Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Essays in international economics
(USC Thesis Other)
Essays in international economics
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Essays in International Economics
by
Bilal Muhammad Khan
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Economics)
August 2017
Copyright 2017 Bilal Muhammad Khan
Epigraph
\In God we trust, all others bring data."
Dr. W. Edwards Deming (19001993)
\Where there is ruin, there is hope for a treasure."
Jalal ad-Din M. Rumi (12071273 )
ii
Dedication
To my family, for their consistent support and love,
to my teachers, for inspiring me to explore,
and to mankind, for making the world an interesting place.
iii
Acknowledgements
It would not have been possible to complete this dissertation without the guidance of my committee
members.
Foremost, I am extremely grateful to Ayse Imrohoroglu and Joshua Aizenman for their invalu-
able guidance, endless patience, immense knowledge and unconditional support during my years at
the University of Southern California (USC).
I am also highly indebted to Joel David, Jerey Nugent and Selale Tuzel, who served in my
qualifying committee, for their insightful suggestions. I would also like to express my sincere
gratitude to Caroline Betts, Kensuke Teshima, Rahul Giri and Rubina Verma for their helpful
comments.
This dissertation beneted greatly from discussions with colleagues and seminar participants at
USC as well as conference participants at WEAI, RCEF and Midwest International Trade Meeting.
I sincerely appreciate support from my colleague and co-author Junjie Xia.
I am thankful to the economics department sta Young Miller, Fatima Perez and Morgan
Ponder for their various administrative support throughout my years at USC. I am also grateful to
fellowships and grants from the Department of Economics, the USC Graduate School and Dornsife
College of Letters, Arts and Science at USC.
iv
Table of Contents
Epigraph ii
Dedication iii
Acknowledgements iv
List Of Tables vi
List Of Figures viii
Abstract ix
Chapter 1: Export Destination, Skill Utilization and Skill Premium in Chinese
Manufacturing Sector 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Relevant Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Income Inequality and Trade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Chapter 2: Productivity Dierences Between and Within Firms-Case of Taiwan 32
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Multi-plant Firms and Labor Productivity . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Exporter Premium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Reference List 48
Appendix A
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
v
List Of Tables
1.1 Summary Statistics for Chinese Manufacturing Firms in 2004 . . . . . . . . . . . . . 25
1.2 Relationship between Price of the product and Export Destination . . . . . . . . . . 26
1.3 Relationship between proportion of skilled workers and Export destination . . . . . . 27
1.4 Relationship between Average wage and Export destination using broader measure of Skill 28
1.5 Relationship between Average wage and Export destination using detailed Skill level 29
1.6 Quantile Regression between Average Wage and Export Destination . . . . . . . . . 30
2.1 Summary Statistics of Key Variables at Plant and Firm Level . . . . . . . . . . . . . 43
2.2 No. of Multi-plant Firms and Its Share in Employment and Sales: 1998-2005 . . . . 43
2.3 Number of Multi-industry Firms and Its Shares in Employment and Sales Within
Multi-plant Firms: 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Importance of Most Important Industry for Multi-industry-Multi-Plant Firms: 1998-
2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Importance of Top Plant for Multi-Plant Firms (Aggregate): 1998-2005 . . . . . . . 44
2.6 Importance of Top Plant for Multi-Plant Firms (Single Industry): 1998-2005 . . . . 44
2.7 Importance of Top Plant for Multi-Plant Firms (Multi Industry): 1998-2005 . . . . . 45
2.8 Heterogeneity between the most labor productive plants and other plants within rms 45
2.9 Labor Productivity (1) across and within industry and (2) across and within rm . . 45
2.10 Heterogeneity between bigger plants and other plants within rms . . . . . . . . . . 46
vi
2.11 Exporter Premium with both Firm Exporter Dummy and Plant Exporter Dummy.
2000-2002 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.12 Exporter Premium with both Firm Exporter Dummy and Plant Exporter Dummy.
2000-2002 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
A.1 No. of Firms Switching Plant-Type (Single to Multi & Vice versa): 1998-2005 . . . . 55
A.2 No. of Firms Switching Industry-Type (Single to Multi & Vice versa): 1998-2005 . . 55
A.3 No. of Firms Switching Industry-Type (Single to Multi & Vice versa): 1998-2005 . . 55
A.4 No. of Multi-plant-Multi-industry Firms Entering or Exiting: 1998-2005 . . . . . . . 56
A.5 No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
A.6 No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
A.7 No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.8 No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.9 No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
vii
List Of Figures
1.1 Relationship between average wage and the skill premium at dierent deciles of
average wage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
viii
Abstract
This dissertation explores the implication of rm's interaction in the international market and its
impact on optimal decisions by rm.
The rst Chapter, titled \Export Destination, Skill Utilization and Skill Premium in Chinese
Manufacturing sector", joint work with Junjie Xia, analyzes the link between export destination,
skill utilization and skill premium. We propose a mechanism behind these links: the dierence
in quality valuation of the product across exporting destinations and the distribution of level of
skill among the skilled workers in the labor market. We test this theory using cross-section of
more than 160,000 single product Chinese Manufacturing rms survey data, of which nearly 22,000
are exporting to more than 200 countries across the world. Contrary to the earlier literature, we
explicitly use education as a measure of skill. We nd that rms exporting to high income countries
tend to charge a higher price, pay higher average wages to employees, hire more skilled workers
as dened by education level, and pay higher skill premium as compared to rms exporting to
middle or low income countries or selling domestically. As in similar recent studies, we did not nd
exporting per se to signicantly impact the proportion of skilled workers or the skill premium in
the rm.
The second Chapter \Productivity Dierences Between and Within Firms-Case of Taiwan",
joint work with Kensuke Teshima and Rahul Giri, highlights the existence of heterogeneity in
ix
economic characteristics of plants belonging to the same rm and producing the same product.
Using the plant level data for Taiwanese manufacturing rms data from 1998 to 2005, we found
some interesting results for multi-plant rms and the exporter premium among these multi-plant
rms.
x
Chapter 1
Export Destination, Skill Utilization and Skill Premium in
Chinese Manufacturing Sector
1
1.1 Introduction
Much of the traditional literature in International Trade has been focused on the exporting behavior
of the rm. Yet there has not been much of consensus in the literature on whether more productive
rms self-select into exporting activity (Bernard and Jensen 1995; Clerides, Lach, and Tybout 1998)
or on whether the exporting activity helps to improve the productivity of the rms by using more
ecient technologies for production or hire more skilled workers (Bustos 2011; Matsuyama 2007).
Some of the more recent literature, however, e.g. Bastos and Silva (2010), Verhoogen (2008) and
Manova & Zhang (2012) suggest that characteristics of the exporting destination country such as
income, distance, transportation costs etc. might contribute to the determination of rm behavior
and choice of production techniques. Serti and Tomasi (2008) have analyzed the role of international
1
Joint work with Junjie Xia
1
trade in explaining the intra-industry heterogeneity in the Italian manufacturing rms involved in
exporting or importing activities. They have shown that the rms that involve both importing
and exporting activities outperform the rms that involve only one of these activities in terms of
productivity, size, capital intensity and skill intensity. Brambila et al. (2012) and Brambila & Porto
(2016) have established a causal link between the export destination and the proportion of skilled
workers in the rm whereas Frazer (2013) has established a similar link between the importing
source and skill utilization at the rm level.
In this paper, we elaborate upon the theoretical literature on the role of export destinations and
skill utilization and also empirically establish a link between the export destination and the skill
premium being awarded at the rm level using Chinese manufacturing rms data. Building on the
work of Brambila et al. (2012), we explore whether the provision of higher quality goods requires
only more intensive use of skilled workers or it also requires higher quality of skilled workers. To
test our hypothesis, we use a cross-section of annual Chinese manufacturing rms data for 2004. It
has detailed information on the individual rms including sales, exports, number of workers, wage
bill, capital ownership etc. We match that data with the monthly customs data with information on
every export or import transaction by these rms. We also use the rm level data on the distribution
of workers by education level for these manufacturing rms. After carefully matching these data-
sets, we nd a statistically signicant relationship between export destination, skill utilization and
skill premium at the rm level. The availability of the data on the distribution of education among
the workers helps us to cleanly dene the measure for skills in the rm and to analyze its relation
to export destination.
2
The remainder of the paper is organized as follows. The next section will discuss the recent literature
in international trade that is relevant to our paper. Section 3 will discuss the basic economic
intuition for this paper and develops a model to describe a link between export destination and
the skill premium in the rm setting. In section 4, we would discuss the data and section 5 would
outline the empirical strategy employed to test the results of the model.
1.2 Relevant Literature
Using cross-sectional data for 60 countries, Hallak (2006) showed that the consumers in the richer
countries demand higher quality products. Building on this idea introduced by Hallak (2006),
Verhoogen (2008) developed a model linking trade and wage inequality in developing countries. In
that model with heterogeneous plants and quality dierentiation, more productive plants were able
to produce better quality products and to export them to richer countries. In order to produce
higher quality products, rms were deemed able to attract better quality labor by paying them
higher wages. Using panel data for manufacturing plants in Mexico, he showed that this quality
upgrading leads to higher wage inequality in exporting rms and non-exporting rms within the
industry. Subsequently, however, Matsuyama (2007) found another channel for the eect of export
activity on the proportion of skilled workers in the rm, namely a "skill biased globalization" in
which exporting requires tasks that are more skill intensive in nature. These tasks require workers
who are more familiar with the international business practices and can communicate with the
foreign customers in their language and be careful with respect to the intricacies of the foreign
cultures. All these activities require proportionally more skilled workers than the rm that is
selling domestically.
