Pomfret and Sourdin (2010). Why Do Trade Costs Vary

8/17/2019 Pomfret and Sourdin (2010). Why Do Trade Costs Vary

1/22

O R I G I N A L P A P E R

Why do trade costs vary?

Richard Pomfret • Patricia Sourdin

Published online: 14 September 2010 Kiel Institute 2010

Abstract As tariffs have fallen, it is apparent that trade costs are a significant

obstacle to international trade and that they vary from country to country. The gap

between the cif and fob value of a trade flow is a useful measure of aggregate trade

costs, but only if the measure is based on a consistent volume of trade; mirror

statistics are unsuitable. Using high quality Australian import data disaggregated at

the HS 6-digit level, we find large country-by-country variations in trade costs.

Distance, weight and size account for part of the variation in trade costs. Indicatorsof institutional quality pick up some of the variation in trade costs, but the rela-

tionship is not uniform across mode of transport and commodities; exporting

countries’ institutional quality is more strongly related to trade costs for air freight

than sea freight, and the relationship is commodity-specific and strongest for

manufactured goods. Country-specific characteristics influencing trade costs pro-

vide a link between institutions and economic development.

Keywords Trade costs Trade facilitation

JEL Classification F10 F13 O24

R. Pomfret (&) P. Sourdin

School of Economics, University of Adelaide, Adelaide, SA 5005, Australiae-mail: [email protected]

P. SourdinThe Johns Hopkins University Bologna Center,via Belmeloro 11, 40126 Bologna, Italy

e-mail: [email protected]

1 3

Rev World Econ (2010) 146:709–730DOI 10.1007/s10290-010-0072-8


2/22

1 Introduction

Transport and related trade costs are often viewed as technologically determined,

but in practice they vary considerably across different bilateral trade flows. Some of

the variation is due to distance and other geographical constraints and some reflectscommodity composition of trade. However, port infrastructure, corrupt customs

officials and other ‘trade costs’ are policy-related trade barriers, while other

determinants of trade costs may be indirectly policy-related (e.g. lack of

competition among shippers may be due to low volumes or to non-implementation

of anti-monopoly policy). Country variations related to institutions such as poor law

enforcement increase trade risks and hence affect insurance rates and inventory

costs. This paper aims to get inside the black box of measured trade costs, to

understand which are policy-related (and can be reduced by trade facilitation

measures) and which are exogenously determined. After discussing how to measuretrade costs, the paper addresses the question: why do trade costs vary?

The missing trade mystery and literature on the border effect suggest significant

trade costs, but we have little direct information on the size of trade costs and only

limited evidence on their determinants.1 Anderson and van Wincoop (2004)

highlighted the potential significance of trade costs, with estimates that in the high-

income countries trade costs amount on average to a 170% ad valorem barrier to

trade. However, they use a very broad definition of trade costs, i.e. all costs of

getting a good to the final user apart from the marginal cost of producing the good

itself, and the estimates relied on indicative case studies or indirect evidence fromgravity models. Direct measures of trade costs, such as the World Customs

Organization’s Time Release Studies are more informative, but too narrow and have

been done for too few countries.

The gap between free-on-board (fob) values when a good reaches the port of exit

in the exporting country and import values which include cost, insurance and freight

(cif) provides an economically meaningful and operational measure of international

trade costs.2 The cif-fob gap is an economically meaningful measure of the wedge

between the cost of producing and moving a good to the exporter’s port and the

price paid by the importer upon the good’s arrival in the destination country. The

measure has, however, been difficult to implement because mirror techniques,

matching fob values reported by exporting countries to cif values reported by

importing countries, are unsatisfactory due to large measurement errors (Hummels

and Lugovskyy 2006). Nevertheless, some national statistical offices now collect

consistent data on fob and cif values at disaggregated levels, making the cif-fob

1 Despite large reductions in tariffs and other barriers to trade since 1947, levels of international trade areless than would be expected from relative factor endowments (Trefler 1995). Even across frontiers as

open as that between the USA and Canada, trade between a state and a province is less than between twoUS states or two Canadian provinces ceteris paribus.2 Measurement is important for policy as well as for testing theories. Trade facilitation is included in theDoha Development Round of multilateral trade negotiations, and has featured increasingly prominently inregional trade agreements (Pomfret and Sourdin 2009). In 2001 Asia–Pacific Economic Cooperation(APEC) members adopted a goal of reducing trade costs by five percent over 5 years, and thecommitment was repeated in 2006, although without an agreed measure of trade costs it is difficult to

monitor progress towards such a goal.

710 R. Pomfret, P. Sourdin

1 3


3/22

price gap operationally useful. In this paper we utilize such data for Australian

imports since 1990 at the 6-digit HS level to measure cross-country differences in

trade costs.

Of the few countries collecting consistent cif and fob data, Australia is

particularly well-suited to this exercise.3

Australia is an island; no imports arrive byland and there is no need to allow for geographical contiguity.4 Because Australia is

a reasonably large trading nation—the world’s 14th largest importer in 2006 (WTO

2007, Table 1.9)—there are relatively few empty cells at the 6-digit level of

aggregation. Apart from trade with New Zealand and other Pacific islands, no

significant preferential trading arrangements influence Australia’s trade. Hence,

Australia provides a good natural experiment of the trade costs associated with each

of the 228 trade partners identified in the Australian Bureau of Statistics data.

Several studies show that trade costs vary considerably among country pairs and

are not simply related to distance. Limao and Venables (2001) found a largevariation in the cost of shipping a container from Baltimore to different countries,

some of which is physically determined (landlocked countries have higher transport

costs) but much of it is due to differences in infrastructure. Clark et al. (2004) came

up with similar results for the costs of shipping a container from Latin American

countries to the USA, and emphasised the quality of institutions (corruption,

logistical efficiency, and so forth) as the key determinant of port efficiency. A

similar conclusion informs research on bilateral trade flows; in the micro-founded

gravity model of Anderson and van Wincoop (2003), country-specific trade

resistance terms can be accounted for by exporting-country fixed effects, althoughthe source of the country fixed effects is indeterminate. Recent research has moved

beyond aggregated gravity models to analyse with data disaggregated by

commodity the interaction between variables such as weight/value and timeliness

requirements and the choice of mode of transport and their joint impact with

distance on bilateral trade patterns.5 The present paper complements this work by

using disaggregated data to analyse the variability of trade costs across countries.

