CUSTOMS OBSTACLES AND

DECISION TO IMPORT

by

Anton Vorush

A thesis submitted in partial fulfillment of

the requirements for the degree of

MA in Financial Economics

Kyiv School of Economics

2013

Thesis Supervisor: Professor Elena Besedina

Approved by _________________________________________________
Head of the KSE Defense Committee, Professor Wolfram Schrettl

_____________________________________________________
_____________________________________________________
_____________________________________________________

Date ________________________
 Kyiv School of Economics

Abstract

CUSTOMS OBSTACLES AND

DECISION TO IMPORT

by Anton Vorush

Thesis Supervisor: Professor Elena Besedina

This thesis examines time to clear the customs as a factor of influence on the
decision to import. Implicit cost of the transportation is a much greater barrier
for imported products than direct monetary costs of crossing the border - a
variety of duties, thus requiring detailed study and subsequent implementation
of solutions.

We used the data on import for companies from all available rounds of the
BEEPS. To fill the missing data interpolation and imputation techniques were
used. To study the effect of time clearing the customs on the percentage of
imported inputs, OLS regressions were used on all the samples – initial
benchmark, interpolated and imputed.

The results show that a 10-day delay in customs clearing of imported goods, on
average, reduces their imports by 1.6% (4.1% for Ukraine). Government
subsidies, being a tool to protect domestic producers, also reduce imports. At
the same time 100% foreign ownership of the company is associated with a
15% increase in imports.
 TABLE OF CONTENTS

Chapter 1 INTRODUCTION ..................................................................................1

Chapter 2 LITERATURE REVIEW ......................................................................4
Chapter 3 METHODOLOGY AND DATA .................................................... 10
3.1 Variables overview .......................................................................................... 10
3.2 Dealing with variables..................................................................................... 12
3.2.1 Interpolation .................................................................................................... 15
3.2.2 Multiple imputation ........................................................................................ 16
3.3 Limitations ........................................................................................................ 19
Chapter 4 EMPIRICAL RESULTS ...................................................................... 20
4.1 Benchmark sample: OLS ............................................................................... 20
4.2 Interpolation ..................................................................................................... 21
4.3 Multiple imputation ........................................................................................ 22
4.4 Case of Ukraine ............................................................................................... 26
4.5 Tests ................................................................................................................... 26
4.6 Policy implications .......................................................................................... 27
Chapter 5 CONCLUSIONS................................................................................... 29
WORKS CITED ....................................................................................................... 31
 LIST OF TABLES

Number Page

Table 1. Missing variables pattern ...................................................................... 13

Table 2. Summary statistic on the benchmark OLS sample........................... 14
Table 3. Summary statistic on the interpolated sample ................................... 15
Table 4. Efficiency of an imputed estimate ...................................................... 16
Table 5. Comparison of all samples ................................................................... 18
Table 6. Comparison of all results...................................................................... 25
Table 7. Tests results ............................................................................................ 27

ii
 Chapter 1

INTRODUCTION

Throughout the time, international trade develops more and more impetuously,
opening the borders and increasing the number of suppliers (exporters) as well
as clients – importers. Many firms in order to obtain some scarce inputs or
inputs of the better quality, or inputs at a lower price, import them rather than
buying them from domestic suppliers. Moreover, availability of importing
sources increases products (in particular, inputs) variety - not only in terms of
range, but also in terms of choice of variability of trade conditions with different
trade partners as well. Furthermore, importing inputs may help firms to
diversify the chain of suppliers in order to reduce specific company or country
risk that any firm may be subject to. Nevertheless, despite the greater trade
openness and globalization, both of which makes foreign goods being
perceived at the same level of availability as domestic, firms suffer from various
obstacles when importing. These obstacles are mainly represented by the
delivery time (which, in fact, contains many aspects) that in its turn influences
importing costs.

Delivery time is indeed a very broad definition that may be viewed from
different perspectives and has dozens of pitfalls. In general, customers are ready
to pay more for goods transportation in order to prevent downtime. However,
customers may not be willing to overpay for the faster delivery if a supplier is
relatively close, but they may be willing if the transportation time constitutes
several weeks. Similar situation concerns the product price – if the
transportation price is too high comparing to the product price, customers may
refuse to pay high transportation tariffs in favor of slower delivery. Implicit
costs of transportation time are much more an obstacle for foreign goods than
direct costs of crossing the border – import or export tariffs. Another
determinant in transporting time is the time needed to deliver goods from the

1
factory to the border of exporting country and from the border of importing
country to the customer.

One of the delivery time components that has not been researched much yet is
the time needed for goods clearing at the customs. This, in its turn, depends on
many other factors such as regulation, the level of corruption, border openness,
size of the parcel waiting for export or import and its declared contents. About
24% of the transportation time is spent at borders, while a more reasonable
customs-clearing time should amount to 2% of the total transportation time
that is pre-Schengen target (Nordås et al., 2006). Meanwhile Djankov (2010)
estimates that 75% of transportation delays were caused by administrative
barriers — various customs and tax procedures, customs clearances and cargo
control. If the time spent on customs is too long, firms may decide to stop
importing and change their suppliers to the domestic ones.

I will be using micro-level data from the BEEPS survey conducted by the
EBRD and the World Bank in order to study the relationship between the time
for customs clearing of imported inputs and the percentage of foreign inputs
used by firms. All the data are publicly available. There were already conducted
four rounds of the survey (in 1999, 2002, 2005 and 2009) based on the results
from 29 countries. However, there is an additional round of the survey for
Russia, which became available at the end of the last year that provided more
observations (the sample of more than 4000 firms), so the results will be
ultimately up-to-date and highly relevant.