3
Building on Verhoogen (2008) and Matsuyama (2007), Brambilla et al. (2012) explored the link
between the export destination of the rms and the proportion of skilled workers hired by the rm.
Their intuition is that consumers in higher income countries demand higher quality products as
they value high quality products more than the consumers in low income countries (whose marginal
valuation on income is relatively low). To produce higher quality products, rm need to hire skilled
workers in higher proportion. Using panel data for manufacturing rms in Argentina, they have
found that the exporting to higher income countries matter, but exporting per se does not. Firms
that tend to export more to high income countries use more skills and as a result pay higher
average wages compared to the rms that export to middle-income or low-income countries. They
have used average wage per worker and the proportion of non-production workers as a measure of
skill intensity for the rms.
In still another paper, Brambilla & Porto (2016) have established a link between the income level
of the export destination and the level of average wages in the exporting country across the world.
They have found robust evidence, worldwide, that the industries exporting their product to high
income destination pay higher wages to their workers. Using an instrumental variable approach,
they have shown a causal link for this phenomenon. They have shown that the consumers in
high income destination demand higher quality products and that the provision of higher quality
is costly and requires more intensive use of higher waged skilled labor. Hence, the production of
higher quality products at the industry level creates a wage premium. They have used the data for
82 countries from 1990-2000. They have dis-aggregated the data into 28 manufacturing sectors.
Similar to Brambila et al. (2012), Frazer(2013) explored the link between imports, import destina-
tion and the skill utilization for rms in Rwanda. He found that the importers, in general, and in
4
particular, the ones importing materials from richer countries, pay higher wages (and consequen-
tially, utilize more skills). Nevertheless, all of the above mentioned papers relating to skill utilization
used only crude proxies for such skill. For example, Brambila et al. (2012) and Frazer (2013) used
above average wages or the share of non-production workers proxies for higher utilization of skills
in the rm. In contrast to these papers, our paper explicitly uses education as a measure for level
of skill.
1.3 Theory
In this section, we develop partial equilibrium model analyzing the link between export destination
and the level of the skill premium. On the demand side, we assume that the products are dier-
entiated horizontally as well as vertically. The preferences are non-homothetic in order to capture
the idea that consumers in high income countries value high quality goods more than consumers
in low income countries. For simplicity, we assume a representative consumer for each country and
adopt the multi-nomial logit model as introduced by Verhoogen (2008). Customers in high income
countries have lower marginal utility of income and are willing to pay higher prices for the same
quality good as compared to their counterparts in low income countries. consumer i in country c
has the following utility in consuming product j of quality , a price p and a random deviation
following type-I extreme value distribution,
c
ij
as
U
c
ij
=
c
j
c
p
c
ij
+
c
ij
Using these assumptions, we obtain the following demand function for the product j
x
c
j
(p
c
j
;
c
j
) =
M
c
W
c
e
(
c
j
c
p
c
j
)
5
where M
c
is the number of consumers in country c and W
c
is an index that summarizes the
characteristics of all products available in country c ( i.e. W
c
=
zZ
ce
(
c
z
c
p
c
z
)
) where Z
c
denes
the set of available products.
Implicitly, we notice that the
e
(
c
j
c
p
c
j
)
W
c
is the probability of choosing the product j of quality by
a representative consumer in countryc and we multiply it with the number of consumers in country
c to nd out the expected demand for rm j product.
c
measures the marginal utility of income,
and as per Verhoogen (2008),
1
c
measures the quality valuation in countryc.
c
will determine the
relationship between and p in the consumer's utility function. The higher the level of per capita
income in country c, the lower will be the marginal valuation of income and hence, consumers will
be willing to pay more for the same quality product as compared to the consumers in lower income
countries.
On the supply side, there areJ monopolistically competitive rms in the source country. Each rm
produces a dierentiated product and can export it to multiple destinations or sell it domestically.
The rm can also choose a dierent quality of it's product for dierent exporting destinations,
based on the quality valuation in destination countries. We assume that output of the product can
be produced by using labor only. The rm j, in the source country, can produce the variety of the
quality for it's product as follows:
j
= (
bj
aj +bj
)(
A
1+e
s
j
)
Here a
j
and b
j
are the number of unskilled and skilled workers in each rm j respectively. In the
above production of quality, notice that in order to produce higher quality, the rm needs, not only,
more skilled workers but also, skilled workers of higher quality i.e. the workers are heterogeneous
6
in their level of skill as introduced by Yeaple (2005). We assume that all unskilled workers in the
economy are identical and receives a wage of $1 whereass
j
is the level of skill of the skilled workers
over the entire range (1;1). For current purposes, we can assume the level of skill is uniformly
distributed among the skilled workers. The higher is the value of s
j
, higher is the quality of the
skilled worker. Fors(1;1); takes the value between 0 andA whereA is the maximum quality
of the product.
Now, since the skilled workers vary by quality, we also dene the compensation of the skilled workers
as function ofs
j
. The skilled worker's wage of quality s
j
is given by (1 +
K
1+e
s
j
) i.e.
K
1+e
s
j
is the
premium that the skilled worker of quality s
j
gets over and above the unskilled worker.
Given the above assumptions, a monopolistically competitive rm, producing product j of quality
and selling it to consumers in country c at a price p, would maximize the prot as:
Max
p
c
j
;sj;bj
c
j
= [p
c
j
a
j
b
j
(1 +
K
1+e
s
j
)]x
c
j
(p
c
j
;
c
j
)F
c
or
Max
p
c
j
;s;bj
c
j
= [p
c
j
a
j
b
j
(1 +
K
1+e
s
j
)]e
(
c
j
c
p
c
j
)M
c
W
c
F
c
For the rmj in the source country,M
c
andW
c
will be exogenous. F
c
is the xed cost of exporting
to country c. It can be thought of as transportation costs, or other regulatory costs involved in
exporting to country c.
The rst order conditions for the above maximization problem would be:
7
p
c
j
: e
(
c
j
c
p
c
j
)
(1
c
p
c
j
+
c
a
j
+
c
b
j
(1 +
K
1 +e
sj
)) = 0 (1.1)
b
j
:
e
(
c
j
c
p
c
j
)
1 +e
sj
[
Aa
j
p
c
j
(a
j
+b
j
)
2
Aa
2
j
(a
j
+b
j
)
2
(1 +K +e
sj
)
Aa
j
b
j
(1 +K +e
sj
)
(a
j
+b
j
)
2
)(1 +e
sj
)
] = 0 (1.2)
s
j
:
e
(
c
j
c
p
c
j
)
e
sj
(1 +e
sj
)
2
[
p
c
j
Ab
j
(a
j
+b
j
)
Aa
j
b
j
(a
j
+b
j
)
b
j
K
Ab
2
j
(1 +K +e
sj
)
(a
j
+b
j
)(1 +e
sj
)
] = 0 (1.3)
Now, from FOC of the price, we would have
p
c
j
=a
j
+b
j
(
1 +K +e
sj
1 +e
sj
) +
1
c
(1.4)
First we put this equation of price into the FOC for s
j
, we would get
[(a
j
+b
j
(
1 +K +e
sj
1 +e
sj
) +
1
c
)
Ab
j
(a
j
+b
j
)
Aa
j
b
j
(a
j
+b
j
)
b
j
K
Ab
2
j
(1 +K +e
sj
)
(a
j
+b
j
)(1 +e
sj
)
] = 0
(1.5)
8
After simplifying, we obtain
b
j
=
A
c
K
a
j
Propostion 1: Firms, that export more to high income countries, will hire more skilled (educated)
workers in comparison to the rms that export to middle or low income countries.
Similarly, now we put the value of p
c
j
and above found b
j
into the FOC for b
j
, we would have
[(a
j
+b
j
(
1 +K +e
sj
1 +e
sj
) +
1
c
)
Aa
j
(a
j
+b
j
)
2
Aa
2
j
(a
j
+b
j
)
2
(1 +K +e
sj
)
Aa
j
b
j
(1 +K +e
sj
)
(a
j
+b
j
)
2
)(1 +e
sj
)
] = 0
(1.6)
After simplifying, we obtain
1 +K +e
sj
=
aj
c
K
2
A
Propostion 2: Firms, that export more to high income countries, will hire better quality skilled
(educated) workers in comparison to the rms that export to low or middle income countries.
9
From the above, it is clear that
c
is inversely related with the level of skills
j
of the skilled worker.
Now replacing 1 +e
sj
and b
j
into the rst order condition for the price, we would get
p
c
j
=a
j
+ (
A
c
K
a
j
)(1 +
K
(
a
j
c
K
2
A
K)
) +
1
c
Propostion 3: Firms, that export more to high income countries, will charge higher price to
customers in high income countries as compared to the customers in low or middle income countries.
From the above, one can see that the rms that are exporting to high income countries will be
charging a higher price within the same product category as the consumers in high income countries
would demand higher quality of the product. In addition, our model also shows that rms exporting
larger proportions of exports to high income countries would proportionally hire more skilled workers
and also ones of better quality. Given the wage schedule dened above, it would lead to higher
average wage in the rm.
