Administrative Barriers to Trade∗
Cecília Hornok and Miklós Koren†
September 2010
Abstract
How do exporters react to the administrative burden of trading, i.e. documentation,
customs clearance and technical control? Such administrative costs are variable costs
that typically occur per shipment. Every shipment is to be accompanied by the necessary trade documents and must go through the inspection and customs clearance
processes. Hence, exporters who can sell their products in fewer and larger shipments
bear less of these costs.
We use US shipment-level export data with more than 170 importers to identify the
margins of adjustment at which trading rms respond to administrative costs. Data
on administrative trade barriers are the number and list of trade documents and the
time and nancial cost of administrative tasks from the World Bank's Doing Business
database. We decompose total trade by country into ve margins: the number of
shipments, the price and the physical size of a typical shipment, the transport mode,
and the product composition. Preliminary results show that exporters respond mostly
with the number and size of shipments and less by changing the transport mode or the
product composition.
Preliminary. Please do not quote.
Koren : Central European University, IEHAS and CEPR, E-mail: KorenM@ceu.hu. Hornok : Central
European University and Johannes Kepler University Linz, E-mail: cphhoc01@ceu-budapest.edu. Koren is
grateful for nancial support from the European Commission under the FP7 program "EFIGE". Hornok
thanks for nancial support from the Marie Curie Initial Training Network "GIST".
∗
†
1
1 Introduction
With the diminishing use of tari-type trade restrictions, the focus of trade policy makers
has been increasingly shifted towards less standard sorts of trade barriers, including administrative barriers to trade. As opposed to taris, which can be sources of government revenues,
an important characteristic of administrative barriers is that they are a deadweight loss in
a welfare sense.
We dene administrative barriers to trade in a restricted way. We do not mean the
presence of administrative regulations as product standards, technical or health regulation
per se. Rather, we dene administrative barriers as bureaucratic procedures/administrative
tasks ("red-tape barriers") that a trading rm has to get through when shipping the product
from one country to the other. This may involve, for example, the burden of preparing health
certicates, but not that of making the product itself to comply with the health requirements.
We argue that administrative barriers to trade, as dened above, are typically trade costs
of a 'per shipment' nature. They are not an iceberg type, for they are not proportional to
the value of the product. Nor are they per unit costs. The tasks of trade documentation,
cargo inspection, or customs clearance occur per shipment, and shipments may contain
varying amounts of product units. Exporters who can sell their products in fewer and larger
shipments bear less of these administrative costs.
While keeping total export sales unchanged administrative costs can be lowered by reducing the number of shipments, which in turn may require an increase in the size of a
shipment, a change in the transport technology, or a move towards typically large-shipment
(bulky) products in the product composition. At the extreme, the administrative burden
may entirely prevent the sale of some goods to some destinations.
Bunching goods into fewer and larger shipments, involves tradeos, however. An exporter
waiting to ll a container before sending it o sacrices timely delivery of goods and risks
losing orders to other, more exible (e.g., local) suppliers. We believe the welfare costs of
such inexibility may be large. In future work, we wish to calibrate a structural model that
illustrates these costs.
Our emphasis on shipments as a fundamental unit of trade ows follows Armenter and
Koren (2010), who discuss the implications of the relatively low number of shipments on
empirical models of the extensive margin of trade. They do not focus on the determinants
of transport mode or shipment size.
The dominance of iceberg trade costs in trade theory was challenged by Hummels and
Skiba (2004), who argue that trade costs are at least partly per unit costs. This has important
implications for trade theory: in contrast with ad valorem costs, per unit trade costs do
not necessarily leave relative prices and relative demand unaltered. Demonstrating their
2
hypothesis on a rich panel data set they get an interesting side result. The per (quantity)
unit freight cost depends negatively on total traded quantity. Hence, the larger the size of a
shipment in terms of product units, the less the freight cost per unit is, which in fact suggests
the presence of per shipment trade costs.
Since bureaucratic trade barriers rather represent time barriers than nancial costs, our
approach is strongly related to the literature on the cost of time in trade. An important
message of this literature is that time in trade is far more valuable than what the rate of
depreciation of products (either in a physical or a technical sense) or the interest cost of
delay would suggest. Hummels (2001) demonstrates that rms are willing to pay a disproportionately large premium for air (instead of ocean) transportation to get fast delivery. The
estimated premium (0.5% of the product value per day) far exceeds the daily interest rate.
A series of papers (Harrigan and Venables (2006), Evans and Harrigan (2005), Harrigan
(2010)) look at the implications of the demand for timeliness on production location and
transport mode choice. Harrigan and Venables (2006) point out that the high cost of time
in trade is ultimately related to uncertainty. Uncertainty about the changes in consumer
tastes forces producers to design, produce and deliver products to the nal market within the
shortest possible time. The transit of product components between production stages, which
are often geographically fragmented, should also be synchronized in a timely manner, since
the delay of one single component can hold up the whole production process. As a result
industries tend to agglomerate. In Evans and Harrigan (2005) the demand for timeliness
inuences the decision on whether to import products from places close to the nal market
at the expense of higher wages, or source from more distant importers with lower labor costs.
Harrigan (2010) builds an Eaton-Kortum type model of comparative advantage with demand
for timeliness, in which he also incorporates transport mode choice. Timely delivery can be
assured either by sourcing from nearby producers, or by shipping the product via - the more
expensive and faster - air transportation. The model predicts that faraway suppliers will have
comparative advantage in small weight-to-value goods, which are more easily transported by
air.
Beside the above two, location choice and transport mode choice, trade-os, our paper
deals with a third trade-o. The choice is on the frequency/size of a shipment and the
trade-o is timeliness versus smaller per shipment administrative costs. The demand for
timeliness requires relatively small and frequent shipments, while the burden of per shipment
administrative costs can be mitigated by reducing the frequency, and increasing the size, of
shipments.