Therefore, the dependent variable of my interest is “Percentage of material

inputs and supplies of foreign origin in the last fiscal year”, while the exogenous
variable of the main interest is “Average number of days for imported goods
to clear customs in the last fiscal year”. Other variables that are relevant to the
model should be also controlled for as well as the robustness check performed.

2
Similar research has been done on slightly different topics – like delivering time
(including using different transportation modes (Hummels et al, 2012)) and its
influence on the import quantities (Nordås et al., 2006), the influence of trade
on the financial sector (Bourgeon et al., 2012). I will scrutinize them more
meticulously in the next section of the thesis. Therefore, this thesis work
explores new aspects in the topic of import, specifically import of inputs as well
as dealing with customs.

Results show that most of the variables used are significant. Every 10 days of
custom-clearance-delay on average leads to 1.6% decrease in foreign inputs
import (4.1% for Ukraine). Customs obstacles lead to frequent informal
payments when clearing the customs that, in its turn, facilitate international
trade increasing inputs import by 4%. Fully-foreign-owned firms tend to
import 15% more. However, government subsidies serving as a protectionism
policy and the firm size disincentive import by 10% each.

The thesis is organized as follows. Chapter 2 covers the existing literature

related to trade barriers. Chapter 3 describes the methodology and data used,
variables overview and limitations. Chapter 4 talks about obtained results and
tests, while Chapter 5 concludes the thesis, gives some recommendations and
policy implications.

3
 Chapter 2

LITERATURE REVIEW

The issue of trade barriers, trade costs and, as a consequence, trade facilitation
has been studied much since the boom in the international trade. Some works
were done in the previous century; however, the most prominent and valuable
are from the current millennium. This is because progress in technology
development allows us to differentiate trade costs by using different
transportation modes and cut down on some costs due to the faster and easier
procedures of clearing the customs. Surely, over time, the trading time
importance has decreased because its scale is much less now, but due to the
faster pace of life, every day is on count anyway. Therefore, I will explore this
issue.

In the overwhelming majority of the papers, trading costs are primarily

associated with transportation costs and time required for delivery. One of the
most important contributions in this area Hummels et al. (2012) study the
impact of transportation time associated with different transportation modes
on the trade sensitivity to time. Long transportation time implies larger
inventories that, in its turn, entail larger costs for the time holding and
depreciation costs as well. Bigger inventories are needed in order to prevent
disruptions in assembly line if the cargo is being delayed. All these extra costs
are ready to be paid by customers to prevent downtime.

Hummels et al (2012) develop a model that reveals consumers’ perception of

the service quality that directly depends on delivery time. This factor makes
producers improve customer experience by shipping products by air to obtain
faster delivery time. The costs of transportation are charged proportionally to
quantity rather than value of goods shipped by cargo carriers. Therefore, the
ad-valorem cost of airfreight diminishes in product prices, i.e. the higher prices
are for the product, the less part transportation will constitute in it. Hence, firms

4
with the high price products will be more likely to use airfreight, while firms
with low-price products will be using cheaper ocean shipping. A consumer may
not be willing to overpay for the fast delivery to save just a few days if a supplier
is relatively close to him, but he may want to if a supplier is overseas and the
ocean transit takes many weeks.

The authors show that the airfreight is used not only because consumers value
the transportation time, but also because consumers care about ad-valorem
premium. Hence, the more time ocean transit takes, the more often the
airfreight is chosen. Researchers find that a day of transportation is equivalent
to 0.6 - 2.3 % of an ad-valorem tariff and that customers who buy inputs for
their production as parts and components, seem to be time sensitive and more
likely to use air freight. Generally speaking, those, for whom time is crucial
criterion in the transportation, would incur higher costs of air freight in order
to gain higher customer satisfaction and loyalty.

However, there is no just single effect. In one of his previous papers Hummels
(finds that firms are not only willing to pay a bit less than 1 percent of ad-
valorem tariff for reducing one day shipping by ocean in favor of air
transportation, but also that each additional transportation day lessens the
probability of importing goods on average by 1 percent (Hummels et al., 2001).

Although this work estimates time differences that arise due to different
transportation modes, it is recognized to be the reference paper when talking
about trading time in general. It reinforces the significance of any time lag in
goods transportation – such as transportation within the exporting country,
international shipment (whatever mode is chosen), delivery within destination
county and, finally, the time spent to clear the customs (both, of exporting and
destination country) itself.

In addition to shipping time, some works consider the effect of time needed
for import and export procedures on the international trade flows. In his other

5
paper, Hummels (2007) tries to predict the influence of the exogenous shock
on trade. He estimates that after the September 11, 2001 terrorist attack,
strengthening the security would impose bigger costs for international
transactions. He assumes that all the security checks for imported goods would
add a day to the time of shipping. Taking into account that manufactured
import is over $800 billion a year, this extra day delay would constitute a loss
of about $7 billion annually. "You could be talking about very serious expenses
if goods are subject to even minimal scrutiny on the way in" says Hummels.

In the same paper, the author researches the topic closely related to the topic
of this thesis. He uses the World Bank “Doing Business” data on the number
of days and number of required documents for preparing international
transportation to estimate the tariff equivalent of delays at customs and in ports.
He uses both estimates of daily transportation costs and data on days spent to
clear customs and port delays and finds that the time spent at customs is a much
larger obstacle for trading than import/export tariffs are. The author also
notices that it is essential to take into account the time of goods transportation
from the factory to the border (port). This time may be much longer than
international transportation time itself and there is larger heterogeneity among
countries for this indicator rather than for the distance between them.