1.4 Data
For this paper we use balance sheet data for Chinese Manufacturing rms from the Annual Survey of
Industrial Firms (ASIF) conducted by China's National Bureau of Statistics(NBS) for 2004. ASIF
cover all rms with sales above 5 million RMB
4
during the survey year. There are multi-product as
well as single product rms in this data. The survey reports information on the rm name, rm id,
total sales, total capital, total xed assets, total employment, total wage bill, total exports, number
of computers, rm subsidy, rm ownership etc. In addition to ASIF, for trade-related data, we
use the comprehensive data-set provided by the General Administration of the Chinese Customs,
4
In 2004, 1 USD was equivalent to 8.27 RMB
10
known as Chinese Customs Trade Statistics (CCTS) for 2004. CCTS reports the rm name, rm
id, value of all rm-level exports and imports and export or import destination at the monthly
level. We rst aggregate the CCTS to the annual level and then match it with ASIF for 2004.
In addition, for 2004, ASIF also reports data about the distribution of number of employees by
level of education. For example, in rm x, we know how many workers are with middle school,
high school, technical diplomas, bachelors and postgraduate degrees. Contrary to Brambilla et al.
(2012), we use these education levels to split the workers into skilled and unskilled workers. We
have dened the workers to have less than high school education to be unskilled workers and those
with more than or equal to high school be considered as skilled.
We matched these 3 data-sets using rm ids and rm names. ASIF also reports the top 3 products
of the manufacturing rms. We drop all the rms that have more than one major product as in case
of multi-product rms, we cannot identify how much of the labor is assigned to what product in the
rm. After dropping multi-product rms from the data, we are still left with nearly 22,000 exporting
rms and nearly 140,000 non-exporting rms. These rms are distributed over 524 industries at 4
digit level and across 399 cities in China. Among the left-over rms, 471 industries and 311 cities
have atleast one exporting rm.
We split the countries into high income, middle income and low income countries using 2004 income
per capita range as listed by World Bank. Countries having per capita income to be higher than
$9000 in 2004 are considered to be the high income countries.
11
1.5 Empirical Results
We start by reporting the summary statistics for the Chinese manufacturing sector in 2004 in Table
1.1. We notice that, on average, exporting rms are larger by both sales and employment, pay
higher average wages and are more capital intensive as compared to the non-exporting rms. As
the data is about the manufacturing rms in China, we notice that more than half of the workers
are unskilled as dened by level of education. For the exporting rms, nearly two-thirds of the
exports are directed towards the high income countries. In addition, nearly a quarter of the rms
are pure exporters i.e. with their entire output sold abroad. However, we notice that in contrast
to some of the earlier literature referred to above, in aggregate data, there is not much dierence
between exporting and non-exporting rms in terms of proportion of workers with high school or
above education.
In order to perform the empirical analysis, we start by trying to ascertain the validity of our
hypothesis that rms export the better quality products to higher income countries. Since all the
rms in this analysis are single product rms, we will use the price charged as a proxy for the
quality of the product. We use the CCTS data for 2004. As noted above, CCTS reports the price
and quantity for the exported (or imported) goods as well as the export (or import) destination for
the exporting rms at monthly level. We convert this monthly data to annual data by calculating
the average price being charged by a rm for each destination.
Log(Price)ij =0 +1HI Dummyi +2MI Dummyi +3log(Quantity)ij +4Distancei +j +ij
We start in column 1 of Table 1.2 by regressing logarithm of price against a dummy for High Income
countries as well as Middle Income countries. Here, Log(Price)
ij
is the logarithm of average price
12
charged by rm j to customers in Country i.
j
is the dummy for rm j to control for rm xed
eects. We notice that the rm charges, on average, 5.4 percent more for a product exported to high
income countries than to one exported to low income country in 2004 and charge 1.95 percent more
for a product exported to middle income countries than to low income country. Once we add the
logarithm of quantity exported, as in column 2 of this table, the price premium falls down to about
2.8 percent for high income country and 1 percent for middle income country. However, coecient
for the Middle Income country dummy is no longer signicant when we add the Log(Quantity)
ij
to the analysis as in column 2. We also notice from column (2), a negative relationship between
price and quantity. On average, if the rm exports 1 percent more quantity, price of that export
goes down by about 0.25 percent. While there could be several explanations for this, it could be
because of rms oering bulk discount to the consumers in exporting destination. For the above two
regressions, we are using Firm xed eects. The nding that the higher the income of the country,
higher is the price being charged by the rm for the same category of the product is in line with the
earlier literature. Manova and Zhang (2012) have shown that rms vary the quality of their product
across destinations by varying the quality of the input used in the production process. They have
shown that the rms, within the same product line, charge varying prices for their exports by export
destination. Across destinations, within a rm-product, rms set higher prices in richer countries.
Grg et al.(2010) have analyzed the relationship between gravity variables and f.o.b. export unit
values using Hungarian rm-product destination data. They also found a positive and signicant
relationship between the export unit values and the GDP per capita of the destination. Manova
and Zhang (2012) have shown that rms charge higher prices in export destinations which are
bilaterally more distant countries. It might be possible that the rm is charging a higher price to
high income countries just because they are more distant than the middle or low income countries.
13
In order to nd distance between China and exporting destination, we use the database for bilateral
distances between countries prepared by Mayer and Zignago (2011). In column (3) of Table 1.2,
we add the distance in kilometers as a control variable. We also found positive and signicant
relationship between bilateral distance and price charged by the rm. Nevertheless, even after
controlling for distance, the coecient for high income country dummy still remains positive and
signicant. Another possibility for rm charging varying prices for same product across export
destinations could be the variation in the import taris across destination. However, Baumellassa
et al. (2009) have shown that the average tari for manufactured products was much less in 2004
for high income countries as compared to middle of low income countries. It would imply that for
the same quality products, rm should charge higher prices to customers in middle and low income
countries as compared to the customers in high income countries.
Now that we have established that the rms export higher quality products to the High Income
countries using price as a proxy for quality, we can proceed to examining the impact of export
destination on the labor hired by the rms. From the model, it was seen that the rms, exporting
higher proportion of their exports to high income countries, would be expected to hire proportionally
more skilled(or educated) workers. To test this prediction, Table 1.3 presents results obtained from
a regression similar to that in Brambilla et al. (2012). The estimating equation we have, with
skilled workers dened as ones with high school or more education, is
SKILL
jkl
=
0
+
1
HI
jkl
+
2
EXP
jkl
+X
0
jkl
:
3
+
kl
+
jkl
where SKILL
jkl
is dened as proportion of skilled workers in rm j operating in 4 digit industry
code k and city l, HI
jkl
is dened as the proportion of exports that goes to the High Income
countries, EXP
jkl
is ratio of exports to sales. X
jkl
is a vector of rm specic variables that can
14
possibly aect the proportion of skilled workers. In addition,
kl
is dummy to control for the
industry-city xed eects. That creates more than 30,000 cells for industry-city xed eect. We
also cluster the errors at 4 digit industry-city level.
From column (1) of Table 1.3, it can be seen that exports to high income country positively and
signicantly aects the proportion of skilled workers. However, contrary to the existing literature,
the coecient for export intensity is negative and signicant. Nevertheless, using the same rm
level data for 2004, Zhang (2010) has found that the similar results for the relationship between
export intensity and the proportion of skilled workers in the rm i.e. exporters tend to hire more
un-skilled workers as compared to the non-exporters among the Chinese manufacturers. We will
see later, that once we control for pure exporter dummy, this eect does not remain signicant any
more. Zhang (2010) did not control for the pure exporter dummy.
In order to check the robustness of our results, in column (2) and (3) we add more variables that
can aect the employment of skilled workers in the rm. In column (2), we add age of rm, age
squared, capital per worker and a dummy for State owned rm. In column (3), in addition to the
above,X would include pure exporter (PE) dummy, processing and assembly trade (PAT) dummy
and feed processing trade (FPT) dummy. We assume that the rms that are existing for longer
period of time have better understanding of the labor market as well as the market for it's product.
Though, apriori, we do not expect any sign for the coecient of age. In addition, we are controlling
for Capital intensity of the rm (Cap) dened as the logarithm of capital per worker in the rm.
Our expectation is that the more capital intensive is the rm, rm would hire more skilled workers
to operate the expensive machines and the results in column (2) and (3) supports this. We nd the
similar results as above i.e. positive and signicant eect of high income exporting destination and
15
negative and signicant eect of exporting activity on proportion of skilled workers. In addition,
we nd negative relationship between the age of the rm and proportion of skilled workers in the
rm, though, the coecient is small.
In order to address this negative impact of exporting behavior on the proportion of skilled workers,
we need to be careful in interpreting these results for China. Dai et al(2011), have shown that the
rms which are involved in the processing trade in China are also comparatively less productive.
Processing trade rms import nearly all of the inputs from abroad, assemble it in Chinese Export
Processing zones and then ship all of their output to the rest of the world. In order to distinguish
these types of rms, we will use two types of processing trade dummies i.e. PAT controls for
whether the rm is involved in processing and assembly trade and FPT for Feed and Processing
trade. In addition, we will also use a pure exporting rm dummy (PE) for the rms which ship
all of their output abroad and do not sell anything in China. The intuition is that the processing
trade rms will be exporting all of their output to the foreign rms for which they are processing
the exports. Since usually these processing trade rms are established by foreign rms to take
advantage of cheap labor in China. In addition, building on Melitzs (2003) framework, Lu et al.
