining the effect of landlockedness on export performance from a comparative perspective— LLDCs versus non-LLDCs using the most recent data and methodology. This paper also examines the impact of landlockedness on export performance by considering inter-country differences among landlocked countries. The study additionally looks at whether landlockedness provides an explanation for the slow growth of African developing countries. The second objective of the study is relevant not only because Africa has experienced slow growth from the mid-1970s to the mid-1990s, but also because Africa initiated policy reforms in the early 1990s which led to an increase in trade. Against this background, Collier (2008) suggests that African countries have been adversely affected due to conflict, landlocked neighboring countries, bad governance, and misuse of resources. Coe and Hoffmaister (1999) find that the low level of trade in the African region is caused by economic size, geographical distance, and population. More recently, Bosker and Garretsen (2012) find that improving market access has improved manufacturing trade flows in Africa. Martinez and Mlachila (2013) conclude that the reforms in subSaharan Africa have worked to enhance economic development in this region. But none of these studies are focused on export performance of the region. To achieve these objectives, we use a gravity modeling framework on panel data for developing countries as exporters. Partner countries include both developed and developing countries, for the period 1995–2015. The findings of this study have important implications for policy by contributing to the debate on the role of geography in export performance. The findings of the study suggest that one way of minimizing the negative impact of geographical constraints on trade is to

PAUDEL

AND

COORAY

|

3

adopt a more trade-friendly environment by reducing tariffs and to enter into regional trade agreements with the trading partners. The paper is organized as follows. The next section briefly discusses the landlockedness and export performance literature. Section 3 presents an overview of export performance, comparing export trends and patterns for LLDCs and non-LLDCs in the light of whether trade policies are responsible for the difference in export performance between the two groups of countries. Section 4 presents the data and discusses the research methodologies. Section 5 presents the results. Section 6 concludes.

2

| LANDLOCKEDNESS AND EXPORT PERFORMANCE

The literature on international trade and landlockedness can be broadly divided into two strands: the effect of landlockedness on total trade flows and on export performance. Under the first category, Lim~ao and Venables (2001) suggest that a median landlocked country trades 30 percent less than a non-landlocked country. Similarly, Behar and Venables (2011), in a study of landlockedness including other factors related to economic geography on trade flows, find that landlockedness increases trade costs by almost 50 percent more than the costs imposed by distance, and reduces trade volume by 30–60 percent in a group of developed and developing countries. Coe and Hoffmaister (1999) argue that despite the fact that many African countries are landlocked, the shortfall in trade is offset by trade with former colonial powers. Raballand (2003), in a study of landlockedness on trade in Central Asia, finds that landlockedness reduces trade by as much as 80 percent in these countries. Studies on export performance in developing countries have focused on the relative export performance of landlocked countries from a broader comparative perspective. Djankov, Freund, and Pham (2010) find that a 1 percent increase in relative export time in landlocked countries, reduces trade by approximately 1 percent. Grigoriou (2007), in a study of 167 countries covering 1992– 2004, finds a negative effect of landlockedness for Central Asian trade flows. Faye et al. (2004) undertake a descriptive analysis examining the challenges facing LLDCs. They find that these countries have lower levels of human capital and external trade than their non-landlocked neighbors, and cite distance and high transportation costs as causes. Arvis et al. (2007) similarly argue that landlocked countries are affected by the high cost of freight services, unpredictability in transportation time, and problems in the operation of transit systems. They use a managerial approach by employing the supply chain model put forward by Baumol and Vinod (1970). Clarke (2005) observes that even after controlling for other factors, firms in landlocked countries are less likely to export than firms in countries with access to seaports. Paudel (2014) examines the effect of landlockedness on economic growth in LLDCs. He finds that landlockedness adversely affects economic growth. Other factors, including good governance, trade openness and coordinating infrastructure development with neighbors, are also are found to explain differences in growth rates among LLDCs. However, Paudel does not say much about the export performance of developing countries. The majority of these studies typically examine the effect of landlockedness as one variable among many which affect the export performance of countries, or focus on Africa and Central Asia. They do not attempt to compare the impacts of LLDCs and non-LLDCs on export performance. As mentioned in Section 1 above, the point of departure of our study from the existing literature is that we attempt to identify and quantify the difference in export performance between LLDCs and non-LLDCs.

4

|

PAUDEL

AND

COORAY

3 | EXPORT TRENDS AND PATTERNS IN DEVELOPING COUNTRIES 3.1

| Export trends

Over the past four decades, world exports have been growing at a much faster rate than world gross domestic product (Krugman, Cooper, & Srinivasan, 1995; Krugman, 2008). World exports totaled $124 billion, roughly 10 percent of world GDP, in 1960, and increased to $15,200 billion, almost 25 percent of world GDP, by 2010 (World Bank, 2016a). This suggests that developing countries’ merchandise exports have grown much faster than world exports, but they still account for just one third of total exports. Figure 1 shows that LLDCs’ per-capita non-oil exports is lower than that of non-LLDCs over the period 1995–2015. LLDCs’ share of exports to GDP remains poor compared to the rest of the developing countries as of 2009 (Paudel, 2014). Figure 2 shows that most LLDCs have a low per-capita GDP and percapita non-oil exports for the year 2015, as they are clustered toward the origin of the axis. However, there is a strong positive relationship in both groups between income and exports, suggesting that improving per-capita exports helps to increase per-capita GDP. In order to maintain consistency in the number of landlocked countries, this figure excludes nine landlocked countries which became separate countries after the dissolution of the USSR.1 In sum, the export trend of LLDCs indicates three important points. First, LLDCs’ per-capita exports are much lower than those of other developing countries. Second, the export growth rate of LLDCs has been higher in recent years. However, as LLDCs’ export levels are so much lower than those of their non-landlocked counterparts, the export growth rate required to meet the levels of the non-LLDCs would take a long time. Third, urgent attention is needed to boost the export performance of LLDCs.

3.2

| Export patterns

Non-oil exports of LLDCs are low compared to those of non-LLDCs. For example, for LLDCs the average share of non-oil exports in total exports was about 80 percent in 1999 and reached 87 percent in 2015, while for non-LLDCs it was 87 percent and 88 respectively. If we look at the share for developed countries, the figures were 96 percent and 90 percent respectively. Thus, the share of non-oil 1000 900 800 700 600 500 400 300 200 100 0 1995

1997

1999

2001

2003

2005

Non-landlocked developing countries

2007

2009

2011

2013

Landlocked developing countries

F I G U R E 1 Developing Countries’ Per-Capita Non-Oil Exports, Current Price, US$ Source: Based on data from World Bank (2016a)

2015

PAUDEL

AND

|

COORAY

5

exports of developed countries is declining and of LLDCs is increasing. Furthermore, developed countries have a larger share of manufacturing exports, while LLDCs have a larger share of primary products. Non-LLDCs’ share of manufacturing has been declining over the period. The rate of growth of exports differs across countries according to income group. In addition, the sources of exports are not unique in all LLDCs. The share of manufacturing and primary exports stood at 27 percent and 60 percent respectively in 2015, compared to 37 and 43 percent respectively in 1999; the figures for non-LLDCs were 42 percent and 47 percent respectively in 2015 (Table 1). These data show that manufactured goods are not the dominant exports from LLDCs, and are in lower than those from non-LLDCs. At the individual country level, market shares have varied substantially over time in a few countries. Based on data from 2015, of the 34 LLDCs, Kazakhstan is the largest exporter, but about 70 percent of its exports come from the oil sector. It is followed by Belarus, also an oil exporter (with 29 percent of merchandise exports). Azerbaijan and Bolivia are the other notable oil exporters. Primary commodities dominate the export structures of most LLDCs. Only four countries, Belarus, Macedonia FYR, Nepal, and Botswana, experienced a contribution of about 50 percent from manufacturing exports in 2015 (Armenia shifted ground from manufacturing to primary products in 2015). The contribution from manufacturing increased between 1999 and 2015 in only five countries: Bhutan, Kyrgyz Republic, Niger, Rwanda, and Uganda.

3.3

| Are trade policies responsible for the difference?