A paper closest in spirit to ours is Alessandria, Kaboski and Midrigan (2008), who also
build on the timeliness literature and on the evidence of the lumpiness of shipments. They
argue that lumpy trade is due to xed transaction costs, which force trading rms to ship
3
products infrequently and maintain large inventory holdings. They dene "xed transaction
costs" as the time lag between the order and the delivery of goods and other xed costs
of transacting. When these costs are large, rms send shipments less frequently and hold
larger inventories of imported goods. On rm level data they document that importers hold
signicantly larger inventory holdings than non-importers and, hence, import shipments are
larger and less frequent than domestic ones. Moreover, the same regularity holds within the
rm for international versus domestic purchases.
Here we look at adjustments on the shipment size instead of inventory holdings. Our
paper therefore provides complementary evidence to Alessandria, Kaboski and Midrigan
(2008).
More policy-oriented papers give estimates on the eects of time-related and administrative barriers on trading. Using the Doing Business database Djankov, Freund and Pham
(2006) incorporate the number of days spent with documentation, customs, port handling
and inland transit into an augmented gravity equation and nd that each additional day
delay before the product is shipped reduces trade by more than 1%. Part of the policy
literature is centered around the notion of 'trade facilitation', i.e. the simplication and
harmonization of international trade procedures.1 This line of literature provides ample evidence through country case studies, gravity estimations and CGE model simulations on the
trade-creating eect of reduced administrative burden. An assessment of estimates shows
that trade facilitation can decrease trade costs by at least 2% of the trade value, and this
number may get as large as 5-10% for less developed countries.
2 The eects of administrative barriers
In order to test the hypothesis on the adjustment of shipment size and frequency to administrative barriers one needs data on international transactions at the shipment level.
Standard trade data sets do not provide such information. We use data on US exports to
170 countries in year 2005 from the US Census Bureau. This data set tells the number of
shipments2 and the value (in US dollars) and quantity of trade within each 10-digit product
category and partner country cell. Based on this information one can calculate the value
(physical) shipment size for each product and importer by dividing trade value (quantity)
with the number of shipments. Though the US Census trade database is not a shipment
1 For
more see e.g. Engman (2005) or Francois, van Meijl and van Tongeren (2005).
US Census Bureau denes a shipment accordingly: "Unless as otherwise provided, all goods being
sent from one USPPI to one consignee to a single country of destination on a single conveyance and on the
same day and the value of the goods is over $2,500 per schedule B or when a license is required.", where
USPPI is a U.S. Principal Party in Interest, i.e. "The person or legal entity in the United States that receives
the primary benet, monetary or otherwise, from the export transaction."
2 The
4
level database, the distribution of the calculated shipment sizes should be suciently close
to the true distribution, since the number of shipments within each product-country cell is
typically very small, and almost one-third of the cells contain only one shipment.
The database is also suitable for analysing the transport mode choice. It includes information on the share of trade value transported by air or vessel. If these two do not cover
total trade, we assume ground transportation. We create a database with observations that
are uniquely assigned to one of the three transport modes (air, vessel, ground) by dropping
those observations, where more than one transport mode is used (16% of total trade value).
This way the database becomes three-dimensional along the product code, importer and
transport mode dimensions.
Regardless the rich nature of the trade database, the eect of administrative barriers on
shipments can be measured only on an aggregate country level, since data on the administrative burden is only available for countries. We aggregate US export data to a cross section
of importers in a way that retains most of the richness of shipment information from the
original database, i.e. the information on the frequency versus the size of shipments, the
value versus the physical shipment size, the transport mode choice, and the shipment size
heterogeneity of products.
We decompose US exports to each of the 170 importers into ve margins. These are
the shipment extensive margin, i.e. the number/freqency of shipments, and four margins of
the shipment size: the within-product-transport price and physical shipment size margins,
the transport mode margin, and the margin of product composition. The ve margins
separate ve possible ways of shipment-level adjustment to trade barriers. In response to
trade barriers rms may reduce the number of shipments, they may change the price, pack
larger quantities of goods in one shipment, switch to a transport mode that allows larger
shipments, or change the composition of exported products towards more bulky products.
2.1 Decomposing dierences in average shipment size
Let i index products (10-digit product codes), m modes of shipment (air, vessel, ground),
and c partner countries. Let country 0 be the benchmark country (the average), for which
the share of product-level zeros are the lowest. In fact, we want all products to have nonzero
share, so that the share of dierent modes of transport are well dened for the benchmark
country. (Note that the mode of transport will not be well dened for a product/country
pair if there are no such shipments. This will not be a problem because this term will carry
a zero weight in the index numbers below.)
Let ncim denote the number of shipments of good i through mode m going to country c.
Similarly, qcim denotes the average shipment size for this trade ow in quantity units, pcim
5
is the price per quantity unit (unit value). We introduce the notation
ncim
scim = P
k ncik
for the mode composition of good i in country c, and
P
k ncik
sci = P P
j
k ncjk
for the product composition of country c. We dene s0im and s0i similarly for the benchmark
(average) country.
We decompose the ratio of total trade value (X ) to country c and the benchmark country,
P
P
P P
n
p
q
n
s
s p q
Xc
cim
cim
cim
c
ci
P i Pm cim cim cim ,
= Pi Pm
=
X0
n0 i s0i m s0im p0im q0im
i
m n0im p0im q0im
as follows,
P
P
P
P
sci m scim p0im qcim
sci m scim pcim qcim
Xc
nc
i
i
P
P
·P
·
=
·P
sci m scim p0im q0im
X0
n0
i sci
m scim p0im qcim
iP
P
P
P
sci m scim p0im q0im
sci m s0im p0im q0im
i
i
P
P
P
·P
.
m s0im p0im q0im
m s0im p0im q0im
i sci
i s0i
Or let us express the same decomposition identity in short as
Xc
≡ Xc,total = Xc,extensive · Xc,price · Xc,within · Xc,transport · Xc,prodcomp ,
X0
(1)
where, on the right-hand side, the rst, second, third, etc. terms correspond to the rst,
second, third, etc. terms in the previous equation.
The rst term is the extensive margin: how does the number of shipments dier in the
two countries? The second term is a price eect: how much more expensive are goods
shipped to country c, relative to the benchmark country? The third term we call the within
eect: how do physical shipment sizes dier in the two countries for the same product and
mode of shipment ? The fourth term is a mode of transportation eect: how does the mode
of transportation dier for the same product? Some transport modes allow for larger sized
shipments, as it is reected in Table 2. Are those modes overrepresented? The last term is
a product composition eect: how does the production composition dier? Are bulky items
and/or items that typical travel by large shipment modes overrepresented?