The data allow us to decompose, at least partially, country and commodity

characteristics which impact on trade costs. A country’s geographical characteristics

such as distance from major market are immutable and distinct from institutional

and other characteristics which are amenable to policy change. In general, a country

selling bulky goods will have higher transport costs than a country selling high

value/bulk goods.6 Once geographical characteristics and weight have been

controlled for, we can analyse variations in trade costs using measures of

3 Similar data sets for the USA, New Zealand, and some South American countries are described inHummels (2007, 152–153) and in Korinek and Sourdin (2009).4 Hummels (2007), reviewing the literature on trade costs, emphasises the difficulty of measuring costs of land transport (the mode used by over a fifth of international trade) and how they interact with costs of sea

and air transport, which may be substitutes to varying degrees. For Australia the only substitution optionis between sea and air transport.5 Harrigan and Deng (2008), Berthelon and Freund (2008), Egger (2008), Moreira et al. (2008) and

Hummels and Schaur (2009) contribute to this literature and provide references to other work.6 Although bulk accounts for some commodity characteristics, we are unable to take account of other

characteristics such as perishability or fashion which influence the choice of air or sea transport.

Why do trade costs vary? 711

1 3


4/22

institutional quality and other explanatory variables. The determinants of trade costs

are estimated separately for both sea and air freight. However, the choice of

transport mode may be endogenous, e.g. the preference for air is likely to be

increasing with distance and air freight may be a way of avoiding inefficient internal

transport and ports in the exporting country.7

Anderson and Marcouiller (2002) argued that corruption reduces the volume of

trade because insecurity increases trade costs. The higher price of trade affects the

relative attractiveness of producing traded and non-traded goods and services,

causes substitution away from traded goods, and by reducing the gains from trade

has an indirect negative impact on trade through the income effect. Thus, more

corrupt countries trade less, have distorted trade patterns and have lower incomes.

Despite their emphasis on the price of insecurity, the evidence presented by

Anderson and Marcouiller is based on estimating a gravity model, i.e. analysis of

quantities traded. Levchenko (2007) also shows that institutional differences,measured by a composite indicator of protection of property rights and strength of

the rule of law, are a significant determinant of trade flows; countries with good

institutions trade more, and this is more apparent in institution-dependent sectors.8

Our data permit testing for a direct link between corruption and trade costs and we

can analyse the relationship between institutions and trade at a finer aggregation

level than Anderson and Marcouiller (2002) or Levchenko (2007).

A connection between institutions and trade costs will impact on levels of trade

and economic development. Markusen and Venables (2007) relate the degree of

specialization in an economy to the interaction of comparative advantage and tradecosts; high trade costs inhibit a country from taking advantage of potential gains

from specialization and trade in order to promote economic development. In a

global model of the pattern of bilateral trade, Waugh (2009) finds that the calibrated

trade costs are systematically asymmetric, with poor countries facing higher costs to

export their goods relative to rich countries; removing the asymmetry in trade costs,

cross-country income differences decline by up to 34 percent. Importers may be

concerned about time rather than financial costs; Evans and Harrigan (2005), using

proprietary data from a major US department store chain, find that the retailer’s

demand for timely deliveries influenced its choice of source countries, and

Hummels (2001) has estimated that the cost of a day’s delay in transport adds on

average 0.8% to the value of a manufactured good. This is related to a growing

literature on global supply chains, the importance of trade in intermediate goods,

and the costs of having to keep larger inventories if trade is slow or unreliable.

Countries with high trade costs are likely to be excluded from international supply

chains and hence miss out on one of the most dynamic areas of growth in trade and

incomes.

7 The time advantage of air is more pronounced over longer distances. To the extent that transport costs

are related to weight rather than value, they are closer to a specific than an ad valorem charge, and hencetrade costs are declining with respect to unit value; if the charge is by ton-kilometer, then for a givenvalue the preference for air is likely to be increasing with distance. Hummels and Schaur (2009) argue

that, when demand is volatile, air may be preferred because it permits a faster response to price changes.8 A sector’s institutional dependence is measured by complexity, proxied by the Herfindahl index of

intermediate input use.


1 3


5/22

The next section describes the data and provides summary statistics of tradecosts. Section 3 reports baseline results showing that distance and bulk are

significant determinants of trade costs, but that their impact varies across mode of

transport and a substantial part of cross-country variation remains unexplained.

Section 4 analyses the relations between institutions and trade costs, identifying

transport-mode and commodity-specific patterns. The final section draws

conclusions.

2 Data

The Australian Bureau of Statistics data provide annual fob and cif values of

Australia’s imports for 1990–2007 at the HS 6-digit level of aggregation, as well as

reporting weight for about a quarter of the observations and separating out sea, air

and parcel post. After deleting parcel post, re-imports into Australia, country

categories such as ‘‘Unidentified’’, ships supplies and Australian forces overseas,

and the miscellaneous category (HS 99), we had a dataset of 2,097,969

observations, or between 103 and 133 thousand observations per year.9

Overall, average trade costs associated with imports into Australia fellcontinuously and substantially from 8.0% in 1990 to 4.9% in 2007, despite the

huge increase in the price of oil after 1998 (Fig. 1).10 Average trade costs are higher

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

0.090

1990 1995 2000 2005

sea

all imports

air

Fig. 1 Average trade costs,Australian Imports, 1990–2007.

Note: the means are import-weighted (ad valorem tradecosts = Rcif/ Rfob - 1) and

hence biased downwardsbecause goods or tradingpartners with higher trade costswill be underrepresented

9 With more than 5,000 HS 6-digit categories and over 200 trade partners, there are over a millionpotential trade flows, but the data set has just over 100,000 observations per year. Potential biases fromthe truncated sample could be addressed by a two-step sample selection model, but there is unlikely to bea consistent explanation of empty cells, e.g. some reflect high trade costs precluding goods without a

pronounced comparative advantage whereas others are an artefact of size (one tiny Pacific island has fourobservations and over 5,000 empty cells).10 There is a slight increase between 1999 and 2000 and a more substantial increase between 2003 and2004, both of which may be related to oil price increases, but in every other year the average trade cost isconstant or falling from the previous year. The decline in trade costs may be understated due to acomposition effect; if air costs fell faster than sea costs, the lightest or most time-sensitive goods formerlyshipped by sea may now be airfreighted, increasing average transport costs by both modes while

providing more cost-effective transport for all.


1 3


6/22


7/22

specification, a doubling of distance increases ad valorem trade costs by less than a

tenth.15 For the half million observations identified by consistent measures of

weight, the correlation between weight and costs is 0.0013.16

In sum, ad valoremtrade costs are positively related to distance and to weight, but in the Australian data

both of these are weak correlations implying that the variation in ad valorem trade

costs is principally determined by other variables.