12000 16000 8000 4000

KNA URY PLW PANARG

SYC CHL

CRI

KAZ

MYS

GRD RUS TUR MUS SUR ROM MEX BRA LBN GAB CHN LCA MDV DMA TKM VCT BGR BWA DOM PER COL ECU THA BLR ZAF AZE JAM FJI JOR BLZ MKD NAM LBY IRQ SLV DZA BIH GUY PRY MNG AGO ALB WSM LKA GTM TUN GEO EGY ARM IDN CPVBOL PHL SWZ NGA BTN MARHND UZB UKRSLB VNM NIC MDA COG LAO IND PAK CIV KHM KEN GHA ZMB KIR CMR MMR BGD TMP KGZ TJK ZWE SEN TZA HTI BEN TCD SSD NPL MLI UGA RWA SLE ETH BFA AFG GNB SOM TGO MOZ GIN ZAR LBR MDG MWI NER CAF BDI

0

Per capita exports

In this subsection, we examine whether trade policies in LLDCs are responsible for their poor export performance. A notable number of studies (Weiss, 1999; Greenaway et al., 2002; SantosPaulino & Thirlwall, 2004; Awokuse, 2008; Athukorala, 2011), suggest that the greater the magnitude of trade liberalization accompanied by efficient management, the greater the possibilities for improving export performance. Notably, many of these developing countries (including LLDCs) initiated liberalization and reform in the early 1990s. Table 2 presents the five-year average tariff rate structure for developing countries classified by region. LLDCs are scattered across five regions. East Asia and the Pacific (EAP) comprises two,

0

2000

4000

6000

8000

Per capita GDP

Non-Landlocked

Landlocked

Fitted values

F I G U R E 2 Per-Capita GDP and Non-Oil Exports in Developing Countries, 2015, US$ Source: Based on data from World Bank (2016a)

6

|

PAUDEL

AND

COORAY

T A B L E 1 Exports Scenario of Landlocked Developing Countries Country/region

Year

Total non-oil exports (%)

Manufacturing exports (%)

Primary exports (%)

Total exports (US$ million)

1999

–

–

–

–

2007

–

–

–

–

2015

100

28

71

2,985

1999

100

20

80

358

2007

91

5

86

1,887

2015

79

2

78

4,579

1999

92

59

32

232

2007

99

56

43

815

2015

94

16

78

1,312

1999

21

9

13

929

2007

19

6

12

6,058

2015

13

3

10

11,327

1999

91

75

16

5,909

2007

65

53

12

24,275

2015

71

48

22

26,660

1999

56

24

33

5,871

2007

34

13

21

47,748

2015

32

14

18

45,954

1999

–

–

–

–

2007

–

–

–

–

2015

–

–

–

–

1999

88

20

68

454

2007

88

35

53

904

2015

95

25

70

1,441

1999

98

66

32

1,191

2007

95

76

19

3,356

2015

99

82

17

4,490

1999

100

27

73

428

2007

100

32

68

846

2015

100

34

65

1,304

1999

–

–

–

–

2007

–

–

–

–

2015

–

–

–

–

EAP Lao PDR

Mongolia

ECA Armenia

Azerbaijan

Belarus

Kazakhstan

Kosovo

Kyrgyz Republic

Macedonia, FYR

Moldova

Serbia

(Continues)

PAUDEL

AND

|

COORAY

7

T A B L E 1 (Continued)

Country/region Tajikistan

Turkmenistan

Uzbekistan

Year

Total non-oil exports (%)

Manufacturing exports (%)

Primary exports (%)

Total exports (US$ million)

1999

87

13

74

692

2007

–

–

–

–

2015

–

–

–

–

1999

36

12

24

1,187

2007

–

–

–

–

2015

–

–

–

–

1999

–

–

–

–

2007

–

–

–

–

2015

–

–

–

–

LAC Bolivia

Paraguay

1999

95

38

57

1,402

2007

52

7

45

4,813

2015

54

4

49

8,726

1999

100

15

85

741

2007

60

8

52

4,724

2015

75

10

65

8,328

1999

–

–

–

–

2007

–

–

–

–

2015

97

16

81

571

1999

58

40

18

116

2007

63

38

25

675

2015

–

–

–

–

1999

100

77

23

524

2007

–

–

–

–

2015

100

68

32

660

SA Afghanistan

Bhutan

Nepal

SSA Burundi

Burkina Faso

Botswana

1999

100

0

100

62

2007

96

21

76

156

2015

100

20

80

112

1999

99

15

84

236

2007

100

7

93

453

2015

100

5

95

2,177

1999

100

90

10

2,763

2007

100

73

27

5,073

2015

100

90

10

6,319

(Continues)

8

|

PAUDEL

AND

COORAY

T A B L E 1 (Continued)

Country/region

Year

Total non-oil exports (%)

Manufacturing exports (%)

Primary exports (%)

Total exports (US$ million)

Central African Republic

1999

100

61

39

110

2007

100

22

78

131

2015

100

80

20

97

1999

–

–

–

–

2007

–

–

–

–

2015

–

–

–

–

1999

100

7

93

449

2007

100

13

87

1,277

2015

100

7

93

4,194

1999

100

95

5

336

2007

–

–

–

–

2015

–

–

–

–

1999

100

9

91

438

2007

100

11

89

868

2015

100

15

85

1,080

1999

100

5

95

472

2007

100

3

96

1,441

2015

–

–

–

–

1999

100

2

98

181

2007

98

10

88

627

2015

81

9

72

790

1999

100

3

97

57

2007

100

5

95

155

2015

100

14

86

409

1999

–

–

–

–

2007

99

70

29

1,086

2015

–

–

–

–

Chad

Ethiopia

Lesotho

Malawi

Mali

Niger

Rwanda

Swaziland

Uganda

Zambia

Zimbabwe

1999

100

3

97

506

2007

99

21

78

1,099

2015

99

24

75

1,744

1999

99

18

81

1,063

2007

99

13

87

4,617

2015

98

11

87

6,983

1999

98

27

71

1,887

2007

99

48

51

3,185

2015

98

13

86

2,704

(Continues)

PAUDEL

AND

|

COORAY

9

T A B L E 1 (Continued)

Country/region

Year

Total non-oil exports (%)

Manufacturing exports (%)

Primary exports (%)

Total exports (US$ million)

Landlocked developing

1999

80

37

43

24,803

2007

86

27

59

116,269

2015

87

27

60

144,948

1999

87

65

21

979,950

2007

83

38

45

3,628,270

2015

88

42

47

5,179,914

1999

96

81

15

3,988,681

2007

91

74

17

8,345,468

2015

90

72

18

9,661,838

1999

93

77

15

5,292,515

2007

86

69

17

13,100,000

2015

89

70

19

15,300,000

Non-landlocked developing

Developed

World

Note: “–” indicates figures are not available. Source: Based on data compiled from World Bank (2016b).

Eastern Europe and Central Asia (ECA) 12, Latin America and the Caribbean (LAC) two, South Asia (SA) three, and sub-Saharan Africa (SSA) 15 countries. South Sudan has been excluded due to the lack of data. In the EAP region, the average tariff rate of LLDCs is slightly higher than that of non-LLDCs over the period 1995–2010. The average rate for LLDCs is lower than for nonLLDCs in the ECA, LAC, SA, and SSA regions. This implies that LLDCs are more open to foreign trade than non-LLDCs. We also update the widely used Sachs–Warner trade liberalization index, which was developed by Sachs and Warner (1995) to cover the period until 2009 including all LLDCs, which were not covered in the previous update of the index by Wacziarg and Welch (2008). This index defines a country as liberalized when it has an average tariff rate of not more than 40 percent, a black market premium rate of not more than 20 percent, non-tariff barrier rates of not more than 40 percent, no state monopoly on major exports, and no socialist economic system. The table in Appendix A shows the liberalization status of all LLDCs based on this index. According to the index, 23 LLDCs are open, while 11 of them remained closed until 2009. Lao PDR, Belarus, Kazakhstan, Kosovo, Serbia, Turkmenistan, Uzbekistan, Bhutan, Afghanistan, and Central African Republic are classified as closed because of existing non-tariff barriers. Zimbabwe remains closed because its black market premium rate exceeds the 20 percent criterion. Only five countries, Chad, Lesotho, Malawi, Rwanda, and Swaziland, have graduated to open, satisfying all of the criteria since 1999. As this table shows, based on the average tariff rate, only Zimbabwe has a tariff rate greater than 20 percent, followed by Bhutan at 18 percent, and both the Central African Republic and Lesotho at about 15 percent. The rest of the LLDCs have average tariff rates less than 15 percent. Notably, only seven countries have an average tariff rate less than 5 percent. Turkmenistan has the lowest average tariff rate, 1.4 percent; however, based on other criteria reported in Appendix A, it is still classified as a closed economy.