(2012) identied the condition for the existence of pure exporters and showed their productivity
levels to be above the productivity of the non-exporters but less than the productivity of the rms
which sell both in the domestic and the foreign market. It might also be that the rms with high
export to sales ratios, might make less use of skilled workers in the rm. Once we account for the
pure exporter eect, we notice that the coecient of the export intensity hover around zero and
is no longer signicant. Note that Brambila et al.(2012), also did not nd a signicant eect of
exporting behavior on the proportion of skilled workers in the rm. As in Lu et al. (2014) and Dai
et al. (2011), we also nd signicant negative coecient for pure exporter dummy. Indeed, a rm
16
which is a pure exporter and not exporting to a high income country, on average, tends to hire 4.5
percent less skilled workers as compared to a rm selling all or part of it's output in the domestic
market.
Our results suggest that rms which export only to the High Income countries hire 4.5 percent
more skilled workers as compared to the rms that export to middle or low income countries or
sell only in the domestic market once we control for the Pure Exporter dummy. Brambilla et al.
(2012) found it to be about 5 percent, when they employed a simple OLS regression and used share
of non-production workers as a measure for skill for Argentinian Manufacturing rms.
Next we proceed to estimate the impact of export destination on the skill premium in the rm. We
rst run the regression using the denition of skilled workers as used above. Later, we split the
skilled workers into ve categories dened by their education level. Then, we explore the impact of
export destination on extra premium for each category of education for skilled workers.
To test our main hypothesis of the paper, we run the following regression to estimate the impact
of export destination on the skill premium in the rm.
Average wage
jkl
=0 +1SKILL
jkl
+2HI
jkl
+3HI
jkl
SKILL
jkl
+4EXP
jkl
+X
0
:5 +
jkl
+
jkl
whereAverage wage
jkl
is the logarithm of the average wage for rmj operating in 4 digit industry
k and city l. HI
jkl
SKILL
jkl
is an interaction between the proportion of skilled workers in
the rm and the proportion of the exports going to the High Income countries. X
jkl
is the same
vector of variables as above and
jkl
is the error term.
3
represents our estimate of the additional
premium being awarded to the skilled workers as a result of exporting to high income destinations.
As above, we nd a signicant negative impact of export intensity on the average wage as well.
17
From the rst column, we notice that a rm which hires only skilled workers i.e. workers with
education of high school and above, pays 27 percent more than a rm which hires only unskilled
workers. In addition, there is also a premium for exporting to high income countries. We notice
that, assuming everything else to be the same, a rm which exports all of the output to the high
income country, pays, on average, 6 percent more compared to the rm which sells the output to
middle or low income countries or sell only in the domestic market. We also notice that, once
we control for high income destination, we nd no signicant impact of exporting activity on the
average wage in the rm but do nd a positive and signicant coecient for the interaction term
HI
jkl
SKILL
jkl
. Even after controlling for the skilled (or educated) workers in the rm, we nd
that there is an extra premium for skilled workers. Notice that rms which only employ skilled
workers and export all the output to high income countries pay 26.6 percent more to the workers
compared to rms which also employ only skilled workers but do not export any output to high
income countries.
Next, in column (2) we add more controls (i.e. AGE;AGE
2
;Cap;StateOwnership) as above.
Using the data about US rms, Ouimet and Zarustkie (2014) had found a positive and signicant
relationship between the age of the rm and age of the employee. They found that young rms
disproportionately hire more young workers and these results hold even when they controlled for
industry or geographical location of the rm. Here, we use the age of the rm as a proxy for the
average age of the employees to control for the experience of the workers since there is no data
available for the age of the employees in the rm. In column (3), we would also add PE;PAT and
FPT to control for the pure exporter dummy as well as two types of processing trade rm dummy.
18
After controlling for these extra measures, the premium for skill, high income export destination
as well as the skill premium for the high income export destination falls a little but still remains
positive and signicant. We also nd a positive and signicant relationship between the age of the
rm and the average wage being awarded by the rm. These results show that our ndings are
robust to the addition of these extra controls.
Next, in Table 1.5 we proceed to dis-aggregate the skilled workers into their education groups and
analyze the impact of export destination on skill premium for each category. In this case, labeling
each dierent category of education as a dierent level of skill. HS refers to high school, Diploma
refers 2 year technical degree which is ranked higher than high school but less than 4 year college
degree, BS refers to 4 year college degree and MS refers to 2 year Masters degree.
We notice that every educational category has an extra premium over and above the unskilled
workers. It can be seen that the highest premium is for a college degree. It is positive and
signicant. Ge and Yang (2012) also found the premium for a college degree to have been the
highest and continuously rising over time. For our cross-sectional analysis, we also found the
highest premium be for the college degree in the rm. We also notice that every education category
of high school and above also gets an extra premium if the rm is exporting more to a high income
destination. Notice that the premium for high income export destination is largest for the college
degree.
In addition, we notice that the State ownership dummy has also positive and signicant aect on
proportion of skilled workers as well as average wage. It seems like the state owned rms hire
higher proportion of skilled workers and also pay higher wages on average. We also nd signicant
19
negative eect on the average wage of the pure exporter dummy. However, we found signicantly
positive coecient for labor intensity.
The above results supports our theoretical results where we found the inverse relationship between
marginal utility of income and the price charged by the rm, proportion of skilled workers as well
as skill premium being awarded by the rm.
1.6 Income Inequality and Trade
Another strand of International Trade literature has focused on the impact of trade activities on the
income inequality of the developing countries and have found con
icting evidence in this regard.
Conventional wisdom dictates that the trade liberalization activities would help the less skilled
workers in the developing economy because the developing country has abundant supply of less
skilled workers. Though after analyzing many studies related to developing countries, Goldberg
and Pavcnik (2007) could not nd support for this conventional wisdom. Rather, in many of these
studies related to developing economies, they found that the higher skilled workers benet more
than the unskilled workers as a result of trade liberalization which leads to worsening of income
inequality in these countries. Amriti and Cameron (2012) have analyzed the impact of import taris
on the Indonesian labor market. They found that the reduction in import taris leads to reduction
in skill premium for the rms that import intermediate inputs. They did not nd any signicant
eect of reducing import taris on nal goods on skill premium within rms. They suggest that
their result diers from the earlier studies as their study is the rst one to separate out the eects of
input taris and output taris whereas earlier studies used to focus on reducing nal goods taris
and changing trade shares.
20
As China has achieved spectacular growth after opening up her economy, there has been many
studies exploring the impact of opening up the economy on income inequality in China. Wei and
Wu (2001) have empirically analyzed the relationship between openness and rural-urban income
inequality in Chinese cities and their adjacent rural areas during between 1988 and 1993. They
dened openness as the ratio of exports to GDP. Using distance to major seaports as an instrument
for openness, they found negative correlation between rural-urban inequality and openness to trade.
On the contrary, Dayal-Gulati and Hussain (2002) found the inter-provincial inequality in income
to worsen between 1978 and 1997 as a result of FDI in
ows and technology transfer. They noticed
that the relatively rich coastal and North-eastern region, despite having relatively expensive labor,
were able to attract higher FDI in
ows compared to inland regions. As a result, these regions
converge to higher level of income as compared to inland provinces, atleast in short run. They cited
more developed infrastructure as major reason for higher FDI in
ows in these regions.
Fleisher et al. (2010) have analyzed the impact of regional dierences in physical, human, and
infrastructure capital as well as dierences in FDI
ows on regional growth patterns in China. They
have shown that FDI had a much larger eect on TFP growth before 1994. After that, they found
the impact of human, physical and infrastructure capital to be more signicant on TFP growth as
compared to the FDI in
ows. They found that while infrastructure investments generate higher
returns in more developed coastal and north-eastern regions than in interior, investing in human
capital generates slightly higher or comparable returns in interior regions. They propose investment
in human capital in less-developed areas, on eciency grounds, to reduce the inter-regional income
inequality in China.
21
Using Chinese Urban Household Survey data from 1988 to 2008, Han et al. (2012) analyze the im-
pact of globalization on wage inequality. They explore the impact of two major trade liberalization
shocks in China i.e. Deng Xiaoping's Southern Tour in 1992 and China's accession to the World
Trade Organization in 2001. Using distance of the region from the coast as a measure of exposure to
globalization, they found that the areas more exposed to globalization experienced larger increase
in wage inequality as compared to less exposed regions. They have also shown that both shocks
contributed to the within region inequality by raising the returns to education i.e. returns to high
school after 1992 and returns to college after 2001.
Similar to Han et al. (2012), our results also show that the highest premium is for college graduates
as compared to high school graduates. Han et al. (2014) found the coecient for the college
premium to be about 0.66 which is not much dierent from our results as they also included
Masters degree holders in college graduates. In addition, we also show that the rms which export
bigger proportion of their exports to the high income destination pay an extra premium to these
more educated workers which in turn, further increases the wage inequality in the region. We did
not nd any signicant impact of the exporting behavior of the rm on the average wages in the
rm. These results indicate that much of the income inequality that is attributed to the exporting
behavior is coming from exporting to high income countries as the workers that are involved in
exports to high income countries tend to get higher average wages.