Nevertheless, it is not always possible to assign monetary values to time costs.

Harrigan and Venables (2004) assert that valuation of these costs differs
because of uncertainty in time. Delivery delays can pluck production, so
producers have to order their supplies before their actual needs. In their study,
the authors have focused on just-in-time production methods when supplies
are bought and delivered exactly when they are needed, so as to avoid
inventories costs. This issue is closely related to my thesis topic as customs
delays cannot be estimated or predicted in advance, so they may become a real
obstacle in just-in-time production method by creating large inventories. In
addition, the authors argue that delays also matter because of physical

6
depreciation or obsolesce of goods during the long shipping time (including all
stages of transportation). When the situation is uncertain, advantages of the
established and certain timely delivery of components changes the efficient
dimensional production organization and leads to clustering.

Nordås et al. (2006) study the influence of time for import and export as well
as logistics on gross international trade flows. They find that the time of the
procedures for importing and exporting not only reduces volumes of trade, but
also the probability that the company will enter international markets or use
international resources. Grossman and Helpman (2005) in a theoretical model
show that firm‘s decision whether it is worth outsourcing intermediate inputs
depends on factors such as barriers to enter the market, contracting costs as
well as customization of inputs technology. Halpern, Koren, and Szeidl (2011)
explore the firm’s decision to import intermediaries. Nordås with co-authors
also find that more and more products become time-sensitive (especially labor-
intensive ones), but the liberalization of transport services may be implemented
with comparatively few costs in developing countries. Moreover, they suggest
that bureaucracy related to import and export in developing countries should
itself reduce incentives for local producers to export or import time-sensitive
products.

Trading time becomes much more crucial in the situation with only a few trade
barriers. Martinez-Zarzoso and Novak-Lehmann (2006) find that when there
are no official trade limitations, trading time becomes the main obstacle for
international trade. Taking into account that trade barriers have substantially
decreased during the last couple of decades, we could logically assume that
trading time is a significant determinant of trade patterns.

Later on, this issue has been further investigated by using time of cargo
transportation from the factory to the ship in the nearest port (including
clearing customs) by Djankov et al. (2010). They find that each additional day
of delay reduces trade volumes on average by 1% or equally, increases the

7
distance between trade partners by 70 km. They also find that customs delays
negatively affect trade concerning mostly time-sensitive goods such as
agricultural products.

Some papers consider different sides of international trading, for example

financial aspect of the issue, in addition to trading time. In particular, Bourgeon
et al. (2012) study the effect of financial restrictions and cash flows on the trade
flows. They use the notion of trading time that constitutes the time lag between
production of manufactured goods and receiving payments. While studying
financial restrictions, they are primarily concerned about uncertainty in
payments from the importer. In their estimations, the distance indicator is used
as a compound of shipping time and time spent at the border. They find that
financial constraints have substantial impact on trade as well as trading time.

Relationship between imported inputs and firm’s efficiency is another

interesting topic that worth studying, so Amiti and Davis (2012) study the
influence of import tariffs on wages paid. Amiti and Konings (2005) empirically
show that importing firms win in productivity the most when input tariffs are
abated. A 10% decrease in input tariffs leads on average to 3% productivity
rise. The results are shown to be robust to inclusion of separate effects for the
period of crisis in Asia. Moreover, when regressing the firms’ productivity only
on final goods tariff (as is commonly done in the other works), the coefficient
of marginal effect is even more than doubled. This means that if input tariffs
are excluded, an omitted variable problem may arise.

Empirical studies for Ireland show that firms’ productivity can be improved by
outsourcing inputs, which is more often observed in companies with foreign
ownership (Görg and Hanley (2004); Görg and Hanley (2005); Görg, et al.
(2007)). This result is also confirmed for Hungary (Halpern et al. 2005). The
authors find that imported inputs affect positively the plant efficiency through
complementarity and quality channels.

8
The works analyzed above do not seem to give a complete picture of the
decision to import inputs. This thesis aims to remedy this gap and is positioned
as a complement to the work on export decision. Even though decisions to
export and import may be related somehow, I am convinced that importing
and exporting firms are guided by different drivers in their decision-making.
The export decision is based on the foreign markets demand and its
peculiarities, while importing-inputs-decisions take its grounds in the
production process features at the firm’s level, government protectionism
policy.

Since transportation time studied in the above papers is related to the customs
clearing time I will refer to the methodology employed in the above papers. In
the next section I will present methodological issues in more detail.

9
 Chapter 3

METHODOLOGY AND DATA

The main question of the thesis is whether the time spent at the border for
customs clearing can explain the variation in the firms’ input source structure.
In order to answer this question the below mentioned baseline model was used.
It is developed by the thesis author to become as follows:

𝐼𝑚𝐼𝑛𝑝𝑢𝑡𝑖𝑡 = 𝛽0 + 𝛽1 ∙ 𝐶𝑢𝑠𝑡𝐷𝑎𝑦𝑠𝑖𝑡 + 𝛽2 ∙ 𝑋𝑖𝑡 (1)

where Xit is a vector of secondary priority exogenous variables.