10

|

PAUDEL

AND

COORAY

T A B L E 2 Regional Tariff Structure in Developing Countries Region EAP

ECA

LAC

SA

SSA

1995–99

2000–04

Landlocked

na

12.6

Non-landlocked

2005–10 7.4

Average percent 1995-2014 9.0

12.1

8.3

5.4

7.9

Landlocked

4.2

5.1

3.7

4.4

Non-landlocked

5.9

4.9

3.1

4.6

Landlocked

9.0

8.8

4.1

6.5

Non-landlocked

11.5

9.2

6.3

8.3

Landlocked

15.3

14.4

11.4

12.5

Non-landlocked

33.2

17.2

10.6

17.3

Landlocked

15.4

11.1

9.4

10.8

Non-landlocked

17.7

11.8

9.3

12.1

Note: na denotes not available, average for 1995–2014 due to more data gap in 2015. Source: Based on data compiled from World Bank (2016a).

4 4.1

| DETERMINANTS OF EXPORT PERFORMANCE | Model, Variables and Data Description

Tinbergen (1962) proposed the original gravity model, which is known as a “work horse” among international trade economists (Bergeijk & Brakman, 2010). This model explains trade flows in terms of GDP of reporting and partner countries and geographic distance between countries:
ln Xij;t ¼ a þ b1 lnðGDPi;t Þ þ b2 ln GDPj;t Þ þ b3 lnðDISij;t Þ þ eij;t (1) It is postulated here that GDP represents gravitational forces and geographic distance represents trade costs. The basic gravity model is augmented here with additional variables of interest to identify the export performance of LLDCs and non-LLDCs:
ln Xij;t ¼a þ b1 LLOCKi þ b2 lnðOPENi;t Þ þ b3 lnðGDPi;t Þ þ b4 lnðGDPj;t Þ þ b5 lnðDISij Þ (2) þ b6 lnðRERij;t Þ þ b7 lnðLANij;t Þ þ b8 lnðBORij;t Þ þ b9 lnðRFEi;t Þ þ eij;t : Then the model is tested with various alternative specifications, as suggested by the literature for robustness:
ln Xij;t ¼b1 LLOCKi þ b2 lnðOPENi;t Þ þ b3 lnðGDPi;t Þ þ b4 lnðGDPj;t Þ þ b5 lnðDISij Þ þ b6 lnðRERij;t Þ þ b7 ðLANij;t Þ þ b8 lnðBORij;t Þ þ b9 lnðRFEi;t Þ þ b10 ðRTAij;t Þ þ b11 ðAFRICAi Þ þ b12 ðEUTCi Þ þ eij;t

(3)

where ln denotes the natural logarithm, subscripts i and j refer to the exporter and partner country, and t refers to time. The variables used in the analysis and their detailed definitions, together with postulated signs for the regression coefficients, are listed in Appendix B. The last term of equations (2) and (3) is the error term. The error component structure is given by

PAUDEL

AND

|

COORAY

eij;t ¼ lij;t þ ht þ uij;t

11

(4)

where lij;t is a fixed effect that might be correlated with explanatory variables, ht captures timespecific effects common to all cross-section units, and /ij,t is an error term uncorrelated across cross-section units and over time periods. The dependent variable is non-oil exports (X) measured in US dollars and expressed in log form. The reasons for selecting non-oil exports are as follows. First, oil prices fluctuate widely and are different in nature from prices of mining and other manufacturing products. The fluctuations in oil prices would make the results of the estimations more unreliable. Second, oil product exports depend on geography and do not explain the role of policy in a country. Finally, only a few countries export oil products in the LLDCs group. These specific features make oil products different from the rest of merchandise products. Nominal exports have been converted into real terms by deflating them with the annual US import price index for non-oil commodities using the base year 2010 (for all real values in this paper, year 2010 = 100). Among the explanatory variables, real GDP has been measured in US dollars. Distance (DIS) is measured in kilometers and is the distance between the most populated cities (business capitals) of partner countries. Landlockedness (LLOCK) is a binary variable, that is, it takes a value of 1 for LLDCs and 0 for non-LLDCs. A negative sign is expected for this coefficient based on the literature. The GDP of exporting and partner countries is included as per the literature. Language (LAN) is a binary dummy variable, which takes a value of 1 if trading countries have a common official language, and 0 otherwise. Similarly, border (BOR) is a binary dummy variable indicating whether trading countries share a common border. Trade reform (OPEN) is measured as the weighted average tariff rate as it helps to compare the level of openness of a country in terms of international trade. It is proxied by the weighted average tariff rate for all products, and a negative sign is expected, suggesting that the lower the tariff rate, the higher the export performance. We note that weighted non-tariff barriers would be a better proxy for openness, but data for the selected countries are not available. The variables LLOCK, OPEN and AFRICA are the main variables of interest in this study. RER is the bilateral real exchange rate index, which is defined as RERi;t ¼ NERi;t ðPp =Pd Þ. Here NER is the official exchange rate of the partner currency in terms of the domestic currency Pp is measured by the partner’s GDP deflator with base year 2010, as a measure of the world price. Pd is measured by the GDP deflator of exporting countries, constructed by using the relationship between nominal and real GDP, in local currency for the base year 2010, as a measure of domestic prices. Relative factor endowment (RFE) is the absolute difference between the per-capita GDP of importers and exporters. However, this variable is not the main variable of interest in our study, and has been included to show the structure of trade between countries with similar income levels and to identify if the Linder hypothesis or the Heckscher–Ohlin theory is supported.2 If the coefficient on RFE is positive, it would support the Heckscher–Ohlin theory and a negative RFE will support the Linder hypothesis. We include a binary dummy variable (AFRICA), noting concerns raised by economists, such as Collier (2008), Coe and Hoffmaister (1999), and Bosker and Garretsen (2012), on the low developmental progress of Africa, to examine whether African LLDCs are different from other LLDCs. This variable takes a value 1 if the country is in Africa and 0 otherwise. The expected sign of the coefficient of this variable is negative. A binary dummy variable (EUTC)

12

|

PAUDEL

AND

COORAY

is also included to test whether the export performance of transitional landlocked countries in Central and Eastern Europe, which have emerged following the disintegration of the former Soviet Union, is different from that of other landlocked countries. This coefficient could take a positive or negative sign. The model is estimated using a panel data set of bilateral export trade over the period 1995– 2015 for developing countries. Here, the exporters are developing countries, and partners are developed and developing countries. Interaction terms are included to detect possible differences between LLDCs and non-LLDCs. Developed countries are not included as exporters, as the objective of the study is to compare the export performance of LLDCs and non-LLDCs. The focus of this study is solely on merchandise (i.e. non-oil) exports. As listed in Appendix B, the data for exports are collected from World Bank (2016b). Real GDP in US dollars, real GDP, and nominal GDP in local currency, are used to calculate the GDP deflator. The nominal exchange rate, weighted average tariff rate, and per-capita GDP of exporters and partners (GDPPC) are from World Bank (2016a). The nominal exchange rate data for the European Union are collected from the website of the European Central Bank (2012) and converted to US dollars using the nominal exchange rate of the local currency to match the series with other countries. The distance, language and border data were compiled from CEPII (2016). The data for regional trade agreements (RTAs) were collected from De Sousa (2012); these are based on regional trade agreements reported to the World Trade Organization by relevant countries. The data for the later period for RTAs are assumed to be the same, as alterations of the RTAs are not indicated in the database. The data for weighted average tariff rates are for non-oil products and are linearly interpolated.