Figure 2 illustrates the pattern of ad valorem trade costs over time, using

exporter-commodity fixed effects to control for distance and commodity charac-

teristics. The pattern for goods arriving by sea is similar to that with the raw data in

Fig. 1. The adjusted costs, however, reveal the higher costs of air transport once

weight/value is taken into account. The adjusted values indicate a larger percentage

Table 1 Average trade costs by country 2007

Ad valorem tradecosts

Number of observations

Less than 2 percent 132–3.9 31

4–5.9 57

6–7.9 43

8–9.9 23

10–11.9 17

12–13.9 8

14–15.9 4

16–17.9 3

18–19.9 320.0 percent or more 9

Total 211

Ten largest import sources Ten lowest trade costs Ten highest trade costs

USA 0.050 Puerto Rico 0.010 El Salvador 0.198

China 0.063 Swaziland 0.011 Bhutan 0.205

Japan 0.048 Chad 0.012 Pitcairn Island 0.269

Germany 0.040 Papua New Guinea 0.013 Tonga 0.285

Singapore 0.042 Grenada 0.014 Norfolk Island 0.456UK 0.029 Anguilla 0.015 Guyana 0.492

Malaysia 0.040 Ireland 0.016 Morocco 0.513

New Zealand 0.049 Laos 0.016 Christmas Island 0.547

Korea 0.045 Gibraltar 0.017 Nauru 0.640

France 0.035 St. Helena 0.017 Yemen 0.648

15 Berthelon and Freund (2008) conclude from their disaggregated gravity model analysis that theimportance of distance over time is related to the substitutability of goods, i.e. distance is more relevant to

the cost of trading differentiated manufactured goods than to trade in homogeneous primary products.16 The quantity data include measures by number, square meters and many commodity-specific units. For

556,468 observations they were in metric tons, kilograms, grams or metric carats.


1 3


8/22

decline in maritime trade costs between 1990 and 2007 than shown by the

unadjusted data. The picture for air transport is of a dramatic decline in adjusted

trade costs during the 1990s, but no clear trend since 1999.

A number of other variables have been identified in the literature as influencing

transport costs. Transport costs are subject to scale economies and may depend uponthe potential size of the bilateral trade. Unbalanced trade can influence trade costs, if

the ship or plane has to travel empty in one direction.17 Both scale economies and

unbalanced trade are likely to be more significant for sea than for air freight. Trade

costs may also be influenced by how many shipping lines or airlines serve the

bilateral route and by how much monopoly power they have.18

Trade costs are also influenced by institutional and policy factors. In this paper

the institutions in the importing country, Australia, are constant for all bilateral trade

flows, and differences will be observed dependent upon the exporting country’s

institutions. Limao and Venables (2001) identified onshore infrastructure as animportant variable.19 Clark et al. (2004) focused on port efficiency.20 Port costs may

be high for geographical reasons (e.g. lack of deep water harbours) or low for scale

Fig. 2 Ad valorem trade costs,adjusted for exporter-commodity effects, 1990–2007.

Note: vertical axis indicates advalorem trade costs, using

exporter-commodity fixedeffects to control for distanceand commodity characteristics

17 Wilmsmeier et al. (2006) find that unbalanced trade (measured by the ratio of imports to exports in acountry’s bilateral trade) is a significant determinant of freight costs in Latin America and they argue thattheir estimated coefficients are too low because the imbalances ‘‘need to be applied to broader trade

routes’’ such as South America’s Pacific coast and North America. This is less relevant to Australia,where the only major non-Australian port for an empty ship to pick up cargo in the Southwest Pacific isAuckland. However, a potential complication from using Australia as the yardstick for measuringcountries’ trade costs is the importance of bulk commodities in Australian exports. Although the trade

costs of Australian exports are not the subject of this paper, there may be an indirect non-random impacton Australian import costs from the empty space in returning bulk carriers.18 Hummels et al. (2009) show that one-sixth of importer/exporter pairs are served by a single linerservice, and over half are served by three or less. They also present evidence of shipping companiescharging higher rates on goods with inelastic demand, which is consistent with the exercise of marketpower. In contrast, the measures of market power in Clark et al. (2004) are not statistically significant.Geloso-Grosso (2008) and Piermartini and Rousova (2008) using a gravity model both find a robust

positive relationship between liberalization and the volume of air traffic.19 Their infrastructure index is based on kilometers of road, paved road and railway per square kilometer

and telephone main lines per capita.20

Their principal measure of port efficiency is survey data drawn from the Global CompetitivenessReport published by the World Economic Forum. Wilson et al. (2003) and Wilmsmeier et al. (2006) usethe same source, and Sanchez et al. (2003) use Latin American survey data. Bloningen and Wilson (2008)show that survey data overstate the importance of port efficiency because respondents include othercountry fixed effects. A problem with using the Global Competitiveness Report data or the Bloningen andWilson econometric estimates of port costs is that the former only cover about fifty countries and the

latter cover 100 ports in 42 countries.


1 3


9/22

reasons (e.g. a Rotterdam or Hong Kong effect, which encompasses more than pure

exporting country variables). They may be high because corruption leads to extra

demurrage costs or because political obstacles restrict investment in port facilities.

Devlin and Yee (2005) document the wide variation in logistics costs among the

Middle Eastern and North African countries and how they can influence shippingcosts, e.g. inefficient trucking services lead to longer stand time on the dockside and

costly inventory accumulation as well as reducing export volumes so that there are

infrequent shipping services.21 There is a large literature on the Digital Divide

between developed and developing countries and on the positive effect of Internet

adoption on economic growth.22 We address this complex of determinants by using

the Transparency International Corruption Perceptions Index as a proxy for

‘institutional quality’ in the exporting country.