3.1 Variables overview

The dependent variable is the percentage of material inputs and supplies of

foreign origin in the last fiscal year (defined as ImInput), while the exogenous
variable of our main interest is the average number of days for imported goods
to clear customs in the last fiscal year (CustDays). Coefficient on this variable is
expected to have a negative sign because every additional day of delay may be
an obstacle for production and the company may want to change foreign
suppliers for domestic ones.

Other variables that should be controlled for constitute the vector X, and are
listed further. Direct supply chain can mean less transportation and transaction
cost as well as faster and more reliable supply, so the expected coefficient sign
on the dummy variable whether any of these inputs and supplies imported
directly (ImDirect) is positive. The level of the obstacle for transportation of
goods, supplies, and inputs (TranspObs), as well as the obstacle level represented
by the customs and trade regulations (CustObs) – are categorical variables. These
two variables are the pure measures of obstacles to import goods in the survey.
Obviously, expected signs on their coefficients are negative.

10
Foreign owners are assumed to be more loyal to foreign purchases and willing
to use foreign (imported) inputs (Halpern, 2011), so the coefficient on
percentage of the firm owned by private foreign individuals, companies or
organizations (Foreign) is expected to have positive sign. Dummy variable
(GovSub) indicates whether the company received governmental financial
support or not over the last 3 years, which may either show the good terms
with the government or operating in a privileged and subsidized (protected)
sector. Hence, if the company is in the protected sector, the government
implements protectionism policy and the expected sign is negative. The variable
of additional payments/informal gifts frequency to deal with customs/imports
by firms like the respondent (InformGift) represents corruption level in the
country. Expected sign of the coefficient on this variable is unclear as from the
one hand, higher corruption imposes additional costs and inconveniences for
importing firms, but from the other hand, this facilitates trade after all. The firm
size (Size) is unclear as well. Positive sign may be caused as larger firms are
assumed to have more need in resources, which may be scarce or very specific
and not present at the domestic market in the required amount, as well as they
are assumed to have more resources to enter international markets. However,
negative sign may be caused by the fact that larger companies may be more
vertically integrated, and produce their inputs themselves. Finally, the control
for year (Year) and 2008 is expected to have a negative influence on imports
due to the global financial crisis.

All the data taken from one source - Business Environment and Enterprise
Performance Survey (BEEPS) conducted by EBRD1 and World Bank as it is
micro-level data and cannot be combined with the data from other sources.
The used dataset is a panel collected from 4 rounds of survey
(29,716 observations) plus an additional round for Russia in 2012

1
http://www.ebrd.com/russian/pages/research/economics/data/beeps.shtml

11
(4,222 observations). Hence, overall there are almost 34 thousands of
observations in the initial data. Although the data are panel, they are very
unbalanced, so pooled data were used in the model.

From the initial whole sample of the data the following statistics can be
presented:

There are representatives that fully depend on imported inputs (10% of the
firms) as well as those that depend fully on domestic suppliers - 44% of firms
did not import at all in the year prior to the survey. From those firms that
import, on average respondents imported 56% of their material inputs and
supplies with roughly one third of them imported directly. In the range of time
spent clearing the customs between zero days and 1 calendar year, firms
reported for their inputs to spend on average 5 days clearing the customs.
However, it should be mentioned that only half of the firms that import
responded to the question. Transportation, clearing customs and trade
regulations were slight obstacles for the firms according to the survey.
Moreover, most of the firms never paid additional payments or gave informal
gifts to deal with customs. Just a small share of firms (8%) received government
subsidy over the last 3 years and the same share of firms were hold by foreign
individuals.

3.2 Dealing with variables

For all the categorical variables, answers that could not be applied to the firm
or refusals to answer that initially produced negative values were substituted by
missing values.

Some tests on missing data patterns show that if the data on ImInput are missing
then there are no data on ImDirect, which is obvious. The next causality is - if
there are no data on GovSub then there are no data on InrofmGift either. There
are 3,275 missing observations, so this is not a coincidence. This may be due to

12
the fact that a person responsible for the survey filling in does not gain such
financial (and often informal) data or the firm did not want to reveal such data.

Initially, only 13% of observations had the data on all the variables, 28% had
one variable missing, 14% had two variables missing and 22% had three
variables missing.

Concerning variables of our interest, ImInput and CustDays, 18% of

observations have answers for both variables. Only these observations were
used in all the regressions for them to be most trustworthy. The half of the
variables have information on the ImInput, but do not have on CustDays – the
exogenous variable of main interest; and one third of observations do not have
information on both of these variables. Such observations were dropped as well
as observations with more than three variables missing because they
contributed very little.

As a result, the sample is reduced to 6 thousands of observations with already

73% of them having information on all the variables, 12% missing only one
variable and 13% missing two variables (Table 1).

Table 1. Missing variables pattern

Missing
vars Freq. Percent
0 4,493 72.97%
1 734 11.92%
2 779 12.65%
3 151 2.45%
Total 6,157 100.00%

Descriptive statistics for all variables from the benchmark sample as well as
expected signs are presented in Table 2 below:

13
Table 2. Summary statistic on the benchmark OLS sample
Variable Obs Mean Std. Min Max Exp.
Dev. sign

% of material inputs and 4493 61.922 33.046 1 100 N/A

supplies of foreign origin

Avg # of days for 4493 5.246 9.947 0 365 -

imported goods to clear
the customs

Were any of these 4493 0.444 0.497 0 1 +

material inputs and
supplies imported
directly?

How much of an 4493 0.705 1.026 0 4 -

obstacle is transportation
of goods, supplies, and
inputs?

How much of an 4493 1.382 1.151 0 4 -

obstacle are customs and
trade regulations?