4.2

| Methodology

Many previous studies have estimated the gravity equations using either pooled ordinary least squares (POLS) estimation, fixed effect (FE) estimation, or random effects (RE) estimation. In this study, our main variables of interest are the landlockedness dummy variable, trade reform (OPEN), the AFRICA dummy variable and its interaction with LLOCK and DIS. All these variables except for OPEN are time-invariant. We cannot estimate the coefficients of time-invariant variables using FE. Therefore, we initially estimate the model with the Poisson pseudo-maximum likelihood (PPML) method as developed in Silva and Tenreyro (2006) and then for purposes of robustness we use pooled OLS, FE and RE. We start off by estimating the basic gravity variables, including LLOCK and OPEN, to detect the impact of landlockedness and trade openness. Then, we estimate the augmented gravity model as shown by equation (2) using PPML estimation, which is more suitable than the other methods for two reasons in our context: (i) it fits well in the semi-log model, so that countries with a small quantity of exports would not be affected by the data; and (ii) it allows us to estimate the coefficients for time-invariant variables (G omez-Herrera, 2013). Moreover, the PPML estimations are elasticities (Genc, 2013).

5

| RESULTS

Table 3 presents results for the basic gravity model including landlockedness and openness for developing countries using PPML estimation techniques. Table 4 presents the results for our

PAUDEL

AND

|

COORAY

13

T A B L E 3 Basic Gravity Model including Landlockedness and Openness Dependent variable: exports

(PPML)

(PPML)

(PPML)

(PPML)

Landlockedness (dummy)

0.261***

4.792***

0.319***

0.497***

GDP (log)

Partner’s GDP (log)

Distance (log)

(0.055)

(0.455)

(0.056)

(0.098)

1.064***

1.064***

1.052***

1.052***

(0.016)

(0.016)

(0.016)

(0.016)

0.881***

0.882***

0.865***

0.865***

(0.017)

(0.017)

(0.017)

(0.017)

0.799***

0.785***

0.749***

0.749***

(0.027)

(0.028)

(0.029)

(0.029)

Openness (tariff rate %)

Distance (log) 3 LLOCK (interaction)

0.058***

0.058***

(0.006)

(0.006)

0.621*** (0.055)

Tariff rates 3 LLOCK

0.036*** (0.013)

Number of observations

257,935

257,935

200,641

200,641

Pseudo R-squared

0.83

0.83

0.82

0.83

RESET test p-values

0.29

0.29

0.28

0.27

Partner effect

Yes

Yes

Yes

Yes

Year effect

Yes

Yes

Yes

Yes

Note: ***, **and *indicate 1%, 5%, and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors under PPML estimation.

augmented gravity model as specified by equation (2). Table 5 presents results for the augmented gravity model for alternative specifications as given by equation (3). Table 6 also presents results for alternative specifications by adding more variables. Appendix C presents results for various specifications under POLS, FE, and RE. Overall, these results suggest that LLDCs have on average about 1 percent lower exports than non-LLDCs. More specifically, the results from Table 3 suggest that, holding the other variables constant, LLDCs export about 0.50 percent less than other developing countries. The costs due to distance are higher for landlocked countries, suggesting that a 1 percent increase in distance in LLDCs leads to an additional 0.50 percent decline in exports, holding other variables in the model constant.3 The results from this table suggest another important finding, That is, reducing tariffs in developing countries by 10 percent helps to increase exports by about 0.6 percent on average. But, when we look at this in the context of LLDCs, this does not hold. The results indicate that lowering the tariff in these countries by 10 percent causes exports to fall by 0.4 percent on average. The results seem to suggest that LLDCs countries may not necessarily benefit from greater openness, so reducing tariffs alone do not have the desired effect on domestic manufacturing sectors. The results for the augmented gravity model, with interaction terms for the landlockedness dummy variable, in Table 4 show that most of the variables have the expected sign. The main gravity variables, such as, GDP, partners’ GDP, distance, border dummy, and common language

14

|

PAUDEL

AND

COORAY

T A B L E 4 Augmented Gravity Model Dependent variable: exports

(PPML)

(PPML 2)

Landlockedness (dummy)

0.309***

13.013***

GDP (log)

Partner’s GDP (log)

Distance (log)

Bilateral RER (log)

Openness (tariff rate %)

RFE (log)

Border (dummy)

Common language (dummy)

(0.054)

(0.878)

1.068***

1.072***

(0.014)

(0.014)

0.855***

0.855***

(0.015)

(0.015)

0.602***

0.593***

(0.019)

(0.019)

0.274***

0.276***

(0.025)

(0.025)

0.091***

0.092***

(0.005)

(0.005)

0.119***

0.124***

(0.017)

(0.017)

0.947***

0.935***

(0.063)

(0.064)

1.087***

1.092***

(0.060)

(0.062)

Interactions

(PPML 2) cont’d  

GDP (log) 3 LLOCK

0.361*** (0.024)

Partners’ GDP (log) 3 LLOCK

0.021 (0.025)

Distance (log) 3 LLOCK

0.317*** (0.053)

Exchange rate 3 LLOCK

0.098 (0.068)

Tariff rate 3 LLOCK

0.001

RFE (log) 3 LLOCK

0.209***

(0.010)

(0.024) Border 3 LLOCK

0.394***

Common language 3 LLOCK

0.276***

(0.090)

(0.101)

Number of observations

158,966

158,966

Pseudo R-squared

0.88

0.88

RESET test p-values

0.28

0.29

Partner effect

Yes

Yes

Year effect

Yes

Yes

Note: ***, ** and * indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are PPML estimation results. The rightmost column is the continuation of results for column (2). LLOCK denotes landlockedness, RFE denotes relative factor endowment.

dummy are statistically highly significant and have the expected sign. The bilateral real exchange rate has a negative and statistically significant sign which is not consistent with our expectations; however, this variable is not out main variable of interest. The reason could be the heavy share of imported technology and raw material in exports. But it has the expected sign and is statistically significant in the tables in Appendix C. Therefore, the result for this variable is ambiguous. The interaction term for GDP and landlockedness shows that GDP in landlocked countries has not contributed positively to boost export performance; in fact, this coefficient is negative and statistically highly significant. The coefficient on relative factor endowment is not significant. The coefficients for common language and common border are statistically significant. Having a common border enables an LLDC to export more, assuming that other variables remain constant. Table 5 presents results for the basic gravity model with alternative specifications. In Table 4, we estimate openness by excluding RTAs. In Table 5, we include a dummy variable for RTAs to identify their impact on the export performance of developing countries. We do not find a

PAUDEL

AND

|

COORAY

15

T A B L E 5 Augmented Gravity Model (Alternative Specification) Dependent variable: exports

(PPML)

(PPML 2)

Landlockedness (dummy)

0.092*

7.011***



(0.053)

(0.845)



1.081***

1.080***

(0.012)

(0.012)

0.849***

0.848***

(0.015)

(0.015)

0.539***

0.529***

(0.019)

(0.020)

0.195***

0.194***

(0.023)

(0.024)

0.363***

0.373***

GDP (log)

Partner’s GDP (log)

Distance (log)

Bilateral RER (log)

Per-capita GDP (log)

Interactions

GDP (log) 3 LLOCK Partners’ GDP (log) 3 LLOCK Distance (log) 3 LLOCK

Per-capita GDP (log) 3 LLOCK

(0.022)

(0.022)

Partners’ per-capita GDP (log) 3 LLOCK

0.097***

0.097***

Tariff rate 3 LLOCK

(0.004)

(0.004)

0.021

0.025

(0.016)

(0.017)

0.872***

0.867***

(0.055)

(0.056)

Common language (dummy)

0.811***

0.815***

(0.053)

(0.055)

RTA (dummy)

0.584***

0.575***

(0.039)

(0.040)

0.284***

0.284***

(0.058)

(0.061)