If per shipment administrative trade barriers make trading rms send less frequent and
larger shipments one should see the extensive margin to respond negatively and the within
margin positively to larger per shipment costs. Firms may also choose to switch to a transport
mode that allows for larger shipments, which should be seen from a positive response of the
transport margin. Or they can alter the composition of the traded product mix, so that the
share of products that are typically transported in larger shipments increases. In this case,
larger per shipment costs are associated with a larger product composition eect.
6
2.2 Indicators of administrative barriers
We use the import side of the 'Trading across borders' section of the Doing Business (DB)
database of the World Bank. The data is based on a survey among trade facilitators at
freight-forwarding companies. It includes country-level data on the time and the nancial
cost of importing a standardized cargo of goods.3 We use data from 2009, because the
procedural breakdown is available only for the most recent survey. Though other data is
from year 2005, we do not see it problematic, given the very strong persistence of the DB
indicators across years.
In the DB survey, the traded product is assumed to travel in a dry-cargo, 20-foot, full
container load via ocean. It weighs 10 tons, is valued at USD 20,000, it cannot be hazardous
and does not require special treatment or standards.4 All trade procedures, except for the
process of ocean transportation, are taken into account. In the case of importing, it means
all procedures from the vessel's arrival at the port of entry to the cargo's delivery at the
warehouse in the importer's city (the largest city by assumption).
The import transaction is broken down into four procedures:
• document preparation,
• customs clearance and inspection,
• port and terminal handling, and
• inland transportation and handling.5
Both the time and the nancial cost are reported for each procedural stage separately.
Time is expressed in calendar days, and it spans the period from the moment a procedure is
initiated until it is completed. The nancial cost variables are in US dollars per container.
They only include the cost of the procedure without taris, trade taxes or bribes.
In our view, bureaucratic barriers to trade are more closely captured by the rst two
procedures, i.e. document preparation and customs clearance and inspection. The latter
two, and especially inland transport, which is dominated by the freight cost of transit, are
more closely related to moving and storing the goods. Hence, in the empirical analysis we
are primarily interested in the eect of the rst two procedures, and leave the second two
for controlling.
3 In
addition, there is data on the number of documents required for customs and other agencies, complemented with a list of document names.
4 http://www.doingbusiness.org/MethodologySurveys/TradingAcrossBorders.aspx
5 Note that port and terminal handling refers to seaport handling, and inland transportation is from the
seaport, since the survey question is specic to ocean transportation.
7
Summary statistics show large variations of indicators across the 170 importers (Table
3). It seems that time and nancial costs to import tend to be larger for less developed
countries; the median is 32 days and USD 1,813 for the African continent, in contrast with
the 12 days and USD 1,160 for Europe. In terms of time, the best performing country is
Singapore, where an import transaction is completed within 3 days, in sharp contrast with
Chad or Uzbekistan, where the required time is 3 months.
Document preparation is the most time-consuming out of the four procedures: it represents half of the total time for the average country (Table 4). In terms of nancial costs,
however, inland transportation is the most burdensome, taking up almost half of the total
cost for the average country. This suggests that the administrative burden of trading is
better represented by the amount of time lost than by a nancial measure.
The relatively weak relationship between the time and nancial burden of administrative
tasks is also evident from correlation coecients between the day and the cost measures of
the same procedure (diagonal in the lower-left quadrat of Table 5). They are quite modest at
0.3-0.4, except for inland transit, where it is 0.8. Correlation coecients between procedures
for either the day or the cost measure are not strong either, the highest being 0.66 for
Document Day and Custom Day.
2.3 Empirical specication
We build on the gravity theory of Anderson and van Wincoop (2003), who derive a gravity
equation for bilateral trade between exporter r and importer c (in logs) as follows,
ln Xrc = ln Yr + ln Yc − ln YW + (1 − σ) ln Trc − (1 − σ) ln Πr − (1 − σ) ln Pc ,
(2)
subject to the constraints on the relationship between Πr and Pc ,
1
# 1−σ
"
Πr =
X
1−σ σ−1
θc Trc
Pc
c
and
1
# 1−σ
"
Pc =
X
1−σ σ−1
θr Trc
Πr
,
r
where Yr and Yc are exporter and importer income levels, YW is world income, Trc are
bilateral trade costs, σ is the elasticity of substitution between domestic and foreign goods,
and θr = YYWr and θc = YYWc are income shares. Πr and Pc are the multilateral trade resistances
(MR) of the exporter and importer countries, which are average measures of trade barriers
that the exporter faces in, and the importer imposes towards, all countries in the world.
8
The need to account for the two MR terms has been recently the main challenge in the
empirical applications of the gravity equation. The most commonly used solution, when
country xed eects (or country-time xed eects in panel data sets) control for the MRs, is
not applicable for two reasons: we have only one country dimension and we want to identify
the eect of a country-specic (and not bilateral) trade cost.
We control for the MRs according to Baier and Bergstrand (2009) (henceforth, BB).
They show that, under the assumption of bilaterally symmetric trade costs6 (Trc = Tcr ), the
MR terms can be expressed by their Taylor series approximations (centered around a world
with symmetric trade frictions, Trc = T ) as
ln Πr =
N
X
l=1
and
N
N
N
N
1 XX
θk θl ln Tkl
θl ln Trl −
2 k=1 l=1
N
X
1 XX
θk θl ln Tkl .
ln Pc =
θk ln Tkc −
2 k=1 l=1
k=1
Substituting these two MR terms into the gravity equation (2) we get
ln Xrc = ln Yr + ln Yc − ln YW + (1 − σ) ln Trc −
" N
#
N
N X
N
X
X
X
− (1 − σ)
θl ln Trl +
θk ln Tkc −
θk θl ln Tkl . (3)
l=1
k=1
k=1 l=1
Note that the last (cross-product) term in the bracket is always constant across country
pairs. Moreover, with only one exporter (US) all the r-specic terms are constants. Hence,
on a cross-section of importers the gravity equation can be simplied as
ln XU S,c = α + ln Yc + (1 − σ) ln TU S,c −
N
X
!