In addition to the above analysis, we also provide the results for table 1.4 by running quantile
regressions for average wage.
2
We note varying skill premium for high school and above education
on dierent percentile of average wage. We notice that there is no premium for lowest decile and it
2
Due to the computing limitations, we have used two digit industry xed eects rather than 4 digit industry code
and city xed eects for this quantile based regression
22
rises up to 35 percent for the top decile. In addition, we also nd varying return to the High Income
export destination for these skilled (or educated) workers. It can be seen that the extra premium
for exporting to high income countries range between 11 percent at the lowest decile to 32 percent
for the top decile for these skilled workers. These results further conrm that the rms exporting
better quality products hire better quality skilled (or educated) workers and also rewards them by
paying higher wages.
1.7 Conclusion
In this paper, we have extended the recent literature analyzing the impact of export destination on
the rm activity in several ways. First, we have developed a theoretical model linking the export
destination to the level of skill (or education) premium being awarded at the rm level. The model
implies that rms exporting to high income destinations should pay higher wages to their skilled
(or more educated) workers. Second, we empirically test this theory using this huge cross-section
of manufacturing rms. We have found signicant and positive eect of export destination on
proportion of skilled workers in the rm and the skill premium being oered by the rm. As such,
it is the rst paper to establish a link between export destination and the skill premium at the
rm level. Third, in keeping with the recent literature, we nd no signicant impact of exporting
activity on either the proportion of skilled workers or the skill premium. Fourth, we have clearly
dened the skill level by education at the rm level while much of the work uses the average wage
or the share of non-production workers as proxy of skills due to the lack of data on the educational
distribution in the rm in the recent trade literature
3
. Finally, in contrast to most other empirical
3
This statement is true to the best of our knowledge.
23
studies that focus on Latin American countries , the present study focuses on China which has
recently become the world largest exporter.
However, to clearly establish the causal link between export destination and the skill premium in
the rm, we have to be careful about the endogeneity issue as we have cross-sectional data. All
these results establish a correlation between our variable of interests. Having only one observation
per rm, we cannot us the rm xed aects in the analysis. We have used the interacting dummies
for 4 digit industry code and the city code, assuming, that the rms in the same city and industry
code would be similar. The results still hold even after controlling at such detailed level.
4
4
It is currently work in progress and we are working on nding suitable instrument for our analysis. We have
used couple of instruments so far and the results still hold for the skill premium at the rm level, though, these are
weak instruments in nature.
24
Table 1.1: Summary Statistics for Chinese Manufacturing Firms in 2004
All Exporting Non-exporting
No. of Firms 162,867 21,977 140886
Skilled workers 0.435 0.42 0.437
Workers with high school degree 0.32 0.307 0.325
Workers with diploma 0.08 0.076 0.08
Workers with Bachelors degree 0.0295 0.035 0.028
Workers with Masters degree 0.0025 0.003 0.0024
Capital per worker (RMB) 65522 75260 63990
Average wage (RMB) 9446 10601 9278
Employees per rm 249 555 202
Fully State-owned 0.1 0.03 0.12
Mean sales (Million RMB) 46.6 104.85 37.53
Export to sales ratio 0.64
Proportion of export to HI countries 0.68
Average number of exporting destinations 21
Pure Exporter 4949
25
Table 1.2: Relationship between Price of the product and Export Destination
(1) (2) (3)
VARIABLES Log Price Log Price Log Price
HI Dummy 0.0543*** 0.0278*** 0.0221***
(0.00890) (0.00798) (0.00801)
MI Dummy 0.0195** 0.0108 -0.000979
(0.00902) (0.00809) (0.00820)
Log Quantity -0.255*** -0.256***
(0.00124) (0.00124)
Distance(km) 4.89e-06***
(5.72e-07)
Constant 1.314*** 3.320*** 3.294***
(0.00807) (0.0121) (0.0125)
Observations 200,988 200,988 200,988
R-squared 0.806 0.844 0.845
Firm Fixed Eects Yes Yes Yes
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: Dependant variable is the logarithm of the price of the product being charged by the rm for
each destination. HI Dummy and MI Dummy are dummy variables for High Income country and Middle
Income country dummy respectively. Log Quantity is the logarithm of the quantity being exported for each
transaction. We have controlled for Firm Fixed eects for this regression and standard errors are robust.
26
Table 1.3: Relationship between proportion of skilled workers and Export destination
(1) (2) (3)
VARIABLES SKILL SKILL SKILL
HI 0.0651*** 0.0453*** 0.0453***
(0.00608) (0.00543) (0.00537)
EXP -0.0259*** -0.0178*** -0.00370
(0.00662) (0.00606) (0.00627)
AGE -0.00363*** -0.00364***
(0.000246) (0.000246)
AGE
2
4.79e-05*** 4.79e-05***
(5.39e-06) (5.39e-06)
Cap 0.0679*** 0.0678***
(0.00118) (0.00117)
State Ownership 0.0859*** 0.0859***
(0.00420) (0.00420)
PE -0.0233***
(0.00596)
PAT -0.0231**
(0.00995)
FPT -0.0113
(0.00782)
Constant 0.425*** 0.107*** 0.107***
(0.000604) (0.00564) (0.00562)
Observations 155,155 155,129 155,129
R-squared 0.432 0.470 0.470
Industry-City Fixed Eects Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: Here Proportion of Skilled workers (SKILL) in a rm are regressed against the proportion of exports
to High Income countries (HI), Export to Sales ratio (EXP), Age of the rm, Logarithm of capital per worker
(Cap), State Ownership Dummy, Pure Exporter (PE) dummy, Feed and Processing Trade (FPT) dummy,
Processing and Assembly Trade (PAT) dummy. Skilled worker has been dened as the one with high school
and above level of education. We have controlled for Industry-City xed eects and standard errors have
been clustered at Industry-City level. There are 524 four-digit industries and 399 cities in the data.
27
Table 1.4: Relationship between Average wage and Export destination using broader measure of Skill
(1) (2) (3)
VARIABLES Average wage Average wage Average wage
SKILL 0.271*** 0.195*** 0.194***
(0.00912) (0.00860) (0.00859)
HI 0.0608*** 0.0493*** 0.0452***
(0.0153) (0.0149) (0.0148)
EXP -0.0201 -0.0161 0.00508
(0.0129) (0.0125) (0.0132)
HI*SKILL 0.266*** 0.231*** 0.230***
(0.0264) (0.0251) (0.0248)
AGE 0.00137*** 0.00134***
(0.000420) (0.000420)
AGE
2
-1.91e-05** -1.87e-05**
(8.01e-06) (8.01e-06)
Cap 0.0933*** 0.0932***
(0.00289) (0.00289)
State Ownership 0.0322*** 0.0323***
(0.0107) (0.0107)
PE -0.0455***
(0.0125)
PAT 0.0233
(0.0233)
FPT -0.00304
(0.0127)
Constant 1.898*** 1.464*** 1.465***
(0.00400) (0.0145) (0.0145)
Observations 155,155 155,129 155,129
R-squared 0.491 0.505 0.505
Industry-City Fixed Eects Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: Here, Logarithm of average wage (Average wage) in the rm is regressed against proportion of Skilled
workers (SKILL), the proportion of exports to High Income countries (HI), Export to Sales ratio (EXP),
interaction of HI and SKILL (HI*SKILL), Age of the rm, Logarithm of capital per worker (Cap), State
Ownership Dummy, Pure Exporter (PE) dummy, Feed and Processing Trade (FPT) dummy, Processing
and Assembly Trade (PAT) dummy. Skilled worker has been dened as the one with high school and above
level of education. HI*SKILL captures the premium for skilled (or educated) workers for exporting to high
income countries. We have controlled for Industry-City xed eects and standard errors have been clustered
at Industry-City level. There are 524 four-digit industries and 399 cities in the data.
28
Table 1.5: Relationship between Average wage and Export destination using detailed Skill level
(1) (2) (3)
VARIABLES Average wage Average wage Average wage
HI 0.0560*** 0.0473*** 0.0405***
(0.0150) (0.0148) (0.0146)
MS 0.798*** 0.669*** 0.670***
(0.181) (0.177) (0.177)
BS 1.096*** 0.907*** 0.908***
(0.0491) (0.0478) (0.0478)
Diploma 0.559*** 0.432*** 0.432***
(0.0269) (0.0264) (0.0263)
HS 0.110*** 0.0781*** 0.0781***
(0.00949) (0.00928) (0.00928)
HI*MS 1.109* 1.003 1.011
(0.646) (0.647) (0.638)
HI*BS 1.206*** 1.103*** 1.103***
(0.170) (0.164) (0.164)
HI*Diploma 0.351*** 0.331*** 0.333***
(0.0979) (0.0940) (0.0939)
HI*HS 0.0877*** 0.0687** 0.0643**
(0.0305) (0.0294) (0.0293)
EXP -0.00430 -0.00294 0.0138
(0.0125) (0.0122) (0.0130)
AGE 0.00201*** 0.00196***
(0.000415) (0.000415)
AGE
2
-2.43e-05*** -2.35e-05***
(7.97e-06) (7.96e-06)
Cap 0.0774*** 0.0773***
(0.00286) (0.00285)
State Ownership 0.0234** 0.0235**
(0.0105) (0.0105)
PE -0.0456***
(0.0123)
PAT 0.0386*
(0.0233)
FPT 0.0190
(0.0123)
Constant 1.904*** 1.537*** 1.538***
(0.00378) (0.0141) (0.0141)
Observations 155,155 155,129 155,129
R-squared 0.505 0.514 0.514
Industry-City Fixed Eects Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: Here, Logarithm of average wage (Average wage) in the rm is regressed against dis-aggregated
measure of education(or skill) i.e. High school (HS), technical diploma (Diploma), Bachelor's degree (BS) or
Masters degree (MS), the proportion of exports to High Income countries (HI), Export to Sales ratio (EXP),
interaction of HI and Y (HI*Y) where Y can be HS, Diploma, BS or MS, Age of the rm, Logarithm of
capital per worker (Cap), State Ownership Dummy, Pure Exporter (PE) dummy, Feed and Processing Trade
(FPT) dummy, Processing and Assembly Trade (PAT) dummy. HI*Y captures the additional premium for
educated (high school, diploma, Bachelor's or Masters) workers as a result of exporting to high income
countries. We have controlled for Industry-City xed eects and standard errors have been clustered at
Industry-City level. There are 524 four-digit industries and 399 cities in the data.