0.520***

0.543***

(0.039)

(0.041)

Africa (dummy)

Post-Soviet countries

0.151*** (0.053)

(0.023)

Border (dummy)

0.313*** (0.041)

Exchange rate 3 LLOCK

0.127***

RFE(log)

0.003 (0.024)

0.125***

Openness (tariff rate %)

0.407*** (0.036)

(0.023) Partner’s per-capita GDP (log)

(PPML 2) cont’d

0.683*** (0.062) 0.064 (0.039) 0.021 (0.014)

RFE (log) 3 LLOCK

0.123*** (0.031)

Border 3 LLOCK

0.207** (0.087)

Common language 3 LLOCK

0.544*** (0.108)

RTA 3 LLOCK

0.522*** (0.090)

African landlocked (dummy)

1.222*** (0.121)

Post-Soviet landlocked (dummy)

0.511*** (0.101)

Number of observations

158,966

158,966

Pseudo R-squared

0.88

0.88

RESET test p-values

0.28

0.29

Partner effect

Yes

Yes

Year effect

Yes

Yes

Note: ***, ** and * indicate 1%, 5%, and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are PPML estimation results. The rightmost column is the continuation of the results for column (2). LLOCK denotes landlockedness, RFE denotes relative factor endowment, RTA denotes regional trade agreement.

16

|

PAUDEL

AND

COORAY

T A B L E 6 Augmented Gravity Model (Alternative Specification) Dependent variable: exports

(PPML)

(PPML 2)

Landlockedness (dummy)

0.317***

10.295***



(0.050)

(0.894)



1.105***

1.105***

(0.012)

(0.012)

0.852***

0.852***

(0.014)

(0.015)

0.512***

0.501***

(0.018)

(0.018)

0.235***

0.234***

(0.023)

(0.023)

0.420***

0.429***

GDP (log)

Partner’s GDP (log)

Distance (log)

Bilateral RER (log)

Per-capita GDP (log)

(0.022)

(0.023)

Partner’s per-capita GDP (log)

0.130***

0.132***

(0.021)

Openness (tariff rate %)

Interactions

GDP (log) 3 LLOCK

(PPML 2) cont’d

0.494*** (0.045)

Partners’ GDP (log) 3 LLOCK

0.029 (0.024)

Distance (log) 3 LLOCK

0.250*** (0.043)

Exchange rate 3 LLOCK

0.118** (0.056)

Per-capita GDP (log) 3 LLOCK

0.640*** (0.070) 0.079*

(0.022)

Partner’s per-capita GDP (log) 3 LLOCK

(0.040)

0.101***

0.101***

Tariff rate 3 LLOCK

0.038***

(0.005)

(0.005)

0.017

0.020

(0.015)

(0.016)

0.860***

0.854***

(0.054)

(0.055)

Common language (dummy)

0.798***

0.804***

(0.052)

(0.054)

RTA (dummy)

0.678***

0.670***

(0.037)

(0.038)

RFE (log)

Border (dummy)

(0.013) RFE (log) 3 LLOCK

0.115*** (0.030)

Border 3 LLOCK

0.190** (0.087)

Common language 3 LLOCK

0.272** (0.117)

RTA 3 LLOCK

0.345*** (0.085)

Number of observations

158,966

158,966

Pseudo R-squared

0.88

0.88

RESET test p-values

0.28

0.29

Partner effect

Yes

Yes

Year effect

Yes

Yes

Note: ***, ** and * indicate 1%, 5%, and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are PPML estimation results. The rightmost column is the continuation of the results for column (2). LLOCK denotes landlockedness, RFE denotes relative factor endowment, RTA denotes regional trade agreement.

difference when including or excluding RTAs in the estimations. The RTA variable is highly statistically significant when interacted with landlockedness, which indicates that one way of improving the export performance of LLDCs is to enter into bilateral and regional trade agreements. The results for the rest of the variables are consistent with the results reported in previous tables, but the coefficients have changed slightly in magnitude.

PAUDEL

AND

COORAY

|

17

The coefficient on AFRICA is negative and statistically significant. This suggests that African developing countries, on average, have about 0.30 percent lower exports than developing countries in other regions, other things remaining the same. In this estimation, the results are consistent with those of previous studies such as Coe and Hoffmaister (1999). If we compare African developing countries with other developing countries, the export performance of African developing countries is poor. But, if we compare African LLDCs with other developing countries, African LLDCs, have on average, higher export levels compared to the average level for other LLDCs. This could be due to the benefits of relatively strong regional cooperation, as discussed by Faye et al. (2004). A similar story emerges in the case of the Eastern European transition countries, which are landlocked. Table 6 presents result for our augmented gravity model including per-capita GDP and partners’ per-capita GDP and interaction variables with landlockedness. We note that some studies do not estimate these variables with GDP. Therefore, in the previous tables we undertook estimation without these variables. We include these variables in this estimation as proxies for importers’ and exporters’ infrastructure quality and purchasing power of consumers (the quality of the market). The results suggest that exporters’ per-capita GDP is an important factor for export performance in both groups of developing countries. But when it comes to partners’ per-capita GDP for landlocked countries, the sign turns negative. This change of sign indicates that landlocked countries lose market share as their partners’ income status increases, suggesting that policy-makers should focus more on quality products as their partners’ income level increases. The results show that the estimation for the main variables of interest is consistent with those of previous tables (Tables 3–5). Results for the landlockedness dummy variable remain unchanged, maintaining the same level of statistical significance. Variables such as openness, regional trade agreements, common border, common language, and distance also have the same level of statistical significance and expected signs, although the coefficients are slightly different in magnitude.

5.1

| Robustness checks

Next, we test whether the results are consistent with alternative specifications and alternative estimation methods. For this, we test the model using POLS, RE, and FE estimation methods. The results are reported in Tables 1–5C–C in Appendix C. Estimation with these methods also shows that LLDCs have a greater disadvantage due to distance, by at least 0.30 percent. Results for the bilateral exchange rate here are as expected, and contrary to previous results. Thus, the impact of the bilateral exchange rate seems to be ambiguous as the results are not consistent among the PPML and POLS, RE, and FE estimation methods. Table 1C presents results for the basic gravity model including landlockedness for developing countries. The results for POLS are reported in columns (1) and (2) and for RE estimation in columns (3) and (4). These results suggest that LLDCs have, on average, about 1 percent lower exports than non-LLDCs. Table 2C presents the results for the basic gravity model variables plus landlockedness and openness. The coefficient on tariffs show that a 1 percent decrease in tariff rates causes an increase in exports by 0.20 percent on average if other factors remain constant. Table 3C presents results for the basic gravity model including landlockedness and interaction terms employing random effects and FE estimation. The dummy variable landlockedness times distance has been dropped in the FE estimations. The RE results are consistent with the results reported in the previous tables but the coefficients have changed slightly in magnitude. Table 4C presents results for our augmented gravity model and interaction variables with landlockedness using RE. The results for the landlockedness dummy show that the export performance