3 Determinants of trade costs

In our estimating equation, ad valorem trade costs ((cif - fob)/ fob)ik for commodity

k from country i at time t depend on the distance between the country and Australia

(d i,A), the value/weight ratio (VW ik = cif value divided by weight in kilograms),

exporting-country GDP (Y i) or total bilateral trade to capture scale effects, and the

Transparency International Corruption Perceptions Index for the exporting country

(TI i):

cif fobð Þ= fobð Þk it ¼ f d iA; VW k i ; Y it ; TI it

ð1Þ

Table 2 reports OLS regression results using 2007 data and including product fixed

effects.23c

In the full sample of 23,803 observations, distance and the value/weight ratio

have the expected signs and are statistically significant at the one percent level.24

Exporting country GDP and the corruption index both have the expected negative

relation to ad valorem costs. An interaction term between the corruption index and a

dummy variable to indicate if the exporting country is an OECD member is also

21 The World Bank logistics perceptions index provides proxy measures for cross-country variations inlogistic quality (http://info.worldbank.org/etools/tradesurvey/mode1a.asp).22 Freund and Weinhold (2004) found that internet use had no impact on world trade in 1995 but after1997 it had an increasing impact. Andrés et al. (2007), using data from the International Telecommu-nications Union database on the number of internet users, document for 199 countries the wide variationsin internet diffusion and how this is influenced by policy decisions such as the degree of competition

among providers. Unfortunately data on the quality of internet access, intensiveness of use or geographicconcentration are not available for a large enough number of countries to use in cross-country analyses.23 The Transparency International Corruption Perceptions Index is on a scale from 0 to 10, with a highernumber indicating less corruption; 163 countries were covered in 2006 and 180 in 2007. The GDP data

are the current dollars series from the Penn World Tables. Distance (the great circle distance between thelargest city in each country and Sydney) and the landlocked dummy are from the CEPII database referredto in the previous section.24 A dummy for landlocked countries had a negative sign and was statistically significant, which isdifficult to explain as the literature strongly indicates that landlockedness is associated with higher trade

costs (Arvis et al. 2007).


1 3

http://info.worldbank.org/etools/tradesurvey/mode1a.asphttp://info.worldbank.org/etools/tradesurvey/mode1a.asp


10/22

Table 2 Baseline regressions, 2007 data: dependent variable ladvalik : log ((cif - fob)/( fob))i

k

Full sample Sea Air

(a) ladvalik = b0 ? b1ldist iA ? b2lVW ik ? b3 lg dpi ? b4OECDi * TI i ? b5TI i ? ui

Log distance 0.292*** 0.352*** 0.167***

0.138 0.197 0.076

(0.013) (0.013) (0.027)

Log value/weight -0.319*** -0.382*** -0.262***

-0.575 -0.595 -0.465

(0.005) (0.006) (0.008)

Log gdp -0.017*** -0.023*** -0.002

-0.029 -0.047 -0.003

(0.004) (0.004) (0.008)

Sea -1.542***

-0.725

(0.017)

OECD 9 TI 0.005*** 0.020*** -0.017***

0.018 0.077 -0.059

(0.002) (0.002) (0.003)

TI corruption index -0.033*** -0.029*** -0.033***

-0.075 -0.082 -0.066

(0.003) (0.003) (0.006)

Constant -3.088*** -5.053*** -2.281***

(0.103) (0.111) (0.206)

R-squared 0.409 0.367 0.228

N 23,803 15,704 8,099

(b) ladvalik = b0 ? b1ldist iA ? b2lVW ik ? b3limportsi ? b4OECDi * TI i ? b5TI i ? ui

Log distance 0.249*** 0.289*** 0.162***

0.117 0.161 0.076

(0.012) (0.013) (0.014)

Log value/weight -0.317*** -0.381*** -0.108***

-0.573 -0.593 -0.196

(0.005) (0.006) (0.004)

Sea -1.521***

-0.715

(0.017)

OECD 9 TI 0.004** 0.017*** 0.012***

0.013 0.066 0.041

(0.002) (0.002) (0.002)


-0.065 -0.070 -0.023

(0.003) (0.003) (0.003)


1 3


11/22

included to allow the influence of corruption to vary by developing versus

developed economy status, in order to control for the fact that the corruption index

may merely be capturing a development country effect; a positive coefficient on this

interaction term suggests that the effect of corruption is less important as a

determinant of ad valorem transport costs in OECD countries. The mode of

transport, captured by a dummy variable of 1 for sea and 0 for air in the first column

of Table 2, indicates that sea transport is less expensive than air transport once

commodity and country characteristics are controlled for.

Table 2 continued

Full sample Sea Air

Log imports -0.020*** -0.023*** -0.069***

-0.039 -0.054 -0.139

(0.003) (0.003) (0.003)

Constant -2.544*** -4.319*** -2.330***

(0.130) (0.148) (0.159)

R-squared 0.408 0.366 0.058

N 24,010 15,866 24,010

(c) ladvalik = b0 ? b1ldist iA ? b2lVW ik ? b3ltradei ? b4OECDi * TI i ? b5TI i ? ui

Log distance 0.263*** 0.315*** 0.167***

0.124 0.175 0.076

(0.011) (0.012) (0.022)

Log value/weight -0.329*** -0.401*** -0.264***

-0.594 -0.625 -0.468

(0.005) (0.006) (0.007)

Sea -1.458***

-0.685

(0.017)

OECD 9 TI 0.005*** 0.019*** -0.015***

0.018 0.074 -0.050

(0.002) (0.002) (0.003)


-0.067 -0.073 -0.055

(0.002) (0.003) (0.005)

Log trade -0.066*** -0.059*** -0.105***

-0.149 -0.164 -0.185

(0.002) (0.002) (0.006)

Constant -2.394*** -4.337*** -1.317***

(0.105) (0.114) (0.208)

R-squared 0.427 0.390 0.260

N 24,010 15,866 8,144

***, **, * denote significance at the level of 1, 5 and 10% respectively. All regressions include product

effects. Standard errors in parentheses. Standardized coefficients are reported below main estimates


1 3


12/22

To examine whether the determinants of trade costs differ according to the mode of

transport, the last two columns of Table 2 split the sample into goods arriving by sea

and goods arriving by air. Distance and weight have the expected signs with both

modes and, unsurprisingly, the coefficients are larger for imports arriving by sea than

for air freight.25

Exporting country GDP has the expected negative sign for onlyseaborne trade and is significant at the 1% level for sea but not statistically significant

for air, suggesting that scale may be important for seaborne trade. We are aware that

GDP may be picking up other relationships including good institutions, and ran the

same regression replacing GDP by the sum of imports from the trading partner

(Table 2b) and by total imports of commodity k from country i (Table 2c); the results

reported in the three panels of Table 2 indicate that scale, captured by the volume of

trade, is significant for both modes of transport. The institutional quality variable has

the expected negative sign for both air and sea transport, but for imports arriving by air

the coefficient is more economically significant and good institutions in OECDcountries matter more for goods arriving by air than by sea.