29
Table 1.6: Quantile Regression between Average Wage and Export Destination
(1) (2) (3) (4) (5)
VARIABLES 10% 25% 50% 75% 90%
SKILL 0.0124* 0.0179*** 0.0876*** 0.199*** 0.354***
(0.00732) (0.00627) (0.00564) (0.00709) (0.0109)
HI 0.112*** 0.123*** 0.116*** 0.113*** 0.0741***
(0.0164) (0.0141) (0.0119) (0.0133) (0.0195)
EXP 0.0193 0.0199 0.00962 -0.0237* -0.0245
(0.0141) (0.0122) (0.0105) (0.0126) (0.0158)
HI*SKILL 0.116*** 0.154*** 0.188*** 0.290*** 0.320***
(0.0277) (0.0233) (0.0209) (0.0249) (0.0360)
AGE 0.00925*** 0.00599*** 0.00151*** 0.000511 -0.000587
(0.000602) (0.000486) (0.000424) (0.000487) (0.000636)
AGE
2
-0.000245*** -0.000148*** -4.56e-05*** -2.11e-05* -8.59e-06
(1.57e-05) (1.22e-05) (1.08e-05) (1.18e-05) (1.35e-05)
Cap 0.0988*** 0.134*** 0.158*** 0.184*** 0.209***
(0.00190) (0.00162) (0.00146) (0.00178) (0.00266)
State Ownership -0.166*** -0.0790*** 0.0221*** 0.0986*** 0.131***
(0.00968) (0.00713) (0.00701) (0.00805) (0.00997)
PE -0.0671*** -0.0632*** -0.0449*** -0.0236** -0.00462
(0.0139) (0.0117) (0.0107) (0.0112) (0.0152)
PAT 0.0350** 0.0406** 0.0414** 0.0599** 0.0860***
(0.0163) (0.0177) (0.0163) (0.0272) (0.0166)
FPT 0.0106 0.00957 0.0131 0.00320 0.00539
(0.0167) (0.0114) (0.0101) (0.0108) (0.0156)
Constant 1.078*** 1.115*** 1.205*** 1.248*** 1.295***
(0.0224) (0.0177) (0.0156) (0.0162) (0.0256)
Observations 162,541 162,541 162,541 162,541 162,541
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: This table reports the quantile regression results for each decile of Logarithm of average wage against
proportion of skilled workers (SKILL), Export to Sales ratio (EXP), proportion of export to High Income
countries (HI) and interaction of HI and SKILL (HI*SKILL), Age of the rm, Logarithm of capital per
worker (Cap), State Ownership Dummy, Pure Exporter (PE) dummy, Feed and Processing Trade (FPT)
dummy, Processing and Assembly Trade (PAT) dummy. Here, we have controlled for Industry Fixed eects
at two digit level due to the computational limitation and standard errors are robust.
30
Figure 1.1: Relationship between average wage and the skill premium at dierent deciles of average
wage
0.00 0.10 0.20 0.30 0.40
SKILL
0 20 40 60 80 100
Quantile
Fig.1a
-0.10 -0.05 0.00 0.05
EXP
0 20 40 60 80 100
Quantile
Fig.1b
0.00 0.05 0.10 0.15
HI
0 20 40 60 80 100
Quantile
Fig.1c
0.00 0.10 0.20 0.30 0.40
HI*SKILL
0 20 40 60 80 100
Quantile
Fig.1d
Notes: This gure corresponds to Table 1.6 and traces out the impact of proportion of skilled workers
(SKILL), Export to Sales ratio (EXP), proportion of export to High Income countries (HI) and interaction
of HI and SKILL (HI*SKILL) on each decile of logarithm of average wage using the quantile regression.
Here, we have controlled for Industry Fixed eects at two digit level due to the computational limitation.
31
Chapter 2
Productivity Dierences Between and Within Firms-Case of
Taiwan
1
2.1 Introduction
Traditional economic theory assumes that the rm can easily replicate the production and as a
result, the economic decisions analyzed at the sector level, rm level or the plant level should be
identical. In the recent literature, the heterogeneity in productivity across rms within the sec-
tor has been acknowledged and found to be useful in understanding the working of the economy,
especially in the area of economic growth and international trade. Melitz (2003) develops a dy-
namic industry model with heterogeneous rms to analyze the intra-industry eects of international
trade. He showed the rm's response, to the exposure of trade, will depend on the rm's level of
productivity. He argued that the higher exposure to trade will eventually shift resources from
the low productivity rms to the higher productivity rms in the sector. Restuccia and Rogerson
(2008) have modied the standard growth model by incorporating the heterogeneity among the
plants and calibrate it to US data. They argued that the dierences in the allocation of resources
1
Joint work with Rahul Giri and Kensuke Teshima
32
across heterogeneous plants can be an important factor in accounting for dierences in output per
capita across countries. Hsieh and Klenow (2009) have also highlighted the issue of misallocation
of resources across heterogeneous producers in India, China and U.S. can explain the dierence in
manufacturing TFP across these countries. They calculated the manufacturing TFP gains to be in
range of 30%50% for China and 40%60% for India if China and India hypothetically reallocate
labor and capital to equalize the marginal products to the extent observed in U.S.
In much of these literature, researchers use the available rm level data or the plant level data to
empirically test the economic theories. Though, comparing these data-sets across countries, they
usually do not acknowledge the dierence between the rm level data or the plant level data. In
Hsieh and Klenow (2009), Chinese manufacturing data is available only at the rm level whereas
a the U.S. and Indian data is at the plant level. If there is some reallocation happening at the
plant level within the multi-plant rm, it might underestimate the role of reallocation across the
manufacturing establishments. It is important to recognize the rm-plant ownership structure even
if the rm is producing a single product across all its plants. It will help us to analyze the allocation
of resources (capital and labor) within the rm and how would this allocation changes within the
rm in response to some external shocks. In addition, failing to recognize the rm-plant structure
might lead to misleading interpretation of empirical results either at the plant level or the rm level.
Bernard et al. (2011) and Mayer et al. (2014) have recognized the multi-product structure within
the multi-product rms and have built theoretical models to analyze the impact of competition on
the allocation of resources across products within the rm.
Giri and Teshima (2013) have used the Mexican manufacturing plant-level data from 2003 to 2010
and they can identify the rm-plant structure in this data. They observed some important facts
33
about the multi-plant rms and within-rm across-plant heterogeneity. They found that realloca-
tion of labor across plant within rms contribute positively by 8 percent to the aggregate labor
productivity changes observed in the data. 8 % might be a smaller contribution at the aggregate
level but there are some industries (e.g. Fabricated Metal Product) where this reallocation of labor
across plants reduces the aggregate productivity by nearly 58 %. In addition, they noticed that the
exporter productivity premium is vastly underestimated in plant-level regressions as compared to
the rm-level regressions. Finally, for the trade crisis of 2008-2009, rm-level regressions suggest
that the export oriented rms did not loose many workers whereas plant-level regressions suggest
that exporting plants did suer signicantly as a result of the crisis. As a result, the conclusions
drawn on the impact of crisis on employment are starkly dierent between the rm-level results
and the plant-level results. The inability to recognize the rm-plant ownership structure, might
also lead to dierent results for the decisions at the rm-level and the plant-level. This dierence
exists for both multi-product as well as single-product multi-plant rms.
This paper, along with co-authors, is an extension of Giri and Teshima (2013). Primarily, though,
Mexican manufacturing rms data is diversied across industries and regions, but is smaller in size.
Taiwanese data has lot many more observations per year for single plant as well as multi-plant rms
and is also very diverse in terms of level of employment, types of products, sales etc. Using the same
analysis as used by Giri and Teshima (2013), we will test these results for Taiwanese Manufacturing
rms data and notice the similarities or dierences with the results for Mexican manufacturing
rms. Rest of the paper is organized as follows: In the next section, we will describe the data set
in detail and provide descriptive statistics for plant-level as well as rm-level variables. Section 3
will highlight the importance of multi-plant rm and analyze the contribution of within-rm across-
plant reallocation on aggregate labor productivity. Section 4 will analyze the exporter premium at
34
the rm-level and plant-level and compare the results. In the last section, we will conclude and
discuss the current extensions of this paper.