What % of this firm do 4493 21.622 37.684 0 100 +

the private foreign
individuals, companies
or organizations own?

Over the last 3 years, has 4493 0.112 0.316 0 1 -

this establishment
received any government
subsidies?

How often do firms like 4493 1.103 1.473 0 5 +/-

you pay additional
payments/informal gifts
to deal with
customs/imports?

Size of the firm 4493 1.086 0.824 0 2 +/-

14
As was mentioned, half of the variables are either categorical or dummy.
Categorical variables to become more relevant were transformed by eliminating
answers, driven by survey specification of answer refusals, “Do not know” and
”Does not apply” that produced negative values.

3.2.1 Interpolation

In addition to the initial (benchmark) sample, interpolation was made within all
the variables for them to have no missing observations, thus OLS to use more
observations in the regression. At first, as it was already mentioned, all missing
observations on ImInput and CustDays were dropped as there were three fourth
of the data missing, and interpolating would mean actually more generating new
data rather than interpolating of the existing data. After this, 6157 observations
were left. All missing values for other variables were generated from normal
distribution with mean and standard deviation taken from observations within
a country of missing observation. See summary statistic for the interpolated
data at the Table 3 below.

Table 3. Summary statistic on the interpolated sample

Variable Obs Mean Std. Dev.
ImInput 6157 60.7125 32.9363
CustDays 6157 5.4598 10.8483
ImDirect 6157 0.4829 0.4997
TranspObs 6157 0.7676 1.0815
CustObs 6157 1.4012 1.1856
Foreign 6157 20.0408 36.6863
GovSub 6157 0.1189 0.3237
InformGift 6157 1.1267 1.4175
Size 6157 1.0520 0.8172

15
3.2.2 Multiple imputation

The other technique used to deal with missing data is multiple imputation.
Imputation is a Monte Carlo technique that replaces missing data with the
simulated one using an M number of imputations (iterations). M is
recommended to be equal to the percentage of missing values (Rubin, 1987),
given that in our case missing values account for 22% we have generated
M = 22 imputations. Each of the imputed complete dataset is further analyzed
using standard techniques separately, but the results are combined to produce
estimated and standard errors to incorporate the uncertainty on missing values.
This method was initially developed to deal with large public surveys with
missing data present, so its use is quite native. This method was further
developed by Schafer (1997) to deal with missing data within several variables.

The efficiency of an estimate is calculated by the formula:

𝛾
(1 + 𝑚)−1 (2)

where γ is a non-response rate (Rubin, 1987). The efficiency comparison under

different levels of missing observations can be found in Table 4.

Table 4. Efficiency of an imputed estimate

γ
m
0.1 0.3 0.5 0.7 0.9
3 97 91 86 81 77
5 98 94 91 88 85
10 99 97 95 93 92
20 100 99 98 97 96

Hence, having 22% non-response rate and the same number of imputations
should produce extremely accurate estimate.

16
Indeed, summary statistics does not change much after imputations are made
(Table 5).

17
 Table 5. Comparison of all samples
Number of observations Mean Standard Deviation
Variable Raw Raw Raw
OLS Interpol. Imput. OLS Interpol. Imput. OLS Interpol. Imput.
sample sample sample
ImInput 4493 6157 6157 6157 61.9220 60.7125 60.7125 60.7125 33.0460 32.9363 32.9363 32.9363
CustDays 4493 6157 6157 6157 5.2460 5.4598 5.4598 5.4598 9.9470 10.8483 10.8483 10.8483
ImDirect 4493 6157 6157 6157 0.4440 0.4829 0.4829 0.4829 0.4970 0.4997 0.4997 0.4997
TranspObs 4493 6104 6157 6157 0.7050 0.7664 0.7676 0.7682 1.0260 1.0823 1.0815 1.0843
CustObs 4493 6064 6157 6157 1.3820 1.4014 1.4012 1.3986 1.1510 1.1877 1.1856 1.1869
Foreign 4493 6084 6157 6157 21.6220 19.7922 20.0408 19.7884 37.6840 36.6469 36.6863 36.6512
18

GovSub 4493 5281 6157 6157 0.1120 0.1129 0.1189 0.1150 0.3160 0.3164 0.3237 0.3190
InformGift 4493 4770 6157 6157 1.1030 1.1000 1.1267 1.0867 1.4730 1.4696 1.4175 1.4640
Size 4493 5894 6157 6157 1.0860 1.0512 1.0520 1.0555 0.8240 0.8215 0.8172 0.8196
3.3 Limitations

There seems to be an involved problem with using time to clear the customs
as an exogenous variable for an amount to import. Transport capacity and
frequency of travels obviously depend on trade volumes; hence, the direction
of causality can therefore be opposite. (Djankov et al., 2010)

Test on statistical difference across categories of transportation obstacles show

that they are not statistically different for both benchmark and imputed
samples. Categories of other variables are statistically different.

It should be also mentioned that the dataset may be subject to the selection
bias. It may occur if some small companies (which, nevertheless, import their
inputs) refuse to answer a long survey due to the lack of free human resources
or they simply were not asked to complete the survey because of the limit
number of interviews for each country or because company did not met formal
criterion for the survey2.

When interpolating, there could be a mistake in choosing the correct data

distribution, so interpolated results may not be the best and should not be
considered as a reference.