In order to compare the relative importance of each independent variable to the

determination of ad valorem transport costs, Table 2 also reports standardized

coefficients. In all specifications, the value to weight ratio is the most important—a

one standard deviation increase in the ratio leads to a fall in ad valorem transport

costs of between 0.6 to 0.5 standard deviations. As expected, value to weight and

distance matter less for ad valorem air transport costs.

Table 3 reports similar regressions with a panel, 1998–2007.26 We experimented

with a number of scale variables; total imports from the trading partner as inTable 2b and total imports of commodity k from country i in year t (log trade) as in

Table 2c. Compared to Table 2, the standard errors are much smaller due to the

larger number of observations. The only qualitative change in the results is the

positive sign on the scale variable when it is measured by total imports by air. In

sum, the results are fairly robust but there is a slight concern about the appropriate

scale variable.

The Transparency International Corruption Perception Index is a proxy for

evaluating the importance of institutions for trade costs. The regressions reported in

Tables 2 and 3 were run with alternative measures such as the Heritage Foundation

Index of Economic Freedom and the World Bank’s Ease of Doing Business index;

the results were identical in sign and statistical significance, with just small

variations in coefficients’ size. To test for variations between high- and low-income

countries (or that a statistically significant coefficient on TI is picking up difference

in technology or sets of traded goods between nations at different levels of

development rather than the effect of institutions), we included a dummy for OECD

members and an interaction term between OECD and TI ; the results were

inconclusive, with differing signs on the coefficient of the interaction term for air

25 The coefficients are significant at the one percent level for both sea and air, with larger coefficients andbigger t -statistics for sea.26 The range was determined by availability of all variables, notably the Transparency InternationalCorruption Perceptions Index. Results with the borders (landlocked) dummy variable are not reportedbecause the coefficient was not positive and significantly different from zero. The regressions underlying

these and the results reported in the following paragraph are available from the authors on request.


1 3


13/22

and sea.27 These results suggest that a more disaggregated analysis may be

appropriate.

4 Trade costs and institutions

This section reports results similar to those reported in Table 3, but disaggregated to

the HS 2-digit level. Disaggregation is likely to be especially important for the

relationship between institutions and trade costs. The aggregate results are

consistent with the hypothesis that in the presence of corruption, exporters prefer

air transport. This could be in order to minimize costs and delays within the

exporting country; goods for which poor institutions may be unimportant, e.g. bulk

commodities, are shipped by sea whereas more time-sensitive, easily pilfered or

Table 3 Baseline regressions, 1998–2007 data: dependent variable ladvalik : log ((cif - fob)/( fob))i

k

Full sample Air Sea

Log distance 0.271*** 0.283*** 0.190*** 0.204*** 0.318*** 0.339***

(0.006) (0.006) (0.011) (0.011) (0.007) (0.007)

Log value/weight -0.227*** -0.240*** -0.166*** -0.173*** -0.269*** -0.296***

(0.005) (0.005) (0.007) (0.007) (0.005) (0.005)

Log imports -0.009*** 0.013*** -0.015***

(0.002) (0.004) (0.002)

Log trade -0.045*** -0.058*** -0.045***

(0.003) (0.006) (0.002)

Sea -1.409*** -1.345***

(0.015) (0.013)

OECD 9 TI 0.007*** 0.010*** -0.010*** -0.005*** 0.020*** 0.023***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

TI -0.034*** -0.033*** -0.042*** -0.034*** -0.031*** -0.030***

(0.001) (0.001) (0.003) (0.003) (0.001) (0.001)

Constant -3.240*** -3.084*** -3.123*** -2.511*** -4.762*** -4.750***

(0.074) (0.061) (0.124) (0.106) (0.080) (0.064)

R-squared 0.359 0.369 0.089 0.102 0.179 0.198

N 245,238 245,238 91,483 91,483 153,755 153,755

***, **, * denote significance at the level of 1, 5 and 10% respectively. Standard errors are in paren-

theses. All models are estimated with product level effects and year dummies but output suppressed. Logimports for air and sea are total imports by mode of transport

27

The OECD dummy should capture differences in average ad valorem trade costs between OECD andnon-OECD countries, but perhaps due to collinearity including both the OECD dummy and theinteraction term renders both coefficients not statistically significant at any standard level of significance.The interaction term included in Tables 2 and 3 captures the differential impact of the TI index betweendeveloped and less-developed economies, and we therefore felt it better to report only the results with theinteraction term. The coefficient on TI and other variables of interest are not materially different whether

the OECD dummy variable is included or omitted.


1 3


14/22

otherwise institution-sensitive goods are sent by air and among the latter set the

lower the perceived corruption in the exporting country the lower the trade costs.

The impact of corruption on trade costs is likely to vary by commodity. The

relative impact on trade costs may be stronger on trade in intermediates, if

corruption creates uncertainty about timeliness of delivery as well as higher directcosts. Especially if a country seeks to trade intermediates as part of an international

value chain, insecurity of supply with respect to time and cost will reduce

competitiveness. Such commodity variation is important, because a strand of the

trade literature suggests that gains from trade liberalization on intermediate goods

are greater than gains from trade liberalization on final goods.28

To capture industry-specific influences on trade costs, we included dummies for

HS 2-digit categories in regressions similar to those reported in Tables 2b and 3. For

goods arriving by sea, these dummies were almost all not significantly different

from zero.29

For goods coming by air, however, the coefficients on the dummieswere mostly statistically significant, suggesting that industry-specific features

(perhaps capturing timeliness, fragility and so forth) influence air transport costs.30

Table 4 reports results for the basic regression run at the industry level (i.e. by

2-digit HS categories) using 1998–2007 data. The estimating equation includes log

distance, log value/weight, log of total bilateral imports at the product (6-digit HS)

level, and the Transparency International Corruption Perceptions Index, as in

Table 2b.31 For goods shipped by sea, distance and bulk are the key determinants of

ad valorem trade costs in almost all categories, with frequent statistically significant

coefficients on the scale variable. The corruption variable has a statisticallysignificant negative sign at the 1% level for only 16 of the 55 categories, and half of

these are primary products or simple processed goods (HS 1–25). In sum, the sea

28 Amiti and Konings (2007) find that in Indonesia the gains from trade liberalization on intermediategoods are greater than gains from trade liberalization on final goods. See also Kasahara and Rodrigue(2008) and references therein.29 The results are available from the authors on request. Only HS 44 (wood and wood products), 63(miscellaneous textiles) and 71 (pearls and precious stones) had coefficients significantly different from

zero at the 1% level; the first two are heterogeneous and the third is not a major sea-freighted category.