θk ln Tkc ,
k=1
where all the constants are now in α, TU S,c is bilateral trade cost between the US and
P
importer c, and the last term M Rc = Nk=1 θk ln Tkc is the multilateral trade resistance of
the importer, i.e. how restrictive the importer's market is towards all the exporter countries
in the world.
The expression for the multilateral resistance can be simplied further in the case of
importer-specic (i.e. not bilateral) trade costs, such as the Doing Business indicators. In
the case of importer-specic trade costs, Tkc = Tc for all k 6= c and Tkc = 0 for k = c, i.e. the
trade cost aects all the exporters the same way, except for domestic trade. Regardless of the
source country, the import customs procedure and documentation are equally burdensome
6A
similar, though less simple, solution can be derived for the case of asymmetric trade costs.
9
for all shipments, while domestic trade is not subject to such barriers. The expression
P
ln Tc − N
k=1 θk ln Tkc simplies to ln Tc − (1 − θc ) ln Tc = θc ln Tc and the gravity equation
with only country-specic trade costs becomes
ln XU S,c = α + ln Yc + (1 − σ)θc ln Tc .
(4)
Estimating this equation gives consistent estimates of the gravity parameters. We are
however primarily interested in the comparative static eects and, as Anderson and van
Wincoop (2003) shows, the comparative static eects also depend on the country sizes. As
it was rst pointed out in Behar (2009), the dierence between the gravity parameter and
the comparative static eect typically gets very large in the case of country-specic trade
costs. Dierentiating equation (3) with respect to the country-specic trade cost, we get the
expression for the comparative static eect,
∂lnXU S,c
= (1 − σ) θc (1 − θc ) ≈ (1 − σ) θc ,
∂lnTc
i.e. the comparative static eect is the gravity parameter discounted by the country's
income share. Since the importers' income shares in our sample are low (the average is 0.004
and the largest is Japan with 0.1), the above dierence is always large.7
A further empirical problem in applying the Baier and Bergstrand method to countryspecic trade costs appears, when one wants to include more than one country-specic
variables in the regression. In this case, the dierent θc ln Tc variables can become strongly
correlated due to their common θc element and create a multicollinearity problem in the
estimation. A way to counteract the multicollinearity problem and to directly estimate the
comparative static eect is the following. Note that θc ln Tc = θ̄ ln Tc + (θc − θ̄) ln Tc , where θ̄
is the average-sized importer's income share. If we estimate equation (4) by both including
ln Tc and (θc − θ̄) ln Tc in the regression equation, the coecient on the rst term gives
consistent estimate of the comparative static eect for the average-sized importer ((1 − σ)θ̄),
while the second term controls for the remaining part of the multilateral resistance.
Against this background, we estimate the following cross-section gravity equation, where
we assume that the trade cost function is log-linear both in its bilateral and country-specic
elements,
7 The
dierence can also get non-negligible in the case of bilateral trade costs, if either of the countries is relatively large. Taking the US and the average importer's income shares in our sample (0.28 and
0.004, respectively), the comparative static eect of a bilateral trade cost is a factor of 0.7 smaller than
U S,c
the gravity parameter. Formally, the comparative static eect for the bilateral trade cost is ∂lnX
∂lnTU S,c =
(1 − σ) (1 − θc − θU S + θU S θc ).
10
ln XU S,c,z = β0,z + β1,z ln Tc + β2,z [(θc − θ̄) ln Tc ]
+ β3,z ln GDPc + β4,z ln GDP P Cc + β5,z [ln TU S,c − M Rc ] + U S,c,z , (5)
where z ∈ [total, extensive, price, within, transport, prodcomp ]. Hence, the estimation is
done separately for each component of the decomposition identity (equation (1)) in logs on
the left-hand side. We do not assume the gravity equation to provide a good description of
the variation on the margins. We only want to know the contribution of each margin to the
eect on total trade. By construction, the coecients on the ve margins should sum up to
the coecients in the total trade regression.
On the RHS we include the importer-specic trade cost variables (ln Tc ), keeping in
mind that the estimated coecient βˆ1,z gives the comparative static eect for the averagesized importer. In addition, we include the expression that controls for the rest of the
multilateral resistance: (θc − θ̄) ln Tc . The set of importer-specic trade costs includes rst of
all the variable(s) for per shipment administrative trade barriers to import from the Doing
Business database. These may be the time in days and the log of the cost in US dollars
of administrative procedures in the importer country. In addition, we include other nonbilateral trade costs, such as a landlocked dummy and the importers' average tari rate.
Additional controls are the log of the nominal GDP and GDP per capita of the importer.
We include GDP per capita primarily as a proxy for the overall institutional quality of the
importer. This way we can ensure that the administrative burden variable does not pick up
eects from other elements of institutional quality, with which it may be highly correlated.
Table 5 documents the signicant negative correlation between GDP per capita and the
administrative barrier indicators.
ln TU S,c includes the bilateral trade cost variables: log distance from the US, dummies
for common language, colonial ties, Preferential Trade Agreement and Free Trade AgreeP
ment with the US. The M Rc = Nk=1 θk ln Tkc term is the BB-type multilateral resistance
counterparts of the bilateral trade cost variables. They are calculated by taking the sample
countries (170 importers + USA) as the world total.
We estimate equation (5) with simple OLS and robust standard errors in the case of total
trade. In the case of the ve margins, however, we exploit the correlatedness of the errors
across the ve equations and apply Seemingly Unrelated Regressions (SUR) estimation.
2.4 Results
We investigate the eect of the time (in days) and the nancial cost of the trade procedures
separately. Table 6 shows the estimated eects of the total days indicator, Tables 7 presents
the total cost indicator (in log), where total means the sum of the four procedures.