However, imputed sample produced reliable results and will be referred to

when concluding results.

http://www.ebrd.com/russian/downloads/research/economics/microdata/beeps_repo
rt_ebrd_april10.pdf

19
 Chapter 4

EMPIRICAL RESULTS

4.1 Benchmark sample: OLS

The first step to take is to run OLS on the primary sample (without
interpolation) to establish a benchmark for further comparison. The
benchmark sample includes only those firms that have information for all the
variables; hence, reducing the used sample to around 4.5 thousands
observations. All variables are in levels as our dependent variable already
represents percentages. Regression R-squared equals to 10%, with intercept
capturing two thirds of the variance in imports. Despite this, most of the
variables are statistically significant (except for some categories of GovSub,
TranspObs and CustObs).

Tests on joint significance of categorical variables show that transportation

obstacles are jointly significant, but different types of obstacles do not seem to
be particularly important for importing inputs decision. While custom
obstacles, informal payments and firm size are both jointly significant and
significantly different between categories. For example, only major customs
obstacles seem to be highly important for the importing decision.

The variable of our main interest (CustDays) has the expected sign and is highly
significant, but does not appear to be very influential in terms of magnitude –
every 10 additional days clearing customs, ceteris paribus, on average lowers
import by only 1%. Moreover, even small obstacles in transportation lower the
quantity of imported goods by almost 3%. On the contrary, if goods are
imported directly, import is higher by 5.7%.

One counter-intuitive result is that customs and trade regulations obstacles

increase import by 3-6% depending on the magnitude of the obstacle. A
possible explanation is that firms may use informal payments to deal with

20
customs. Moreover, only coefficients on frequent payments are statistically
significant and increase import by 6-11%! This result, in turn, does not come
as a surprise – more frequent payments provide more protection (though
informal and illegal) to companies’ imported inputs and after such a payment
(actually, a bribe) there will be much less, or even no at all, further problems
with clearing customs.

The effect of government subsidy on the import is negative as it was expected

initially – firms provided with a subsidy imported less by 9.7%. This may be
explained by subsidies playing a role of a protectionist policy. The size of the
firm has negative sign that may be explained, as it was mentioned above, by the
fact that large firms may have their own production of all the inputs, while small
firms need to use third-party inputs, imported in case the domestic suppliers
are absent. The coefficient on firm size is highly significant and shows that
medium firms and large firms have lower share of imported inputs (by 8.9%
and 14.7% respectively) compared to small firms. In the year 2008 firms
imported less by 12.3% compared to 2002, presumably due to the world
financial crisis.

Detailed statistics on benchmark OLS results can be found in Table 6. It is easy

to see that the sample used for OLS is very different from initial statistics, so
the results may be biased due to selection problem.

4.2 Interpolation

After interpolation, we can observe that data are very close to our initial sample
and indeed, OLS regression produces very similar R-squared (9.4%).

Regression coefficients have slightly changed as well. As was already said, this
may be due to the sample selection bias or due to the wrong distribution
assumption. The coefficient on the number of days clearing the customs
increases in magnitude. It may seem small, but with a week delay (which is quite

21
common in our country), import will be less by 1.4% instead of 1% decrease in
the benchmark OLS. That, in turn, may be a large amount of money if the
shipment scale is of large production size.

The variable of direct import captures 6.3% of the variability in imports. This
may be due to the collinearity of variables – if a company has direct import,
which means it is an importer a priory. Government subsidies, in contrast,
become less significant in domestic inputs protection. The firm size dummies
also lose its significance drastically.

Transportation obstacles have the expected signs and some of their categories
in fact may capture the effect customs-clearing time as the customs obstacle.

All categories of informal payments now increase consistently – more frequent

payments lead to much more import (up to +12%). Again, their positive effect
is explained by high trade regulations along with customs obstacles needed to
be mitigated.

Moreover, interpolated sample reveals additional statistically significant year

that affected import – 2012. Compared to 2002, import in each 2008 and 2012
was less by 12.5%.

However, this method predicts the values imperfectly and the regression based
on it has both some advantages and disadvantages over other models.
Nevertheless, the results became less significant than in the benchmark sample.
Detailed statistics on interpolated OLS results can be found in Table 6.

4.3 Multiple imputation

The OLS on imputed as well as on interpolated samples contained 37% more

observations comparing to the benchmark OLS. However, the results from the
multiple imputation model are much closer to the benchmark one as well as are
more significant than those from the interpolated sample, sometimes even

22
more significant than in our initial model. That is why, we will treat this results
as plausible and use them as final results.

Answering the main research question, we can conclude that every 10-day delay
of imported inputs to clear the customs results in 1.3% average decrease in
imports. Moreover, if inputs are imported directly without any intermediaries,
imports tends to increase by 6.5%. Firms can take advantage from direct
imports comparing to indirect because intermediaries are eliminated from the
imports chain, reducing cost and time.

Minor and moderate transportation obstacles may decrease import

correspondingly by 2 to 3 percent depending on the obstacle magnitude.
Customs and trade regulation obstacles of the major magnitude increase import
on average by 3.5%. This counterintuitive result will be explained later together
with the results on informal payments and gifts.

Companies fully owned by foreign individuals tend to import 15% more, so

there is quite a significant relationship between foreign ownership and the share
of imported goods. Government subsidies provided over the last 3 years on
average decrease imports by 9.6%. This may imply governmental protectionism
policies when giving a subsidy.

Frequent and usual informal gifts and payments to deal with customs increase
import correspondingly by 5 and 11 percent, being both significant at 99%
level. Positive values can be explained by customs corruption. Informal
payments facilitate import by loyalty of customs’ officers. Furthermore, the
more customs’ procedures are the obstacle – the more payments are required,
leading to increasing officers’ loyalty and, after all, simplifying customs
procedures.