Case studies suggest that at some sea ports corruption is a major problem, e.g. in Durban and Maputocorrupt payments account for up to 600% of customs agents’ official income and queue jumping andavoidance of storage costs are important motives for illicit payments (Sequiera and Djankov, 2008). Thissuggests that corruption is most burdensome for traders to whom time matters. Such behaviour may leadto a tragedy of the anti-commons where over-competition for rents leads to less trade, and Australianimports from ports in which corruption is rife may be too small to influence our econometric results.30 Leinbach and Bowen (2004), reporting on a survey of 126 electronics producers in Malaysia, thePhilippines and Singapore, find that ‘‘Much of the variation in air cargo services usage is related toproduct characteristics that go beyond simply the value-to-weight ratio’’ (p. 316) and one of their ‘‘mostsignificant findings… is the extent to which air cargo usage is associated with the degree to which a firm

has internationalized, not only its production sites and final markets, but also its material procurementsites’’ (p. 317). They did not address national variations in institutions, but their case studies highlight the

importance not only of flight times but also of associated services (electronic tracking, specialized plane-to-market logistical support, and so forth), which are unlikely to be compatible with inefficientinstitutions or widespread corruption.31 Categories with few observations (n\50) were omitted. There may be a selection bias due to theweight variable (i.e. goods whose quantity is measured by number, area, volume and so forth are

excluded).