11
The main message is that rms send larger-sized and less frequent shipments to countries,
where the administrative burden of trading is larger. This is evident from the signicantly
negative (positive) coecient on the total days or total cost indicators in the extensive
(within) margin regressions. In contrast, the eect on total trade, which equals the sum of
the eects on the ve margins, is more muted. For total days the coecient is -0.01, i.e.
total trade decreases by 1%, if bureaucratic tasks take 1 day longer. Notice that this result
is at the lower end of the result of Djankov, Freund and Pham (2006).
Our numerical results for total days can be read as follows. One day increase in the time
to import causes an approximately 1.6% decline in trade due to the fact that rms send
less shipments (extensive margin). This eect is however partly compensated on the within
margin: rms pack more products in one shipment, which is equivalent to a 0.5% increase
in trade. In the case of the nancial cost of trading (Table 6) the results read as follows.
If the total cost to import is twice as large for an importer than for the average importer,
exports to that country are around 35% smaller (100 · ln 2·coecient), which is the result of a
more than 50% decline due to fewer shipments (extensive margin) and 9%-9% increases due
to larger shipment size (within margin) and the change in the product composition towards
larger-shipment goods (product composition margin).
We run regressions also for the day and cost indicators of the four procedures separately.
Results for the day indicators are in Tables 8 and 9, results for the cost indicators are in
Tables 10 and 11, where we only report the coecient estimates on indicators in question.
The rst tables shows the results from gravity equations, where the indicator of only one
procedure is included. The second tables report results from a gravity equation with all the
four indicators.
The signicant eect on the within margin that we observe for total days and total cost is
mainly coming from the two procedures that are more strongly associated with administrative
tasks (documentation and customs). When the other two procedures (port handling and
inland transit) are considered, rms seem to decrease trade by reducing the number of
shipments without signicant adjustment on the shipment size. These latter two procedures
may not constitute pure per shipment barriers, so that increasing the shipment size is not
necessarily the optimal response of rms.8
If procedures are considered separately we also nd evidence for adjustment on the transport margin. Facing longer customs time (Customs Day variable), exporting rms tend to
choose larger-shipment transport modes such as ocean or ground. In contrast, if the time
to reach the nearest port (Transit Day) is large, rms switch to lower-shipment transport
modes. An explanation for this reverse eect is that port in the Doing Business survey
8 Port
and terminal handling and inland transit time and cost can also increase with the size/weight of
the shipment, as they involve the physical movement of the cargo.
12
means ocean port, so in this case switching from the large-shipment ocean transport mode
is rational. The same eect is present for the nancial cost of transit indicator. Finally, we
nd no convincing evidence that administrative burden would signicantly alter prices.
3 Conclusion
The paper estimates the eect of administrative trade barriers on various margins of trade
ows. We nd that exporters respond mostly with the number and size of shipments and only
less by changing the transport mode or their product composition. In markets with larger
administrative burden imports come in fewer and larger shipments, and larger-shipment
transport modes are more often used.
References
[1] Alessandria, G., Kaboski, J. and Midrigan, V. (2008) "Inventories, Lumpy Trade, and
Large Devaluations," NBER Working Paper No. 13790.
[2] Anderson, J. E. and van Wincoop, E. (2003) "Gravity with Gravitas: A Solution to the
Border Puzzle," American Economic Review, 93, pp. 170-192.
[3] Armenter, R. and Koren, M. (2010) "The Balls-and-Bins Model of Trade," CEPR Discussion Paper No. DP7783.
[4] Baier, S. L. and Bergstrand, J. H. (2009) "Bonus vetus OLS: A Simple Method for
Approximating International Trade-Cost Eects Using the Gravity Equation," Journal
of International Economics, 77, pp. 77-85.
[5] Behar, A. (2009) "De Bonus Vetus OLS: Approximating the international tradecost eects of red tape" The Selected Works of Alberto Behar. Available at:
http://works.bepress.com/alberto behar/11.
[6] Francois J., van Meijl, H. and van Tongeren, F. (2005) "Trade Liberalization in the Doha
Development Round," Economic Policy, April 2005, pp. 349-391.
[7] Djankov, S., Freund, C. and Pham, C. S. (2006) "Trading on Time," World Bank Policy
Research Working Paper No. 3909.
[8] Engman, M. (2005) "The Economic Impact of Trade Facilitation," OECD Trade Policy
Working Papers No. 21.
13
[9] Evans, C. and Harrigan, J. (2005) "Distance, Time an Specialization: Lean Retailing in
General Equilibrium," American Economic Review, 95(1), pp. 292-313.
[10] Harrigan, J. (2010) "Airplanes and Comparative Advantage," NBER Working Paper
No. 11688.
[11] Harrigan, J. and Venables, A. J. (2006) "Timeliness and Agglomeration,"
Urban Economics, 59, pp. 300-316.
[12] Head,
K.
(2003)
"Gravity
for
http://strategy.sauder.ubc.ca/head//gravity.pdf.
Beginners,"
Journal of
Available
at:
[13] Hummels, D. (2001) "Time as a Trade Barrier," GTAP Working Paper No. 18.
[14] Hummels, D. and Skiba A. (2004) "Shipping the Good Apples out? An Empirical
Conrmation of the Alchian-Allen Conjecture," Journal of Political Economy, 112(6),
pp. 1384-1402.
Appendix
A Data reference
Foreign trade data
Trade data is exports of the United States to 170 countries in year 2005 from the foreign trade
database of the US Census Bureau. Information is available on the number of shipments,
the value (in US dollars) and quantity of trade, and the transport mode by importer and
10-digit product category.
As for the transport mode there is data only on the trade value shipped by air or vessel. If
these two modes do not cover total trade, we assume ground transportation for the remaining
trade ow. We drop those observations, where nonzero ows are associated with more than
one transport mode, i.e. the transport mode is unknown (16% of total trade value and 21%
of total number of shipments). Hence, one of the three transport modes (air, vessel, ground)
is uniquely assigned to each observation.
A particular feature of this data set is the information on the number of shipments.
Though it is not shipment-level data, the number of shipments in an importer-product cell
is typically very small. There is only one shipment in almost 30% of the observations and
the median shipment number is four. The size of a shipment for each importer-product cell
can be measured either as the value or the quantity of trade over the number of shipments
14
in that cell. The former is referred to as the value shipment size, the latter as the physical
shipment size.