2.2 Data
We are using the Taiwanese manufacturing rms data collected by Ministry of Economic Aairs and
is called Factory Adjustment and Operation Survey. This survey provides Taiwanese manufacturing
plants data from 1998 to 2005. Though, the data for 2001 is missing from this survey. The survey
collects data about the manufacturing plants but also recognizes the rm-plant ownership. We can
check which plants in the data belong to which rm. The survey collects detailed information about
the plant's operation including the product or products being produced in the plant, sales of each
product, proportion of sales being exported, salary of the workers, capital employed by the plant
etc. The information about the type and quantity of the product being exported is only available
for two years. Rest of the plant level data used in this paper is available from 1998 to 2005, except
for the year 2001. There is a big variation among the plants in terms of employment. Nearly half
of the plants have an employment of less than 10 workers whereas the mean employment is around
27.
Table 2.1 gives the summary of employment and export status at plant as well as rm level for
Taiwanese data for 2000 and 2002. These are the years for which we have the export data available.
Though, it should be noted that most of the rms are also single plant rms. For this reason, we
don't see much dierence in summary statistics at plant and the rm level. Nearly 20 percent of the
plants as well as the rms are exporting for each of the reported years. Approximately 9 percent
of the sales are exported abroad.
35
Here we notice that, on average, the plants belonging to the multi-plant rms employ 4 times more
workers as compared to the plants belonging to the single plant rms. Among the multi-plant rms,
nearly half of them are rms selling some proportions of their sales abroad. In addition, nearly
25 percent of the plants belonging to the multi-plant rms are involved in the exporting activity
whereas for single plant rms, this number is around 19 percent.
Table 2.2 lists the share of the multi-plant rms in the employment and sales statistics in the data
over time. We notice that nearly 10 percent of the plants belong to the multi-plant rms in number,
though, their corresponding share is much higher in terms of employment and sales statistics i.e.
30 percent and 50 percent respectively. It shows that the plants belonging to the multi-plant rms
are much bigger in size and operations as compared to the single plant rms. Though, over time,
we see a decline in both the single plant rms and multi-plant rms by the same magnitude.
In Table 2.3, we focus on the multi-plant rms only. We are analyzing the share of the multi-plant
rms that operate in more than one 4 digit industry code, hence called multi-industry rms. We
notice that over the time, their share in number of plants as well as in employment, sales and
exports is going down. Though the share of the multi-plant rms in aggregate data is between 9
and 10 percent from 1998 to 2005, but the share of multi-industry rm has declined by signicant
number especially since 2003. Later,in appendix, we will analyze whether this big decline is because
of switching towards specialization by these rm to single industry or whether they shut down the
plants and became single plant rms or they exited the market all together.
Table 2.4 focuses more on the multi-industry rms and lists the share of the most important
industry, by sales, in number of plants, employment and production. Here we notice that the
number of plants assigned to the production of the good in top industry increases their share from
36
nearly 49 percent to 51 percent. Correspondingly, we also notice increase in employment from 65
percent to 69 percent. Though, production share increases by a smaller margin from 71.5 percent
to 72.7 percent. This table indicates that there is specialization going on among the multi-industry
multi-plant rms.
In Table 2.5, 2.6 and 2.7, we focus on the top plant among the multi-plant rms, in aggregate,
single industry or multi-industry rms respectively. We notice that the share of the top plant in
employment and sales is nearly two-thirds, irrespective of whether we look at single industry rms,
multi-industry or analyze them in aggregate. In addition, their share also stays the same over
time as well. These tables highlight the fact that even these rms are multi-plant rms but there
production is concentrated in the biggest plant of the rm.
2.3 Multi-plant Firms and Labor Productivity
In the previous section, we noticed that the multi-plant rms contribute signicantly towards em-
ployment, production as well as exports in the economy. In this section, we will analyze the labor
productivity across dierent plants within the multi-plant rms. Labor productivity is dened as
the ratio of value addition by the plant to the employment.
Table 2.8 compares the average labor productivity between the most productive plant (Top) and
the remaining plants (Non-Top) within the multi-plant rm and reports their percentile rank for
employment, production, exports and labor productivity. In order to dene a plant to be the most
productive, we do not dierentiate whether the rm is single-industry rm or a multi-industry rm.
To calculate the percentile rank for each plant, we are using all the plants in the 4 digit industry
code, irrespective of either belonging to the single plant rms or multi-plant rms. We can notice
37
that the Top plants are, on average, ranked 9 percentile higher as compared to the average of the
percentile rank of other plants in the rm. We do not nd much dierence in percentile rank for
employment and exports for the most productive plants. Table 2.9 reports the R
2
for the following
y
ijst
=
st
+
ijst
(2.1)
y
ijst
=
it
+
ijst
(2.2)
wherey
ijst
denotes labor productivity for rmi, plantj, industrys and yeart.
st
and
it
represent
the industry xed eects and rm xed eects. We can notice that nearly 15 % variation in plant
level labor productivity can be explained by the 4 digit industry xed eects whereas nearly two
third of the variation can be explained by the rm xed eects. For this analysis, we have only
used the multi-plant rms. Though, one-third of the variation is still unexplained even after using
the rm xed eects.
Table 2.10 compares the size of the plant within a multi-plant rm and the labor productivity,
wages being paid and the export to sales ratio for the plant. Table 2.10 presents the result for the
following regression.
Y
ijt
=
0
+
1
Cost +
it
+
ijt
(2.3)
where Y
ijt
is the logarithm of labor productivity, wages or export to sales ratio for rm i, plant j
in yeart andCost is the share of cost share of inputs for the plant.
it
captures the rm-year xed
eects. We ran this regression for all multi-plant rms together, single-industry multi-plant rms
as well as multi-industry multi-plant rms. The results suggest that bigger plants in the rm have
higher labor productivity, pay higher wages as well as exports higher proportion of their sales.
38
Table 2.8, 2.9 and 2.10 conrm the fact that even within a multi-plant rm, there is a certain
degree of heterogeneity among the plants even if they are producing the same product. This
heterogeneity is bigger if the plants are producing dierent products. The magnitude of coecients
are comparable with Giri and Teshima (2013) for each of the dependent variable, though, they found
bigger heterogeneity among the single industry multi-plant rms as compared to the multi-industry
multi-plant rms. These results conrms that the heterogeneity among the plants exist even if they
belong to the same rm and provide further support to the recent literature in international trade
which analyze the rm-level heterogeneity within the same industry.
2.4 Exporter Premium
In earlier section, we found that the bigger plants, within the multi-plant rms, are more likely to
export bigger share of their output as compared to smaller plants. In this section, we will analyze
the exporting premium for employment, average wage and labor productivity at the plant as well
as rm level.
Table 2.11 presents the results for the following:
log(Y
ijst
) =
0
+
1
FE
it
+
2
PE
ijt
+
3
MPF
it
+
st
+
ijst
(2.4)
Y
ijst
represents the logarithm of employment, average wages and labor productivity for plant j
of rm i operating in industry s in year t. FE is the rm exporter dummy, PE is the plant
exporter dummy,
st
is the industry-year eect and MPF is the dummy for the plant belonging
39
to the multi-plant rm or not
2
. The results in table 11 suggests that the exporting plants within a
multi-plant rm have an extra premium in terms of employment and wages as compared to the non-
exporting plants. Similarly, the non-exporting plants belonging an exporting multi-plant rm also
enjoys a premium in terms of employment, average wages and labor productivity. We also found
employment, wage as well as productivity premium for belonging to a multi-plant rm. Intuitively,
it might be because a rm will set up a new plant if it is big enough in size and as we have seen
in earlier results, that the bigger plants pay higher wages and have higher labor productivity as
well. The more interesting result in table 13 is that there is not a signicant dierence in labor
productivity of a non-exporting plant and an exporting plant within an exporting multi-plant rm.
It suggests that the labor productivity is determined at the rm level rather than at the plant level.
Though, the workers in exporting plant earns 4.5 percent higher wages as compared to the workers
in non-exporting plant despite that there is no signicant dierence in labor productivity. Table
2.12 estimate the following regression
log(y
ist
) =
0
+
1
Exporter
it
+
3
MPF
it
+
st
+
ist
(2.5)
where y
ist
is the labor productivity of the plant (or rm) i in year t. Here, we are comparing the
result for export premium for labor productivity when we analyze the data at the plant level or
rm level. First two columns are at the plant level and column (3) and (4) are at the rm level.
We notice that the premium for exporting is 3 percent higher at the rm level analysis as compared
to analyzing at the plant level. This dierence is much smaller as compared to Giri and Teshima
(2013) for Mexico. They found the exporting premium for productivity at the rm level to be 3
2
Here, we have included the single plant rms as well to the analysis. If a single plant rm is exporting rm, in
that case, the plant exporting as well as rm exporting dummy will be 1. In case of multi-plant rms, we can have
a exporting plant as well as non-exporting plants.
40
times higher as compared to analyzing it at the plant level. Table 2.11 and 2.12 show that there
exists a exporter premium for labor productivity but it is determined at the rm level rather than at
plant level as we noticed that there is no signicant dierence between exporting and non-exporting
plants of the exporting multi-plant rm. There is an added premium in terms of employment and
wages is at the plant level if the plant is an exporting plant in a multi-plant rm. These results,
along with Giri and Teshima (2013) result for Mexico, highlight the fact that we should be careful
in comparing the results for plant level data and rm level data.