1 3


15/22

T a b l e 4

R e g r e s s

i o n b y H S 2 - d i g i t i n d u s t r y ,

1 9 9 8 –

2 0 0 7 d a t a

H S 2

S e a o n l y

A i r o n l y

L o g d i s t a n

c e

L o g v a l / w g t

L o g i m p o r t s

T I

R - s q u a r e d

N

L o g d i s t a n c e

L o g v a l u e / w g t

L o g i m p o r t s

T I

R - s q

u a r e d

N

2

0 . 7

2 5 * * *

- 0 . 3

1 5 * * *

- 0 . 0

3 5

0 . 0

7 6

0 . 2 0

9

1 5 3

3

0 . 1

5 7 * * *

- 0 . 3

3 5 * * *

- 0 . 0

5 1 * * *

- 0 . 0

1 2 *

0 . 1

8 9

3 , 8

5 2

0 . 3

6 6 * * *

- 0 . 1

5 7 *

- 0 . 0

2 6

- 0 . 0

0 6

0 . 0 7

5

1 , 4

0 5

4

0 . 2

6 4 * * *

- 0 . 3

0 3 * * *

- 0 . 0

6 2 * * *

- 0 . 0

2 5 * * *

0 . 2

5 8

1 , 4

6 9

0 . 2

5 4 *

0 . 0

7 9

0 . 0

6 4

0 . 0

5 3

0 . 0 6

4

6 3 3

5

0 . 3

1 6 * * *

- 0 . 3

2 6 * * *

0 . 0

1 3

0 . 0

3 6 * * *

0 . 2

8 2

5 1 1

0 . 4

6 2 * * *

- 0 . 1

7 0

- 0 . 1

6 1 * * *

- 0 . 0

7 1 * *

0 . 2 0

8

3 4 7

6

0 . 4

2 9 * *

- 0 . 2

6 6

- 0 . 0

3 5

0 . 0

7 4

0 . 4

0 0

5 1

0 . 3

1 5

- 0 . 2

7 6

- 0 . 1

1 3

- 0 . 0

3 0

0 . 3 0

3

1 3 1

7

0 . 2

2 7 * * *

- 0 . 3

0 4 * * *

0 . 0

1 3

- 0 . 0

0 2

0 . 2

0 1

2 , 9

3 8

- 0 . 0

1 7

- 0 . 0

2 1

0 . 0

4 7 * * *

- 0 . 0

0 8

0 . 0 7

2

7 7 5

8

0 . 3

5 3 * * *

- 0 . 3

1 2 * * *

- 0 . 0

3 7 * * *

- 0 . 0

1 7 * *

0 . 2

0 7

2 , 9

1 4

- 0 . 0

7 0

0 . 2

2 0 * * *

0 . 0

8 6 * * *

- 0 . 1

0 5 * * *

0 . 1 6

6

6 7 4

9

0 . 3

8 0 * * *

- 0 . 1

6 9 * * *

- 0 . 0

6 1 * * *

- 0 . 0

4 0 * * *

0 . 1

9 2

5 , 0

9 9

0 . 3

7 4 * * *

- 0 . 0

0 5

0 . 0

6 4

- 0 . 0

1 1

0 . 0 5

1

1 , 5

4 0

1 0

0 . 2

9 7 * * *

- 0 . 1

8 9 * * *

- 0 . 0

0 2

- 0 . 0

0 9

0 . 0

9 2

6 9 4

0 . 5

9 6 * *

0 . 0

3 6

- 0 . 1

0 2

0 . 0

2 4

0 . 1 3

5

1 0 9

1 1

0 . 4

8 1 * * *

- 0 . 2

6 2 * * *

0 . 0

3 5 * *

- 0 . 0

1 3

0 . 2

4 0

2 , 3

5 7

0 . 1

6 4

- 0 . 0

8 4

0 . 1

3 5 * *

- 0 . 0

4 8

0 . 1 2

9

3 5 4

1 2

0 . 3

7 5 * * *

- 0 . 3

1 2 * * *

- 0 . 0

3 7 * * *

- 0 . 0

1 6

0 . 2

2 7

2 , 2

0 7

0 . 4

0 7 * * *

- 0 . 3

2 6 * * *

- 0 . 1

4 4 * * *

- 0 . 0

6 6 * *

0 . 3 8

8

1 , 5

7 6

1 3

0 . 2

1 6 * *

- 0 . 2

2 9 * * *

- 0 . 1

0 6 * * *

- 0 . 0

3 6 * *

0 . 2

0 4

1 , 1

9 7

0 . 3

1 4 * * *

- 0 . 3

5 3 * * *

- 0 . 0

8 1 * *

- 0 . 0

4 6 * * *

0 . 2 1

5

9 2 2

1 4

0 . 1

6 9

- 0 . 2

9 9 * * *

- 0 . 0

3 5

- 0 . 1

0 1 * * *

0 . 2

8 6

4 6 7

0 . 5

0 3 * * *

- 0 . 2

2 4 * *

- 0 . 0

5 8

- 0 . 1

0 1 * *

0 . 1 9

0

9 9

1 5

0 . 2

9 0 * * *

- 0 . 3

4 6 * * *

- 0 . 0

5 8 * * *

0 . 0

1 7

0 . 2

4 3

7 4 0

0 . 3

8 9 *

- 0 . 1

5 9 * *

0 . 0

5 8

- 0 . 0

3 4

0 . 0 8

1

2 9 1

1 6

0 . 0

8 4 * *

- 0 . 2

1 8 * * *

- 0 . 0

6 7 * * *

- 0 . 0

0 7

0 . 1

3 0

3 , 1

2 6

0 . 4

4 5 * * *

- 0 . 2

0 8 * *

0 . 0

3 0

0 . 0

2 3

0 . 1 3

1

6 3 6

1 7

0 . 4

0 3 * * *

- 0 . 3

0 0 * * *

- 0 . 0

0 4

- 0 . 0

1 3 * *

0 . 2

1 6

2 , 0

7 6

0 . 2

8 9 * * *

- 0 . 0

4 4

0 . 0

6 7 * * *

0 . 0

0 3

0 . 0 4

1

7 2 6

1 8

0 . 4

2 9 * * *

- 0 . 2

1 0 * * *

- 0 . 0

4 5 * *

- 0 . 0

4 0 * * *

0 . 2

2 8

1 , 7

9 8

0 . 2

8 2 * * *

0 . 1

8 6 * * *

0 . 0

7 2 * * *

- 0 . 0

2 2

0 . 1 1

7

7 2 1

1 9

0 . 4

5 9 * * *

- 0 . 2

2 7 * * *

0 . 0

2 4 * *

- 0 . 0

2 4 * * *

0 . 2

1 2

3 , 9

0 6

0 . 3

4 2 * * *

0 . 1

7 5 * * *

0 . 0

8 6 * *

- 0 . 0

0 8

0 . 0 9

8

9 9 9

2 0

0 . 3

8 4 * * *

- 0 . 2

7 4 * * *

- 0 . 0

2 7 * * *

- 0 . 0

2 4 * * *

0 . 2

1 1

6 , 0

5 4

0 . 1

3 6

0 . 1

1 8 * *

0 . 0

5 5 * *

- 0 . 0

4 1 *

0 . 0 8

1

9 0 1

2 1

0 . 3

6 9 * * *

- 0 . 2

6 6 * * *

- 0 . 0

5 5 * * *

- 0 . 0

2 5 * *

0 . 2

3 1

2 , 9

1 6

0 . 3

8 2 * * *

- 0 . 1

2 6 * * *

- 0 . 0

5 9 * *

0 . 0

1 4

0 . 0 8

5

1 , 3

5 6

2 3

0 . 2

5 3 *

- 0 . 4

2 0 * * *

- 0 . 0

3 3

- 0 . 0

0 3

0 . 2

9 4

8 6 3

0 . 3

8 8 *

- 0 . 0

4 7

- 0 . 0

1 6

- 0 . 0

3 2

0 . 0 6

2

3 4 0

2 4

0 . 2

2 5 * * *

- 0 . 1

9 6 * *

- 0 . 1

3 8 * * *

- 0 . 0

7 5 * * *

0 . 2

7 1

5 8 3

0 . 3

6 9 * * *

- 0 . 1

9 1 * * *

- 0 . 0

9 1 *

- 0 . 0

5 7 * * *

0 . 1 8

3

4 8 4

2 5

0 . 0

7 5 *

- 0 . 3

5 5 * * *

0 . 0

5 7 * * *

- 0 . 0

1 3

0 . 2

9 1

3 , 5

8 5

0 . 1

6 7 * *

- 0 . 0

1 2

0 . 0

8 7 * *

- 0 . 0

2 2

0 . 1 0

4

1 , 1

8 0


1 3


16/22

T a b l e 4

c o n t i n u e

d

H S 2

S e a o n l y

A i r o n l y

L o g d i s t a n

c e

L o g v a l / w g t

L o g i m p o r t s

T I