We drop product lines, which correspond to low value shipments (2% of the total trade
value). In the Census database trade transactions are reported only above a trade value
threshold. Low value shipment lines are estimates based on historical ratios of low value
trade, except for Canada, where true data is available. They are classied under three
product codes as aggregates. Hence, they appear erroneously as three large shipments and
distort the shipment size distribution.9
Price is calculated as a unit value, i.e. value over quantity. It is an f.o.b. price, since
exports are valued at the port of export in the US and include only inland freight charges. In
the database there is no single quantity measure, which would apply to all product categories:
product quantities are measured either in kilograms, numbers, square meters, liters, dozens,
barrels, etc. It is therefore important to calculate the price at the 10-digit product level,
where the quantity measure per product is unique. For some products the quantity measure
is not dened; here we assume that quantity equals value, i.e. the quantity measure is a unit
of US dollar.
Other regressors
GDP and GDP per capita of the importer countries in current USD for year 2005 is from
the World Bank's World Development Indicators database.
Gravity variables (bilateral distance, landlocked, common language, colonial ties) are
from CEPII. Distance is the population-weighted average of bilateral distances between the
largest cities in the two countries, common language dummy refers to ocial language,
colonial ties dummy refers to colonial relationship after 1945.10
Tari rate in the importing country is the un-weighted average of ad valorem tari rates
for all goods in 2005 from the World Bank's "Trends in average MFN applied tari rates in
developing and industrialized countries" database.11 When data for 2005 is missing, either
simple linear intrapolation between the two nearest data points is used or the closest year's
value (usually 2003 or 2007) is taken.
PTA and FTA dummies indicate preferential or free trade agreements, respectively, effective in year 2005. They are based on the Database on Economic Integration Agreements
provided by Jerey Bergstrand on his home page.12 We dene PTA as categories 1 and 2
9 Low
value shipment lines are 9809005000: "Shipments valued USD 20,000 and under, not identied
by kind", 9880002000: "Canadian low value shipments and shipments not identied by kind", 9880004000:
"Low value estimate, excluding Canada".
10 Description of variables by CEPII: http://www.cepii.fr/distance/noticedist_en.pdf
11 siteresources.worldbank.org/INTRES/Resources/tar2008.xls
12 http://www.nd.edu/~jbergstr/#Links
15
(non-reciprocal and reciprocal PTA's) and FTA as categories 3 to 6 (free trade agreement,
customs union, common market, economic union) in the original database.
B Figures and Tables
Table 1: List of importer countries
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Afghanistan
Albania
Algeria
Angola
Antigua and Barbuda
Argentina
Armenia
Australia
Austria
Azerbaijan
Bahamas
Bahrain
Bangladesh
Belarus
Belgium
Belize
Benin
Bhutan
Bolivia
Bosnia-Herzegovina
Botswana
Brazil
Brunei
Bulgaria
Burkina
Burundi
Cambodia
Cameroon
Canada
Cape Verde
Central African Republic
Chad
Chile
China
Colombia
Comoros
Congo (Brazzaville)
Congo (Kinshasa)
Costa Rica
Croatia
Cyprus
Czech Republic
Denmark
Djibouti
Dominica
Dominican Republic
Ecuador
Egypt
El Salvador
Equatorial Guinea
Eritrea
Estonia
Ethiopia
Germany
Fiji
Finland
France
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
Gabon
Gambia
Georgia
Ghana
Greece
Grenada
Guatemala
Guinea
Guinea-Bissau
Guyana
Haiti
Honduras
Hong Kong
Hungary
Iceland
India
Indonesia
Iran
Ireland
Israel
Italy
Ivory Coast
Jamaica
Japan
Jordan
Kazakhstan
Kenya
Korea, South
Kuwait
Kyrgyzstan
Laos
Latvia
Lebanon
Lesotho
Liberia
Lithuania
Luxembourg
Macedonia (Skopje)
Madagascar
Malawi
Malaysia
Maldives
Mali
Mauritania
Mauritius
Mexico
Moldova
Mongolia
Morocco
Mozambique
Namibia
Nepal
Netherlands
New Zealand
Nicaragua
Niger
Nigeria
16
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
Norway
Oman
Pakistan
Panama
Papua New Guinea
Paraguay
Peru
Philippines
Poland
Portugal
Qatar
Republic of Yemen
Romania
Russia
Rwanda
Sao Tome and Principe
Saudi Arabia
Senegal
Seychelles
Sierra Leone
Singapore
Slovakia
Slovenia
Solomon Islands
South Africa
Spain
Sri Lanka
St Kitts and Nevis
St Lucia
St Vincent and the Grenadines
Sudan
Suriname
Swaziland
Sweden
Switzerland
Syria
Tajikistan
Tanzania
Thailand
Togo
Tonga
Trinidad and Tobago
Tunisia
Turkey
Uganda
Ukraine
United Arab Emirates
United Kingdom
Uruguay
Uzbekistan
Vanuatu
Venezuela
Vietnam
Western Samoa
Zambia
Zimbabwe
Table 2: Shipment size by mode of transport
Mode of
Physical shipment size (kg)
Value shipment size (USD)
transport
N.obs.
mean median
N.obs.
mean
median
Air
56,019
1,418
368 199,477
30,461
8,414
Ground
12,104 12,897
4,911
33,172 225,791
14,435
Vessel
84,678 39,163
5,391 208,174
71,291
17,727
All
152,801 23,244
1,990 440,823
64,441
11,940
Notes: Calculated for 10-digit products and 170 importers. Shipment size
is export value or quantity over number of shipments. Physical shipment
size is only for products in kg.
Table 3: Summary statistics of Doing Business indicators by continent
Continent
No.obs
Number of Documents
median min max
Africa
51
9
5
17
America
32
7
4
10
8
3
14
Asia
42
Europe
37
6
2
10
Pacic
8
6.5
4
13
Total
170
7
2
17
Source: Doing Business Survey 2010.