18

|

PAUDEL

AND

COORAY

of LLDCs is on average about 0.60 percent less, which is consistent with the literature. The results for openness have the expected sign, suggesting that on average a 1 percent decrease in the tariff rate results in an increase in exports by 0.01 percent in non-LLDCs and about 0.04 percent in LLDCs.4 These results confirm that trade reform is important in both sets of developing countries, but it shows that it has a higher impact on the export performance of LLDCs, indicating that reducing tariffs may be a way to minimize the effects of landlockedness. The results are consistent with the view that in general trade liberalization promotes exports. The bilateral real exchange rate has a positive and statistically significant impact on exports, suggesting that the depreciation of the domestic currency promotes exports in both sets of developing countries. The coefficients on exporter and partner GDP are highly significant. Distance has a statistically highly significant negative impact, as expected: on average the negative impact is about 1.3 percent on the export performance of developing countries. The coefficient on relative factor endowment is not significant. Regional trade agreements contribute more to LLDCs than to non-LLDCs, but they have a statistically significant impact on the export performance of both groups of developing countries. The coefficient estimates for common language and common border are statistically significant. Having a common border enables a developing country to export more, assuming other variables remain constant. The coefficient on AFRICA is negative and statistically significant. This suggests that African developing countries have on average about 0.75 percent lower exports than developing countries in other regions, other things remaining the same. These results are much stronger than those of the PPML estimation. If we compare African developing countries with other developing countries, the export performance of African developing countries is poor. But if we compare African LLDCs with other developing countries, African LLDCs have on average higher export levels than the average level for other LLDCs. This means that LLDCs in these regions are better off than LLDCs in other regions. This could be due to the benefits of relatively strong regional cooperation, as discussed by Faye et al. (2004). A similar story emerges in the case of the Eastern European transition countries, which are landlocked. These results are again consistent with those of Table 5. Table 5C presents results for our augmented gravity model and interaction of variables with landlockedness under RE for alternative specifications. Here, we include exporters’ per-capita GDP and partner per-capita GDP as a proxy for infrastructure. The results are broadly consistent with those of Table 6, particularly the results for the main variables of interest. To summarize, the results suggest that landlockedness limits the export performance of countries compared to those that are not landlocked. Distance has a larger negative effect on the export performance of these countries. Trade openness does not have the same positive effects on landlocked countries as it does on non-landlocked countries. The evidence does not suggest that African landlocked countries are at a greater disadvantage than LLDCs in other regions.

6

| CONCLUSION

This paper has examined the determinants of export performance of developing countries from a comparative perspective: landlocked versus non-landlocked countries. The results suggest that although LLDCs have been making some progress in expanding exports in recent decades, their export performance remains low compared to other developing countries. While landlockedness remains a constraint, the results indicate that these countries can improve their export performance by creating a more trade-friendly environment through lowering tariffs and entering into regional

PAUDEL

AND

|

COORAY

19

trade agreements. Both demand- and supply-side factors play a significant role in determining the export performance of LLDCs, as indicated by their own and partner GDPs. The results for the distance variables suggest that distance-related trade costs restrict export performance more in LLDCs than in non-LLDCs. This situation requires urgent attention to be paid to building sustainable infrastructure connectivity in LLDCs. Arvis et al. (2007) observe that many landlocked countries depend on one or two routes through “transit neighbors” to undertake international trade. As noted by them, apart from building infrastructure, it is also important to implement proper policies that regulate transport service providers and trade of goods and services along corridors. This requires governments and private stakeholders to undertake discussions on implementing proper regulatory structures for international transit through cost reduction, service delivery, and greater competition and efficiency. We do not find any evidence to suggest that African landlocked countries are at a greater disadvantage than LLDCs in other regions. On the contrary, ceteris paribus, the average export levels for these countries are about 100 percent higher than the average exports level of other LLDCs. This result perhaps reflects the success of reforms undertaken by a few of these countries since the early 1990s, the impact of which has not been adequately captured by the explanatory variables used in the model. Trade liberalization does not have the same beneficial effects on LLDCs as on non-LLDCs. These countries need to find potential export avenues, such as becoming more involved in global production sharing networks, product specialization, and building strong infrastructure relative to the size of their economies. ACKNOWLEDGEMENTS We wish to thank the editor of the journal, an anonymous referee, and Prema-Chandra Athukorala, Peter Warr, Paul Burke, Max Corden, and Ronald Findlay for their comments on an earlier version of this paper. The first author wishes to thank three anonymous thesis examiners. We also thank participants at presentations at the Australian National University. Needless to say, any remaining errors and mistakes are solely ours. ENDNOTES 1

These countries are Armenia, Azerbaijan, Belarus, Kazakhstan, Kyrgyz Republic, Moldova, Tajikistan, Turkmenistan and Uzbekistan (Idan & Shaffer, 2011). 2 The Heckscher–Ohlin hypothesis suggests that more trade occurs if country endowment levels are different. On the other hand, a negative sign for this variable would support the Linder (1961) hypothesis, which suggests that the different levels of endowment affect trade negatively, meaning that more trade occurs where countries are in almost the same income category. 3 The coefficient for the dummy variable is: exp(c) – 1, where c is the estimated coefficient; that is, exp(– 0.621) – 1 = –0.46 for the interaction (distance (log) 9 LLOCK) term in Table 3. 4 To calculate the coefficients for LLDCs, the sum of the coefficients in equation 2 with the respective interaction variables is used. For example, for openness, –0.008 – 0.029= –0.037.

ORCID Ramesh C. Paudel

http://orcid.org/0000-0001-7721-3205

20

|

PAUDEL

AND

COORAY

REFERENCES Arvis, J.-F., Raballand, G., & Marteau, J.-F. (2007). The cost of being landlocked: Logistics costs and supply chain reliability (Policy Research Working Paper 4258). Washington, DC: World Bank. Athukorala, P.-C. (2011). Production networks and trade patterns in East Asia: Regionalization or globalization? Asian Economic Papers, 10(1), 65–95. Awokuse, T. O. (2008). Trade openness and economic growth: Is growth export-led or import-led? Applied Economics, 40(2), 161–173. Balassa, B. (1985). Exports, policy choices, and economic growth in developing countries after the 1973 oil shock. Journal of Development Economics, 18(1), 23–35. Baumol, W. J., & Vinod, H. D. (1970). An inventory theoretic model of freight transport demand. Management Science, 16(7), 413–421. Behar, A., & Venables, A. J. (2011). Transport costs and international trade. In R. L. Andre de Palma, Emile Quinet, & Roger Vickerman (Eds.), A handbook of transport economics (pp. 97–115). Cheltenham: Edward Elgar. Bergeijk, P. A. G. V., & Brakman, S. (Eds.) (2010). The gravity model in international trade: Advances and applications. Cambridge: Cambridge University Press. Bosker, M., & Garretsen, H. (2012). Economic geography and economic development in sub-Saharan Africa. World Bank Economic Review, 26(3), 443–485. CEPII (2016). Gravity [data file]. Retrieved from http://www.cepii.fr/anglaisgraph/bdd/gravity.asp Clarke, G. R. (2005). Beyond tariffs and quotas: Why don’t African manufacturers export more? (Policy Research Working Paper 3617). Washington, DC: World Bank. Coe, D. T., & Hoffmaister, A. W. (1999). North-South trade: Is Africa unusual? Journal of African Economies, 8 (2), 228–256. Collier, P. (2008). The bottom billion: Why the Poorest countries are failing and what can be done about it. Oxford: Oxford University Press. De Sousa, J. (2012). The currency union effect on trade is decreasing over time. Economics Letters, 117(3), 917–920. Djankov, S., Freund, C., & Pham, C. S. (2010). Trading on time. Review of Economics and Statistics, 92(1), 166–173. European Central Bank (2012). Conversion rates from former national currency. Retrieved from http://www.ecb.int/ euro/intro/html/index.en.html Faye, M. L., McArthur, J. W., Sachs, J. D., & Snow, T. (2004). The challenges facing landlocked developing countries. Journal of Human Development, 5, 31–68. Genc, M. (2013). Migration and tourism flows to New Zealand. In A. Matias, P. Nijkamp, & M. Sarmento (Eds.), Quantitative methods in tourism economics (pp. 113–128). Heidelberg: Physica-Verlag. GFDatabase (2011). Global Financial Database. Retrieved from https://www.globalfinancialdata.com/Databases/ GFDatabase.html Gomez-Herrera, E. (2013). Comparing alternative methods to estimate gravity models of bilateral trade. Empirical Economics, 44(3), 1087–1111. Greenaway, D., Morgan, W., & Wright, P. (2002). Trade liberalisation and growth in developing countries. Journal of Development Economics, 67(1), 229–244. https://doi.org/10.1016/S0304-3878(01)00185-7. Grigoriou, C. (2007). Landlockedness, infrastructure and trade: New estimates for central asian countries (Policy Research Working Paper 4335). Washington, DC: World Bank. Idan, A. & B. Shaffer (2011). The Foreign Policies of Post-Soviet Landlocked States. Post-Soviet Affairs 27(3), 241–268. Krueger, A. O. (1990). Asian trade and growth lessons. American Economic Review, 80(2), 108–112. Krugman, P., Cooper, R. N., & Srinivasan, T. (1995). Growing world trade: Causes and consequences. Brookings Papers on Economic Activity, 1995(1), 327–377. Krugman, P. R. (2008). Trade and wages, reconsidered. Brookings Papers on Economic Activity, 2008(1), 103–154. Lim~ao, N., & Venables, A. J. (2001). Infrastructure, geographical disadvantage, transport costs, and trade. World Bank Economic Review, 15(3), 451–479. https://doi.org/10.1093/wber/15.3.451. Linder, S. B. (1961). An essay on trade and transformation. Uppsala: Almqvist & Wiksell. Martinez, M., & Mlachila, M. (2013). The quality of the recent high-growth episode in sub-Saharan Africa (IMF Working Paper, WP/13/53). Washington, DC: International Monetary Fund.