Company size increase leads to the fall in inputs import from 8 to 13 percent
for medium (22-99 employees) and large (more than 99 employees) companies

23
respectively with results significant at 99% level. The underlying explanation
proposed is that large companies are more vertically integrated and have their
own production lines to cover some of own demand in used inputs, having less
need in imported goods. Moreover, having own inputs production line
provides just-in-time production both to avoid extra storage costs and to assure
certainty of supply.

Three out of five years (2005, 2008, 2012) have significant results, decreasing
import comparing to 2002. Hence, such a radical change in significance for
2012 in interpolated sample is confirmed by the imputed one. Coefficients on
both 2008 and 2012 years remained the same – decreasing the inputs import by
about -12.5% comparing to 2002. However, in 2005, the import level was less
than the level of 2002 on average by 1.9% - the difference is not as big as in the
after-crisis years.

Constant term still captures almost two thirds of the variation in import. In
order to improve model fit the underlying survey needs to be extended by
additional questions concerning import (for example, cost to import, customs
and other import duties, whether imported goods have a domestic substitute,
etc.). Detailed statistics on imputed OLS results as well as comparison of all the
obtained results can be found in Table 6 below.

24
Table 6. Comparison of all results
Variable OLS Interpolated Imputed
CustDays -0.1030** -0.1378*** -0.1277***
ImDirect 5.7125*** 6.3307*** 6.4916***
TranspObs
Minor -2.5366** -1.9802* -2.2194**
Moderate -2.6436* -2.4521* -2.7196**
Major 1.1880 -0.3223 -0.5361
Very severe -5.8135 -1.4096 -1.8100
CustObs
Minor 0.9941 0.5332 0.2938
Moderate 2.6621* 1.0024 0.5012
Major 5.4730*** 3.8056*** 3.5391***
Very severe 5.8667 1.6614 0.05604
Foreign 0.1324*** 0.1456*** 0.1480***
GovSub -9.7180*** -9.9676*** -9.6474***
InformGift
Never -1.7794 -1.1457 -0.4029
Seldom 0.0123 -0.7218 1.7047
Sometimes 0.1620 1.0520 1.2670
Frequently 5.6522*** 6.2796*** 4.9810***
Usually 10.7114*** 12.3435*** 10.5246***
Size
Med (20-99 empl) -8.8708*** -8.3698*** -7.9423***
Large (>99 empl) -14.6909*** -13.1631*** -12.8149***
Year

2005 0.8701 -1.4169 -1.8614*

2007 0.5704 1.0796
2008 -12.3047*** -12.5362*** -12.8255***
2009 -1.0208 -1.9273 -2.1046
2012 -12.5660*** -12.3431***
const 65.7988*** 65.2874*** 64.9717***
legend: * p<0.05; ** p<0.01; *** p<0.001

25
4.4 Case of Ukraine

To investigate the effect of customs delays on importing decision of the

Ukrainian firms we use the imputed sample of 250 companies (compared to
benchmark sample of 145 observations without imputation). The results are
close, but with some differences.

Every 10-day delay clearing the customs in Ukraine decreases the quantity of
imports by 4.1%. Direct import is no longer a significant factor for imports
share.

Only very severe transportation obstacles matter, decreasing imports by 46%

as well as government subsidies that lower import by 32%. Apparently,
Ukrainian government has implemented more severe protectionism policy than
average across the whole sample. Firm size still matters – it decreases import
by 14-17% according to the firm size (medium and large firms respectively
comparing to small firms).

Full-foreign ownership increases import by 22%. While frequent informal gifts

and payments decrease imports by 12% being significant at 90% level.

4.5 Tests

Breusch-Pagan test showed heteroscedasticity given the presence of a

significant outlier in the sample (a large company from the Former Yugoslav
Republic of Macedonia electronic industry in 2009 that imported 90% of its
inputs and it took 300 days to clear the customs using direct import). Hence, I
have excluded it from the sample and reestimated the regression as robustness
check. However, all the coefficients proved to be stable under robustness check
in both benchmark OLS and imputed OLS regressions with an effect on
dependent variable within 0.5%. Moreover, we should emphasize that
coefficient on days clearing the customs has increased in its absolute value to

26
0.22; hence, decreasing the imported inputs by 2.2% when talking about 10-
days customs delay.

Ramsey test on omitted variables suggests no evidence of functional form

misspecification. See the results in Table 7.

Table 7. Tests results

Statistics Benchmark OLS Interpretation

Breusch-Pagan test for chi2(1)=9.57 We do reject H0, so the

heteroskedasticity. P>chi2=0.002 residuals are
H0: Constant variance heteroskedastic

Ramsey RESET test for F(3,4470)=1.42 We cannot reject H0, so

omitted variables. P>F=0.234 there are no omitted
H0: no omitted variables variables

4.6 Policy implications

My results suggest that delays in customs clearance can decrease the share of
imported goods. If we believe that this limits firms’ choice in inputs and
production decisions, the government should undertake policies that would
reduce delays. First, corruption at customs’ can be dealt with, but according to
the obtained results, it is unclear whether entities will gain from this.

However, if government is primarily concerned in protectionism it could not

be in its interest to deal with customs delays which eventually favor domestic
producers of the comparable inputs.