R - s q u a r e d

N

L o g d i s t a n c e

L o g v a l u e / w g t

L o g i m p o r t s

T I

R - s q

u a r e d

N

2 6

0 . 2

4 3 *

- 0 . 2

9 5 * * *

- 0 . 0

4 1 *

- 0 . 0

1 7

0 . 2

4 0

5 7 2

0 . 0

8 3

- 0 . 0

6 9

0 . 0

1 9

- 0 . 0

3 4

0 . 1 7

3

3 3 0

2 7

0 . 0

7 3

- 0 . 2

6 1 * * *

0 . 0

0 8

- 0 . 0

1 3

0 . 1

8 8

1 , 2

0 7

0 . 1

1 4

- 0 . 0

6 2 *

0 . 0

8 1 * *

- 0 . 0

3 0

0 . 0 4

6

5 7 7

2 8

0 . 2

9 2 * * *

- 0 . 3

1 4 * * *

- 0 . 0

4 3 * * *

- 0 . 0

1 2 * *

0 . 1

9 6

1 0 , 6

2 3

0 . 0

6 1

- 0 . 1

1 7 * * *

- 0 . 0

5 8 * * *

- 0 . 0

0 2

0 . 0 5

9

5 , 0

9 0

2 9

0 . 3

7 0 * * *

- 0 . 3

3 9 * * *

- 0 . 0

6 8 * * *

- 0 . 0

0 8 * *

0 . 2

5 8

1 8 , 7

8 4

0 . 1

9 4 * * *

- 0 . 2

3 7 * * *

- 0 . 2

0 8 * * *

- 0 . 0

2 7 * * *

0 . 3 0

9

1 3 , 7

7 7

3 1

0 . 2

8 9 * * *

- 0 . 4

4 0 * * *

- 0 . 0

3 2 * *

- 0 . 0

1 1

0 . 3

8 9

1 , 9

3 0

0 . 5

5 5 * * *

- 0 . 0

0 5

0 . 0

0 3

- 0 . 0

6 7

0 . 0 9

2

3 7 3

3 2

0 . 3

5 8 * * *

- 0 . 3

3 3 * * *

- 0 . 0

5 9 * * *

- 0 . 0

0 5

0 . 2

6 2

4 , 9

8 9

0 . 2

8 4 * * *

- 0 . 2

8 8 * * *

- 0 . 0

7 9 * * *

- 0 . 0

2 2 * *

0 . 2 2

4

4 , 0

8 2

3 3

0 . 2

8 9 * * *

- 0 . 2

3 8 * * *

- 0 . 0

6 3 * * *

- 0 . 0

4 8 * * *

0 . 2

2 0

4 , 7

7 2

0 . 3

1 6 * * *

- 0 . 2

4 5 * * *

- 0 . 0

4 6 * * *

- 0 . 0

5 5 * * *

0 . 1 6

3

4 , 0

7 0

3 4

0 . 2

8 0 * * *

- 0 . 2

0 1 * * *

- 0 . 0

0 9

- 0 . 0

4 5 * * *

0 . 1

6 8

4 , 2

9 8

0 . 2

3 6 * * *

- 0 . 1

5 3 * * *

0 . 0

2 6 *

- 0 . 0

5 0 * * *

0 . 1 0

0

2 , 7

0 2

3 5

0 . 2

8 0 * * *

- 0 . 2

5 3 * * *

- 0 . 0

4 5 * * *

- 0 . 0

3 2 * * *

0 . 2

0 9

1 , 6

8 5

0 . 2

1 0 * *

- 0 . 2

2 5 * * *

- 0 . 0

2 0

- 0 . 0

2 5

0 . 1 2

5

1 , 3

5 2

3 8

0 . 2

4 2 * * *

- 0 . 4

2 2 * * *

- 0 . 0

4 3 * * *

- 0 . 0

1 5 * * *

0 . 3

4 9

5 , 4

7 9

0 . 2

4 3 * * *

- 0 . 1

4 5 * * *

- 0 . 0

6 9 * * *

- 0 . 0

3 5 * * *

0 . 1 1

5

3 , 2

2 4

3 9

0 . 2

9 8 * * *

- 0 . 2

9 7 * * *

- 0 . 0

4 0 * * *

- 0 . 0

0 8 *

0 . 1

8 3

1 4 , 0

8 9

0 . 1

8 7 * * *

- 0 . 1

1 0 * * *

0 . 0

2 2 * *

- 0 . 0

5 2 * * *

0 . 0 7

3

9 , 8

2 2

4 0

0 . 2

5 6 * * *

- 0 . 2

3 5 * * *

- 0 . 0

3 6 * * *

- 0 . 0

1 2 *

0 . 1

4 8

2 , 9

8 7

0 . 0

8 3

- 0 . 1

0 8 * * *

0 . 0

0 8

- 0 . 0

8 9 * * *

0 . 1 1

4

1 , 6

5 5

4 1

0 . 0

5 8 * * *

- 0 . 0

9 8 * * *

- 0 . 1

1 4 * * *

- 0 . 0

5 4 * * *

0 . 2

7 2

5 2

0 . 0

3 9

- 0 . 1

2 6

- 0 . 2

3 4 * * *

- 0 . 1

0 6

0 . 3 5

2

6 6

4 4

- 0 . 1

9 6 *

- 0 . 3

1 1 * * *

- 0 . 0

4 0

- 0 . 0

5 1 * *

0 . 2

8 8

2 6 9

1 . 0

1 3 * *

- 0 . 0

4 2

0 . 1

3 2 * * *

- 0 . 2

0 3 * *

0 . 5 1

1

7 4

4 5

0 . 8

4 1 * *

- 0 . 2

2 4 * *

0 . 0

2 4

- 0 . 0

0 2

0 . 1

8 4

5 3

4 8

0 . 2

4 1 * * *

- 0 . 1

8 2 * * *

0 . 0

0 4

- 0 . 0

5 0 * * *

0 . 1

2 9

4 , 1

2 6

0 . 1

5 4 * * *

- 0 . 0

6 7 * * *

0 . 0

4 0 *

- 0 . 0

5 2 * * *

0 . 0 5

0

3 , 7

5 0

5 0

0 . 9

5 7 * *

- 0 . 2

7 2 * *

- 0 . 0

6 4

0 . 0

6 3

0 . 2

9 9

1 3 1

- 0 . 0

3 9

- 0 . 1

4 2 * *

- 0 . 0

9 3 * * *

- 0 . 0

3 2

0 . 1 4

8

2 2 1

5 1

0 . 2

9 1 * * *

- 0 . 3

9 2 * * *

- 0 . 0

8 0 * * *

0 . 0

1 2

0 . 3

0 6

9 0 9

0 . 3

1 1 * * *

- 0 . 2

1 6 * * *

- 0 . 0

8 2 * * *

- 0 . 0

9 5 * * *

0 . 2 5

5

8 4 7

5 2

0 . 4

5 0 * * *

- 0 . 3

0 3 * * *

- 0 . 0

8 3 * * *

- 0 . 0

2 9 * *

0 . 2

0 8

1 , 8

6 9

- 0 . 0

2 1

- 0 . 0

5 6

- 0 . 0

2 0

- 0 . 0

8 4 * * *

0 . 0 6

4

1 , 3

5 0

5 3

0 . 3

3 2

- 0 . 2

7 7 * * *

- 0 . 0

3 3

- 0 . 0

5 2 * *

0 . 1

5 8

3 6 9

- 0 . 3

6 7

- 0 . 0

2 2

- 0 . 0

6 1

- 0 . 0

9 4 * *

0 . 0 7

9

1 5 7

5 4

0 . 5

0 3 * * *

- 0 . 3

8 9 * * *

- 0 . 0

6 2 * * *

- 0 . 0

1 6 * *

0 . 2

7 5

2 , 8

4 9

0 . 2

1 8 * * *

- 0 . 1

7 5 * * *

- 0 . 0

0 7

- 0 . 0

8 3 * * *

0 . 1 1

7

2 , 3

4 3

5 5

0 . 3

7 7 * * *

- 0 . 4

0 8 * * *

- 0 . 0

8 9 * * *

- 0 . 0

1 7 * *

0 . 3

0 5

2 , 6

4 5

0 . 0

9 3

- 0 . 1

5 9 * * *

- 0 . 0

0 5

- 0 . 0

7 6 * * *

0 . 1 1

4

1 , 5

9 6

5 6

0 . 2

4 9 * * *

- 0 . 1

8 0 * * *

- 0 . 0

3 0 * *

- 0 . 0

3 6 * * *

0 . 1

0 8

1 , 8

0 5

0 . 2

2 6 * * *

- 0 . 1

7 4 * * *

- 0 . 0

4 3 * * *

- 0 . 0

7 1 * * *

0 . 1 2

2

1 , 3

9 1


1 3


17/22

T a b l e 4

c o n t i n u e

d

H S �

Pomfret and Sourdin (2010). Why Do Trade Costs Vary

Documents

Transcript of Pomfret and Sourdin (2010). Why Do Trade Costs Vary