PAUDEL

AND

|

COORAY

21

Paudel, R. C. (2014). Economic growth in developing countries: Is landlockedness destiny? Economic Papers, 33 (4), 339–361. https://doi.org/10.1111/1759-3441.12096. Raballand, G. (2003). Determinants of the negative impact of being landlocked on trade: An empirical investigation through the Central Asian case. Comparative Economic Studies, 45(4), 520–536. Radelet, S., & Sachs, J. (1998). Shipping costs, manufactured exports, and economic growth. Paper presented at the American Economic Association Meetings, Harvard University, Cambridge, MA. Sachs, J. D., & Warner, A. (1995). Economic reform and the process of global integration. Brookings Papers on Economic Activity, 1, 1–118. Santos-Paulino, A., & Thirlwall, A. P. (2004). The impact of trade liberalisation on exports, imports and the balance of payments of developing countries. Economic Journal, 114(493), F50–F72. Silva, J. S., & Tenreyro, S. (2006). The log of gravity. Review of Economics and Statistics, 88(4), 641–658. Stiglitz, J. E. (1996). Some lessons from the East Asian Miracle. World Bank Research Observer, 11(2), 151–177. Tinbergen, J. (1962). Shaping the world economy: Suggestions for an international economic policy. New York: Twentieth Century Fund. US Department of Labor, Bureau of Labor Statistics (2016). Database, tables & calculators by subject. Retrieved from http://data.bls.gov Wacziarg, R., & Welch, K. H. (2008). Trade liberalization and growth: New evidence. World Bank Economic Review, 22(2), 187–231. https://doi.org/10.1093/wber/lhn007. Weiss, J. (1999). Trade reform and manufacturing performance in Mexico: From import substitution to dramatic export growth. Journal of Latin American Studies, 31(01), 151–166. World Bank (2016a). World Development Indicators. Washington, DC: World Bank. Retrieved from https://da ta.worldbank.org/products/wdi World Bank (2016b). WITS: World Integrated Trade Solution [database]. Retrieved from http://wits.worldbank.org/ WITS/WITS/Default-A.aspx?Page=Default

How to cite this article: Paudel RC, Cooray A. Export performance of developing countries: Does landlockedness matter?. Rev Dev Econ. 2018;00:1–27. https://doi.org/ 10.1111/rode.12389

APPENDIX A: LIBERALIZATION STATUS IN LANDLOCKED DEVELOPING COUNTRIES Updated SachsWarner criteria of liberalization for 1999–2009 Year of opening

Av. tariff %

NTB rate %

BM prm. %

Exp. Mkt. Board

Socialist State

Lao PDR



11.3

na

na

0

0

Mongolia

1997

4.8

0

0

0

0

Armenia

1995

2.2

0

0

0

0

Azerbaijan

1995

4.9

0

0

0

0

Belarus



6.3

na

0

0

0

Kazakhstan

–

4.4

na

na

0

0

Region/country EAP

ECA

(Continues)

22

|

PAUDEL

AND

COORAY

T A B L E A (Continued) Updated SachsWarner criteria of liberalization for 1999–2009 Region/country

Year of opening

Av. tariff %

NTB rate %

BM prm. %

Exp. Mkt. Board

Socialist State

Kosovo

–

na

na

na

0

0

Kyrgyz Republic

1994

4.3

0

0

0

0

Macedonia, FYR

1994

5.3

0

0

0

0

Moldova

1994

2.3

0

0

0

0

Serbia

–

6.6

na

na

0

0

Tajikistan

1996

5.3

0

0

0

0

Turkmenistan

–

1.4

na

na

0

0

Uzbekistan

–

6.6

na

0

0

0

Bolivia

1985

7.5

0

0

0

0

Paraguay

1989

7.7

0

0

0

0

LAC

SA Nepal

1991

15

0

0

0

0

Bhutan

–

18

na

0

0

0

Afghanistan

–

5.5

na

22

0

0

Botswana

1979

7.9

0

0

0

0

Burkina Faso

1998

11.2

0

0

0

0

Burundi

1999

13.2

0

0

0

0

CAR

–

15.5

na

0

1

0

Chad

2001

14.1

0

0

0

0

Ethiopia

1996

12.6

0

0

0

0

Lesotho

2001

15.3

0

0

0

0

Malawi

2001

13.1

0

0

0

0

Mali

1988

9.8

0

0

0

0

Niger

1994

11.1

0

0

0

0

Rwanda

2001

12.5

0

0

0

0

SSA

Swaziland

2001

7

0

0

0

0

Uganda

1988

7.7

0

0

0

0

Zambia

1993

9.3

0

0

0

0

Zimbabwe

–

20.3

0

29

0

0

Note: (1) Updated Sachs–Warner criteria (a country is liberalized when it has non-tariff barrier rate (NTB) of no more than 40%, average tariff rate of no more than 40%, black market exchange rate of no more than 20%, and does not have an export marketing board and socialist economic system). (2) “na” denotes not available, but it is believed the figures exceed the given criteria making these countries remain closed. (3) Av., CAR, BM prm., and Exp. Mkt. stand for average, Central African Republic, black market premium, and export market. “–” denotes closed. Source: Sachs and Warner (1995), Wacziarg and Welch (2008) and GFDatabase (2011).

PAUDEL

AND

|

COORAY

23

APPENDIX B: LIST OF VARIABLES, DATA SOURCES AND EXPECTED SIGN OF COEFFICIENT Variables

Details and expected sign

Data source

X

Real non-oil exports between trading countries, the dependent variable, converted to real deflating by US import price index for non-oil products

World Bank (2016b) and United States Department of Labor (2016)

LLOCK

Landlockedness, binary dummy (–)

OPEN

Openness measured by weighted average tariff rate (–)

World Bank (2016b)

GDP

Real gross domestic product, size of economy (+)

World Bank (2016a)

DIS

The distance between business cities of partners (–)

CEPII (2016)

RER

Real exchange rate (domestic currency/US$) (+)

World Bank (2016a) and ECB (2012)

GDPPC

Per-capita GDP of exporters and partners (+)

World Bank (2016a)

AFRICA

If the country is in Africa, binary dummy (–)

LAN

Common language, cultural affinity (+)

CEPII (2016)

BOR

Common border of trading countries (+)

CEPII (2016)

RFE

Relative factor endowment, either Heckscher–Ohlin or Linder hypothesis (+, –)

World Bank (2016a)

RTA

Regional trade agreements, binary dummy (+)

De Sousa (2012)

EUTC

Eastern European transition countries, binary dummy (+/–)

APPENDIX C

T A B L E C 1 Basic Gravity Model including Landlockedness Dependent variable: exports (log)

(POLS)

(POLS)

(RE)

(RE)

Landlockedness (dummy)

1.048***

1.410***

1.177***

0.429

GDP (log)

Partner’s GDP (log)

Distance (log)

(0.013)

(0.145)

(0.040)

(0.470)

1.063***

1.063***

1.011***

1.012***

(0.003)

(0.003)

(0.007)

(0.007)

0.960***

0.960***

1.004***

1.003***

(0.002)

(0.002)

(0.007)

(0.007)

1.421***

1.366***

1.435***

1.396***

(0.007)

(0.007)

(0.022)

(0.025)

Distance (log) 3 LLOCK (interaction)

0.287***

0.185***

(0.017)

(0.054)

Number of observations

257,935

257,935

257,935

257,935

F-statistics

81,258

65,138

R-squared

0.56

0.56

0.56

0.56

Partner effect

Yes

Yes

Yes

Yes

Note: ***, ** and *indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are pooled ordinary least squares, and columns (3) and (4) are random effect estimation results. LLOCK denotes landlockedness.