Total Days
median min
32
14
18.5
9
20.5
3
12
5
24
8
21
3
max
100
71
92
36
31
100
Total Cost (USD)
median min max
1813
577 6150
1400
730 2868
1125
439 4600
1160
620 1945
849
630 1392
1331
439 6150
Table 4: Procedures in total days and total cost
Procedure
Document preparation
Custom clearance and inspection
Port and terminal handling
Inland transportation from seaport
All
Source: Doing Business Survey 2010.
mean
13.7
3.7
4.5
4.7
26.6
Days
as % of total
51.7
14.0
16.8
17.5
100
s.e.
0.8
0.2
0.3
0.6
1.4
mean
306.1
213.7
317.0
778.0
1614.8
Cost (USD)
as % of total
19.0
13.2
19.6
48.2
100.0
s.e.
14.3
15.9
13.5
64.4
78.2
Table 5: Correlation coecients of the Doing Business indicators
Document Day
Custom Day
Port Day
Transit Day
Document Cost
Custom Cost
Port Cost
Transit Cost
GDP per capita
s.e.
Document
Day
1
0.66
0.35
0.44
0.40
0.22
0.24
0.54
-0.42
[0.000]
Custom
Day
Port
Day
Transit
Day
Document
Cost
Custom
Cost
Port
Cost
Transit
Cost
1
0.37
0.39
0.37
0.29
0.27
0.42
-0.37
[0.000]
1
0.20
0.28
0.21
0.29
0.31
-0.39
[0.000]
1
0.42
0.07
-0.07
0.81
-0.22
[0.003]
1
0.31
0.28
0.37
-0.24
[0.001]
1
0.27
0.10
-0.24
[0.002]
1
0.04
-0.12
[0.129]
1
-0.22
[0.005]
17
Table 6: Total Days
total
extensive
price
within
transport
prod.comp
0.964***
0.846***
-0.006
0.093***
-0.009
0.040**
[0.055]
[0.050]
[0.011]
[0.017]
[0.007]
[0.017]
Ln GDPPC
0.065
0.173**
0.020
-0.075***
-0.013
-0.040
[0.069]
[0.073]
[0.016]
[0.025]
[0.011]
[0.025]
DAYS
-0.010** -0.016***
0.001
0.005**
-0.001
0.001
[0.005]
[0.006]
[0.001]
[0.002]
[0.001]
[0.002]
(θc − θ̄)DAYS
0.056
0.503
0.187
-0.334
0.012
-0.311
[0.693]
[0.965]
[0.206]
[0.331]
[0.143]
[0.329]
LnT-MR Landlocked
61.27
41.34
0.244
-4.878
-3.922
28.49
[62.44]
[83.29]
[17.82]
[28.57]
[12.33]
[28.44]
LnT-MR Tari
-45.98
-97.36
-27.26
13.25
2.84
62.55
[227.70]
[292.68]
[62.61]
[100.38]
[43.32]
[99.93]
-1.538***
-1.739*** 0.074** 0.188*** -0.141***
0.079
LnT-MR Distance
[0.161]
[0.170]
[0.036]
[0.058]
[0.025]
[0.058]
LnT-MR PTA
-0.188
-0.080
0.058
-0.227**
0.063
-0.002
[0.305]
[0.291]
[0.062]
[0.100]
[0.043]
[0.099]
LnT-MR FTA
0.489
0.406
0.124
0.134
-0.149**
-0.027
[0.553]
[0.464]
[0.099]
[0.159]
[0.069]
[0.159]
0.311
0.399
-0.084
-0.111
-0.029
0.136
LnT-MR Common Language
[0.291]
[0.269]
[0.058]
[0.092]
[0.040]
[0.092]
LnT-MR Colonial Ties
1.753***
1.613
0.110
-0.059
-0.250
0.339
[0.641]
[1.151]
[0.246]
[0.395]
[0.170]
[0.393]
Constant
-11.84***
-11.52***
-0.160
-0.528**
0.383***
-0.014
[0.774]
[0.677]
[0.145]
[0.232]
[0.100]
[0.231]
N
170
170
170
170
170
170
Adj. R2 /Pseudo R2
0.86
0.86
0.06
0.40
0.27
0.11
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust standard errors.
Gravity on the margins is estimated with Seemingly Unrelated Regressions estimation.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
Ln GDP
Table 7: Total Cost
total
extensive
price
within
transport
prod.comp
0.956***
0.836***
-0.007
0.094***
-0.008
0.041**
[0.054]
[0.048]
[0.011]
[0.018]
[0.007]
[0.017]
Ln GDPPC
0.043
0.160**
0.021
-0.095***
-0.013
-0.029
[0.065]
[0.066]
[0.015]
[0.024]
[0.010]
[0.023]
-0.512*** -0.746***
0.007
0.134**
-0.041*
0.133***
Ln COST
[0.155]
[0.145]
[0.032]
[0.053]
[0.022]
[0.050]
1.282
1.927
0.370
-0.406
-0.093
-0.518
(θc − θ̄)COST
[0.845]
[1.299]
[0.288]
[0.472]
[0.199]
[0.454]
LnT-MR Landlocked
98.682
92.859
0.084
-11.444
-1.694
18.878
[65.368]
[80.877]
[17.947]
[29.388]
[12.373]
[28.252]
LnT-MR Tari
-172.801
-177.980
-14.747
-34.654
12.961
41.619
[154.502]
[203.039]
[45.056]
[73.778]
[31.061]
[70.925]
-1.637***
-1.887***
0.074** 0.224*** -0.148***
0.101*
LnT-MR Distance
[0.146]
[0.164]
[0.036]
[0.060]
[0.025]
[0.057]
LnT-MR PTA
-0.197
-0.091
0.061
-0.236**
0.060
0.009
[0.284]
[0.280]
[0.062]
[0.102]
[0.043]
[0.098]
LnT-MR FTA
0.401
0.230
0.123
0.205
-0.165**
0.008
[0.496]
[0.441]
[0.098]
[0.160]
[0.067]
[0.154]
LnT-MR Common Language
0.327
0.434*
-0.087
-0.132
-0.025
0.137
[0.284]
[0.258]
[0.057]
[0.094]
[0.039]
[0.090]
LnT-MR Colonial Ties
1.698***
1.549
0.147
-0.082
-0.275
0.360
[0.557]
[1.116]
[0.248]
[0.405]
[0.171]
[0.390]
Constant
-8.119***
-6.325***
-0.203
-1.197**
0.645***
-1.039**
[1.526]
[1.370]
[0.304]
[0.498]
[0.210]
[0.479]
N
170
170
170
170
170
170
Adj. R2 /Pseudo R2
0.87
0.87
0.06
0.38
0.28
0.14
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust standard errors.