Instead of longer customs procedures, some additional tariff may be negotiated

with the World Trade Organization (WTO), a member of which is Ukraine.
This is expected to have a similar effect on import, but the government will
gain from additional cash flows, or additional working places may be created in

27
order to provide less time clearing the customs, creating a win-win situation.
This can be compared to quota and a tariff, where customs-clearance time is a
quota restriction. Tariff is always preferred to quota and their equivalence can
be found. The size of the tariff cannot be calculated within my work due to
insufficiency of data; hence, it may well be the topic for future research.

28
 Chapter 5

CONCLUSIONS

After all the analysis, quite unexpected results were obtained, nonetheless,
conjectural explanation was proposed. The variable of the main interest –
days spent clearing the customs, appeared to be statistically significant, still
having low marginal effect. Hence, every additional 10 days spent on customs
will lower the probability of import by 1.6%. This and all further listed results
are presented from the robust regressions on imputed sample.

Import is decreased on average by 3% in case of moderate transportation

obstacles occurrence. While customs obstacles counterintuitively increase
import. However, such obstacles are treated by informal gifts or payments
that, in its turn, can even facilitate more trade.

Smaller firms tend to import more than larger ones as well as firms owned by
foreign individuals. Furthermore, 100% foreign-owned firm would import
more by 15%, ceteris paribus. While government subsidies lower import (by
10%) in order to protect domestic producers.

In the case of Ukraine, every 10-day delay clearing the customs decreases the
quantity of import by 4.1% with no significant effect from direct import. Only
very slight and very severe transportation obstacles matter, decreasing import
by 7% and 47% respectively as well as government subsidies that lower import
by 32%. Firm size decreases import by 14-17% according to firm size, while
full-foreign ownership increases import by 22%. Regression model on
Ukrainian firms has data only from 2002, 2005 and 2008 with no statistically
significant differences from 2002 base year.

After all, it may be proposed to governments to implement some policies to

reduce customs-clearance time, for example substitute long customs
procedures by some additional customs duty. Surely, this is have to be

29
negotiated with WTO first. This is expected to have similar effect on import,
but the government will gain from additional cash flows or from creation of
additional working places in order to reduce time clearing the customs.
However, if the government adheres to a protectionism policy, it may not be
interested in facilitating import. The size of such a customs duty or tariff cannot
be calculated within my work due to insufficiency of data; hence, it may be the
topic for future research.

Corruption at customs’ may also be somehow dealt with, but the effect on
entities is unclear under given results. Anyway, government concerned about
protectionism may win from eliminating corruption twofold.

30
 WORKS CITED

Amiti, Mary, and Jozef Konings. 2005. Trade Liberalization, Intermediate

Inputs, and Productivity: Evidence from Indonesia. IMF Working Paper.

Amiti, Mary, and Davis R. Donald. 2012. Trade, Firms, and Wages: Theory and
Evidence. Review of Economic Studies 79(1): 1-36. Oxford University Press.

Bourgeon, Pauline, Jean-Charles Bricongne, and Gaulier Guillaume. 2012

Financing Time to Trade: Evidence from French firms, version 1.

Djankov, Simeon, Caroline Freund, and Cong S. Pham. 2010. Trading on Time.
The Review of Economics and Statistics 92(1): 166-173. MIT Press.

Görg, Holger, and Aoife Hanley 2004. Does Outsourcing Increase Profitability?
The Economic and Social Review 35(3): 267-288.

Görg, Holger, and Aoife Hanley 2005. International Outsourcing and Productivity:
Evidence from the Irish Electronics Industry. North American Journal of
Economics and Finance 16(2): 255- 69.

Görg, Holger, Aoife Hanley, and Eric Strobl. 2007. Productivity Effects of
International Outsourcing: Evidence from Plant Level Data. Canadian Journal of
Economics Forthcoming.

Grossman, Gene M., and Elhanan Helpman. 2005. Outsourcing in a Global

Economy. Review of Economic Studies 72: 135-159.

Halpern, Laszlo, Miklos Koren, and Adam Szeidl. 2005. Import and Productivity.
IEHAS Discussion Papers 0509. Institute of Economics, Hungarian Academy
of Sciences.

Halpern, Laszlo, Miklos Koren, and Adam Szeidl. 2011. Imported Inputs and
Productivity. CeFiG Working Papers 8. Center for Firms in the Global
Economy.

Harrigan, James, and Anthony J. Venables. 2004. Timeliness, Trade and

Agglomeration.

Hummels, David, Jun Ishii, and Kei-Mu Yi. 2001. The nature and growth of
vertical specialization in world trade. Journal of International Economics 54: 75-96.

Hummels, David. 2007. Transportation Costs and International Trade in the

Second Era of Globalization. Journal of Economic Perspectives 21: 131–154.

Hummels, David, and Georg Schaur. 2012. Time as a Trade Barrier. NBER
Working Paper No. 17758.

31
Martinez-Zarzoso, Inmaculada, and Felicitas Nowak-Lehmann D. 2007. Is distance
a good proxy for transport costs? The case of competing transport modes.
Journal of International Trade & Economic Development, Taylor and Francis Journals
16(3): 411-434.

Nordås, Hildegunn Kyvik, Enrico Pinali, and Massimo Geloso Grosso. 2006.
Logistics and Time as a Trade Barrier. OECD Trade Policy Papers No. 35.
OECD Publishing.

Rubin, Donald.B. 1987. Multiple Imputation for Nonresponse in Surveys. J. Wiley & Sons,
New York.

Schafer, Joseph L. 1997. Analysis of Incomplete Multivariate Data. Chapman & Hall,
London.

32