24

|

PAUDEL

AND

COORAY

T A B L E C 2 Basic Gravity Model including Landlockedness and Openness Dependent variable: exports (log)

(POLS)

(POLS)

(RE)

(RE)

Landlockedness (dummy)

0.999***

1.031***

1.072***

0.163

GDP (log)

Partner’s GDP (log)

Distance (log)

Openness (tariff rate %)

(0.015)

(0.160)

(0.043)

(0.492)

1.088***

1.087***

1.074***

1.074***

(0.003)

(0.003)

(0.008)

(0.008)

0.985***

0.985***

1.034***

1.034***

(0.003)

(0.003)

(0.007)

(0.007)

1.466***

1.420***

1.502***

1.479***

(0.007)

(0.008)

(0.023)

(0.026)

0.029***

0.028***

0.016***

0.016***

(0.001)

(0.001)

(0.001)

(0.001)

Distance (log) 3 LLOCK (interaction)

0.236***

0.105*

(0.019) Number of observations

200,641

200,641

F-statistics

52,348

43,685

R-squared

0.57

Partner effect

Yes

(0.056) 200,641

200,641

0.57

0.57

0.57

Yes

Yes

Yes

Note: ***, ** and * indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are pooled ordinary least squares, and columns (3) and (4) are random effect estimation results. LLOCK denotes landlockedness.

PAUDEL

AND

|

COORAY

25

T A B L E C 3 Basic Gravity Model including Landlockedness Interaction Dependent variable: exports (log)

(RE)

(RE)

(FE)

(FE)

Landlockedness (dummy)

17.722***

18.659***

(dropped)

(dropped)

GDP (log)

Partner’s GDP (log)

Distance (log)

Distance (log) 3 LLOCK (interaction)

(0.700)

(0.692)

1.096***

1.051***

1.114***

1.289***

(0.007)

(0.007)

(0.022)

(0.031)

1.043***

0.999***

1.366***

1.501***

(0.007)

(0.008)

(0.024)

(0.029)

1.395***

1.395***

(dropped)

(dropped)

(0.025)

(0.024) (dropped)

(dropped)

1.122***

0.187***

0.192***

(0.054)

(0.053)

0.649***

0.716***

1.070***

(0.020)

(0.020)

(0.040)

(0.041)

0.096***

0.071***

0.191***

0.183***

(0.017)

(0.016)

(0.056)

(0.056)

Number of observations

257,935

257,935

257,935

257,935

R-squared

0.56

0.56

0.18

0.18

Partner effect

Yes

Yes

Yes

Yes

Year effect

No

Yes

No

Yes

GDP (log) 3 LLOCK (interaction) Partner’s GDP (log) 3 LLOCK (interaction)

Note: ***, ** and * indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are random effect, and columns (3) and (4) are fixed effect estimation results. LLOCK denotes landlockedness.

26

|

PAUDEL

AND

COORAY

T A B L E C 4 Augmented Gravity Model Dependent variable: exports (log)

(RE)

(RE 2)

Landlockedness (dummy)

0.825***

17.932***

GDP (log)

Partner’s GDP (log)

Distance (log)

Bilateral RER (log)

Openness (tariff rate %)

RFE (log)

Border (dummy)

(0.054)

(1.054)

1.078***

1.133***

(0.012)

(0.011)

1.040***

1.050***

(0.010)

(0.010)

1.294***

1.294***

(0.029)

(0.031)

0.107***

0.113***

(0.016)

(0.010)

0.012***

0.008***

(0.002)

(0.001)

0.017

0.018**

(0.012)

(0.009)

1.136***

0.991***

(0.128)

(0.155)

1.235***

1.299***

(0.057)

(0.060)

0.238***

0.156***

(0.035)

(0.025)

Africa (dummy)

0.632***

0.746***

(0.050)

(0.050)

PostSoviet countries

0.408***

0.215**

(0.068)

(0.098)

Common Language (dummy)

RTA (dummy)

Interactions

(RE 2)cont’d  

GDP (log) 3 LLOCK

0.737*** (0.054)

Partners’ GDP (log) 3 LLOCK

0.036*** (0.028)

Distance (log) 3 LLOCK

0.015*** (0.077)

Exchange rate 3 LLOCK

0.062*** (0.038)

Tariff rates 3 LLOCK

0.029*** (0.007)

RFE (log) 3 LLOCK

0.017*** (0.032)

Border 3 LLOCK

0.269 (0.305)

Common language 3 LLOCK

0.312*** (0.151)

RTA 3 LLOCK

0.574*** (0.156)

African landlocked (dummy)

0.534***

Post-Soviet landlocked (dummy)

0.182***

(0.121)

(0.150)

Number of observations

158,966

158,966

R-squared

0.58

0.58

Partner effect

Yes

Yes

Year effect

Yes

Yes

Note: ***, ** and * indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are random effect estimation results. The rightmost column is the continuation of the results for column (2). LLOCK denotes landlockedness, RFE denotes relative factor endowment, RTA denotes regional trade agreement.

PAUDEL

AND

|

COORAY

27

T A B L E C 5 Augmented Gravity Model (alternative specification) Dependent variable: exports (log)

(RE)

(RE 2)

Landlockedness (dummy)

0.783***

17.871***

-

(0.057)

(1.538)

-

1.069***

1.118***

(0.011)

(0.012)

1.015***

1.017***

(0.011)

(0.012)

1.292***

1.288***

GDP (log)

Partner’s GDP (log)

Distance (log)

(0.029)

(0.032)

0.107***

0.112***

(0.016)

(0.017)

Percapita GDP (log)

0.068***

0.131***

(0.024)

(0.026)

Partner’s per-capita GDP (log)

0.083***

0.103***

(0.020)

Bilateral RER (log)

Interactions

GDP (log) 3 LLOCK

(RE 2) cont’d

0.788*** (0.057)

Partners’ GDP (log) 3 LLOCK

0.010 (0.030)

Distance (log) 3 LLOCK

0.007

Exchange rate 3 LLOCK

0.052

(0.076)

(0.038) Per-capita GDP (log) 3 LLOCK

0.048 (0.065) 0.076

(0.022)

Partners’ per-capita GDP (log) 3 LLOCK

0.012***

0.007***

Tariff rates 3 LLOCK

0.029***

(0.002)

(0.002)

0.040***

0.047***

(0.014)

(0.014)

1.190***

1.066***

(0.128)

(0.154)

Common Language (dummy)

1.221***

1.277***

(0.057)

(0.063)

RTA (dummy)

0.236***

0.151***

(0.035)

(0.035)

Openness (tariff rate %)

RFE (log)

Border (dummy)

Africa (dummy)

Post-Soviet countries

0.568***

0.626***

(0.055)

(0.062)

0.421***

0.232***

(0.069)

(0.082)

(0.061)

(0.007) RFE (log) 3 LLOCK

0.036 (0.043)

Border 3 LLOCK

0.199 (0.304)

Common language 3 LLOCK

0.319**

RTA 3 LLOCK

0.570***

(0.149)

(0.156) African landlocked (dummy)

0.529*** (0.129)

Post-Soviet landlocked (dummy)

0.138 (0.151)

Number of observations

158,966

158,966

R-squared

0.58

0.58

Partner effect

Yes

Yes

Year effect

Yes

Yes

Note: ***, ** and * indicate 1%, 5% and 10% level of statistical significance, respectively. The figures in parentheses are robust standard errors. Columns (1) and (2) are random effect estimation results. The rightmost column is the continuation of the results for column (2). LLOCK denotes landlockedness, RFE denotes relative factor endowment, RTA denotes regional trade agreement.