Gravity on the margins is estimated with Seemingly Unrelated Regressions estimation.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
Ln GDP
18
Table 8: Day indicators by procedure - included one-by-one
total
extensive
price
within
transport
prod.comp
Document Day
-0.012
[0.008]
-0.018*
[0.010]
-0.001
[0.002]
0.007**
[0.003]
-0.001
[0.001]
0.001
[0.003]
Customs Day
-0.002
[0.034]
-0.043
[0.036]
0.008
[0.007]
0.016
[0.012]
0.012**
[0.005]
0.005
[0.012]
Document+Customs Day
-0.009
[0.007]
-0.016*
[0.008]
0.000
[0.002]
0.006**
[0.003]
0.000
[0.001]
0.001
[0.003]
Port Day
-0.059*
[0.030]
-0.067**
[0.028]
0.010
[0.006]
0.002
[0.010]
-0.001
[0.004]
-0.002
[0.010]
Transit Day
-0.022
[0.016]
-0.025
[0.021]
0.004
[0.004]
0.006
[0.007]
-0.006**
[0.003]
-0.001
[0.007]
Port+Transit Day
-0.024** -0.029**
0.003
0.006
-0.004**
0.000
[0.011]
[0.013]
[0.003] [0.005]
[0.002]
[0.004]
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust
standard errors. Gravity on the margins is estimated with Seemingly Unrelated Regressions
estimation. Results are from gravity regressions with only one administrative barrier indicator.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
Table 9: Day indicators by procedure - included jointly
total
extensive
price
within transport prod.comp
-0.011
-0.010
-0.003
0.005
-0.003
-0.001
[0.010]
[0.012]
[0.003] [0.004]
[0.002]
[0.004]
0.070
0.027
0.006
0.002
0.024***
0.010
Customs Day
[0.047]
[0.047]
[0.010] [0.016]
[0.007]
[0.016]
-0.055*
-0.053
0.009
-0.005
-0.002
-0.005
Port Day
[0.031]
[0.033]
[0.007] [0.011]
[0.005]
[0.011]
Transit Day
-0.041
-0.049
0.003
0.013
-0.011**
0.003
[0.027]
[0.032]
[0.007] [0.011]
[0.004]
[0.011]
Document+Customs Day -0.003
-0.009
-0.001
0.005
0.001
0.001
[0.008]
[0.009]
[0.002] [0.003]
[0.001]
[0.003]
Port+Transit Day
-0.023*
-0.025
0.004
0.004
-0.005**
0.000
[0.013]
[0.016]
[0.003] [0.005]
[0.002]
[0.005]
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust
standard errors. Gravity on the margins is estimated with Seemingly Unrelated Regressions
estimation. Results from gravity regressions with procedures included jointly.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
Document Day
19
Table 10: Cost indicators by procedure - included one-by-one
total
extensive
price
within
transport
prod.comp
-0.271**
[0.137]
-0.414***
[0.141]
-0.013
[0.030]
0.062
[0.050]
-0.006
[0.021]
0.100**
[0.047]
Log Custom Cost
0.131
[0.104]
0.004
[0.090]
-0.032*
[0.018]
0.120***
[0.030]
0.011
[0.013]
0.028
[0.030]
Ln Doc+Custom Cost
-0.100
[0.151]
-0.298**
[0.138]
-0.038
[0.029]
0.148***
[0.047]
0.009
[0.020]
0.079*
[0.046]
Log Port Cost
-0.330***
[0.124]
-0.456***
[0.134]
0.040
[0.028]
0.007
[0.048]
-0.008
[0.020]
0.087*
[0.045]
Log Transit Cost
-0.275***
[0.091]
-0.366***
[0.088]
0.015
[0.019]
0.055*
[0.031]
-0.035***
[0.013]
0.056*
[0.030]
Log Document Cost
Ln Port+Transit Cost
-0.448*** -0.599***
0.028
0.067
-0.042**
0.098**
[0.117]
[0.116]
[0.026]
[0.043]
[0.018]
[0.041]
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust
standard errors. Gravity on the margins is estimated with Seemingly Unrelated Regressions
estimation. Results are from gravity regressions with only one administrative barrier indicator.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
Table 11: Cost indicators by procedure - included jointly
Log Document Cost
Log Custom Cost
Log Port Cost
Log Transit Cost
total
extensive
price
within
transport
prod.comp
-0.224
[0.144]
0.212**
[0.101]
-0.294**
[0.146]
-0.262**
[0.108]
-0.288**
[0.146]
0.127
[0.097]
-0.346**
[0.154]
-0.358***
[0.121]
-0.011
[0.033]
-0.036*
[0.022]
0.050
[0.034]
0.016
[0.027]
-0.001
[0.053]
0.108***
[0.035]
-0.045
[0.056]
0.061
[0.044]
0.003
[0.023]
0.013
[0.015]
-0.019
[0.024]
-0.026
[0.019]
0.074
[0.052]
-0.001
[0.035]
0.065
[0.055]
0.045
[0.043]
Ln Doc+Custom Cost
0.039
-0.120
-0.049
0.128**
0.026
0.053
[0.164]
[0.141]
[0.031]
[0.051]
[0.021]
[0.049]
Ln Port+Transit Cost -0.461*** -0.578***
0.038
0.043
-0.050***
0.086**
[0.126]
[0.124]
[0.027]
[0.045]
[0.019]
[0.044]
Notes: Standard errors in brackets. Gravity on total trade is estimated with OLS and robust
standard errors. Gravity on the margins is estimated with Seemingly Unrelated Regressions
estimation. Results from gravity regressions with procedures included jointly.
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
20