2.5 Conclusion
This chapter highlights the existence of heterogeneity in economic characteristics of plants belonging
to the same rm and producing the same product. Using the plant level data for Taiwanese
manufacturing rms data from 1998 to 2005, we found that the bigger plants within the rm have
higher productivity, pay higher wages and export bigger proportion of their output. We also found
that there exists an extra exporting premium at the plant level for employment as well as wages
for exporting multi-plant rm, though, there is no extra exporting premium for labor productivity.
This suggests that there is no dierence in labor productivity of an exporting plant and a non-
exporting plant within an exporting rm, though, there is signicant dierence in wages. Our
results are conrms the results for Mexico by Giri and Teshima (2013), though, the magnitude are
dierent. This suggests that plant level data analysis will underestimate the exporter premium
for labor productivity in contrast to the rm level data analysis. This dierence in productivity
across plants within the rm and reallocation of workers across plants within a rm might contribute
towards variation in aggregate labor productivity of the economy. Giri and Teshima (2013) reported
41
that this reallocation of workers across plants contribute about 8 percent towards aggregate change
in productivity over time.
42
Table 2.1: Summary Statistics of Key Variables at Plant and Firm Level
Single-Plant Firm Multi-plant Firm
Firm = Plant Plant Firm
2000
Number of Workers 20.94 86.95 201.584
Exporter Dummy .18 .24 .50
Export/Total Sales .08 .10 .16
2002
Number of Workers 21.13 94.77 221.31
Exporter Dummy .20 .25 .53
Export/Total Sales .09 .11 .16
Table 2.2: No. of Multi-plant Firms and Its Share in Employment and Sales: 1998-2005
Single-Plant Multi-plant N Plant Employment Sales
1998 71042 7396 .094 .29727 .497
1999 70898 7962 .101 .31229 .514
2000 70123 7990 .102 .31834 .518
2002 63483 7055 .1 .32956 .549
2003 62615 7562 .108 .33089 .497
2004 66456 6811 .093 .30554 .479
2005 57924 5608 .088 .29274 .478
Table 2.3: Number of Multi-industry Firms and Its Shares in Employment and Sales Within Multi-
plant Firms: 1998-2005
Single
Industry
Multi-
industry
N Plant Employment Sales Exports
1998 1461 1745 .579 .695 .77
1999 1689 1776 .548 .643 .707
2000 1739 1728 .53 .643 .725 .729
2002 1620 1435 .503 .593 .688 .632
2003 1895 1354 .455 .524 .64
2004 1973 971 .369 .461 .612
2005 1627 808 .367 .432 .603
43
Table 2.4: Importance of Most Important Industry for Multi-industry-Multi-Plant Firms: 1998-
2005
Year Non Top Top N Plant Emp Production Exports
1998 2184 2100 .4901961 .6548517 .7146152
1999 2216 2145 .4918597 .6222324 .7190678
2000 2130 2106 .4971671 .6295397 .7324819 .869626
2002 1777 1771 .4991544 .6179795 .722629 .8854856
2003 1683 1756 .5106136 .6816633 .7200197
2004 1220 1291 .5141378 .6889977 .7239423
2005 997 1059 .5150778 .695155 .7281231
Table 2.5: Importance of Top Plant for Multi-Plant Firms (Aggregate): 1998-2005
Year Employment Share Sales Share
1998 .6697462 .6481469
1999 .6704286 .630712
2000 .6770762 .637325
2002 .6649355 .6213875
2003 .6664855 .6625078
2004 .6701203 .6652266
2005 .6645018 .6603954
Table 2.6: Importance of Top Plant for Multi-Plant Firms (Single Industry): 1998-2005
Year Employment Share Sales Share
1998 .6787661 .6496539
1999 .6798488 .6357498
2000 .6822401 .644517
2002 .6657429 .6262115
2003 .6720917 .6686428
2004 .6716448 .6691412
2005 .6616883 .6578794
44
Table 2.7: Importance of Top Plant for Multi-Plant Firms (Multi Industry): 1998-2005
Year Employment Share Sales Share
1998 .6621943 .6469617
1999 .6614699 .6241083
2000 .6718794 .6304553
2002 .6640241 .6169262
2003 .6586394 .6515616
2004 .6670226 .6574212
2005 .6701673 .6636224
Table 2.8: Heterogeneity between the most labor productive plants and other plants within rms
Non-
Top
(Pro-
duc-
tion)
Top
(Pro-
duc-
tion)
Non-
Top
(Em-
ploy-
ment)
Top
(Em-
ploy-
ment)
Non-
Top
(Ex-
port)
Top
(Ex-
port)
Non-
Top
(Labor
Pro-
ductiv-
ity)
Top
(Labor
Pro-
ductiv-
ity)
Single Industry .62 .65 .64 .63 .83 .83 .57 .66
Multi Industry .67 .7 .67 .66 .82 .83 .61 .7
Table 2.9: Labor Productivity (1) across and within industry and (2) across and within rm
All Firms Multi-plant Firms
VARIABLES Industry Eects Firm Eects Industry Eects Firm Eects
Constant 872.1*** 872.1*** 1,483*** 1,483***
(2.931) (3.539) (14.52) (12.13)
Observations 431,353 431,353 36,732 36,732
R-squared 0.070 0.935 0.156 0.658
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
45
Table 2.10: Heterogeneity between bigger plants and other plants within rms
VARIABLES Log Labor
Productivity
Log Average Wage Export/Sales
All Firms
Cost Share 0.439*** 0.202*** 0.117***
(0.0114) (0.00646) (0.0209)
Constant 6.665*** 5.764*** 0.141***
(0.00552) (0.00318) (0.0107)
Observations 36,732 36,668 6,153
R-squared 0.857 0.857 0.907
Firm-year Eects Yes Yes Yes
Single-Industry Firms
Cost Share 0.338*** 0.173*** 0.0977***
(0.0139) (0.00865) (0.0319)
Constant 6.614*** 5.727*** 0.153***
(0.00705) (0.00437) (0.0167)
Observations 18,450 18,412 3,094
R-squared 0.901 0.895 0.931
Firm-year Eects Yes Yes Yes
Multi-Industry Firms
Cost Share 0.534*** 0.230*** 0.133***
(0.0174) (0.00947) (0.0273)
Constant 6.725*** 5.803*** 0.131***
(0.00809) (0.00452) (0.0138)
Observations 18,282 18,256 3,059
R-squared 0.817 0.819 0.880
Firm-year Eects Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
46
Table 2.11: Exporter Premium with both Firm Exporter Dummy and Plant Exporter Dummy.
2000-2002
VARIABLES Log Employment Log Average Wage Log Labor
Productivity
Firm Export Dummy 0.825*** 0.193*** 0.319***
(0.0516) (0.0172) (0.0287)
Plant Export Dummy 0.316*** 0.0729*** 0.0801***
(0.0496) (0.0165) (0.0267)
Multi-Plant Firm Dummy 0.439*** 0.132*** 0.253***
(0.0238) (0.0110) (0.0150)
Constant 1.965*** 5.565*** 6.297***
(0.00384) (0.00120) (0.00230)
Observations 119,501 118,724 119,501
R-squared 0.286 0.144 0.162
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 2.12: Exporter Premium with both Firm Exporter Dummy and Plant Exporter Dummy.
2000-2002
VARIABLES Plant Plant Firm Firm
Plant Export Dummy 0.372*** 0.364***
(0.00779) (0.00779)
Multi-Plant Firm Dummy 0.372*** 0.282***
(0.0142) (0.0142)
Firm Export Dummy 0.404*** 0.382***
(0.00879) (0.00825)
Constant 6.345*** 6.314*** 6.314*** 6.307***
(0.00147) (0.00224) (0.00172) (0.00184)
Observations 127,867 127,867 121,726 121,726
R-squared 0.144 0.162 0.144 0.149
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
47
References
[1] Arkolakis, Costas, and Marc-Andreas Muendler. 2010. \The extensive margin of exporting prod-
ucts: A rm-level analysis." NBER Working Paper
[2] Bernard, Andrew B, Stephen J Redding, and Peter K Schott. 2010. \Multiple-Product Firms and
Product Switching." American Economic Review, 100(1): 70-97.
[3] Bernard, Andrew B, Stephen J Redding, and Peter K Schott. 2011. \Multiproduct rms and trade
liberalization." The Quarterly Journal of Economics, 126(3): 1271-1318.
[4] Bastos, Paulo, and Joana Silva. 2010. \The Quality of a Firm's Exports: Where You Export to
Matters." Journal of International Economics 82 (2): 99-111.
[5] Boumellassa Houssein, Debucquet David Laborde & Mitaritonna Cristina, 2009. \A picture of
tari protection across the World in 2004: MAcMap-HS6, Version 2," IFPRI discussion papers
903, International Food Policy Research Institute (IFPRI)
[6] Belton Fleisher, Haizheng Li, Min Qiang Zhao. 2010 \Human capital, economic growth, and
regional inequality in China Journal of Development Economics, Volume 92, Issue 2, Pages 215-231
[7] Bernard, Andrew B., and J. Bradford Jensen. 1995. \Exporters, Jobs, and Wages in U S Manu-
facturing: 1976-1987." Brookings Papers on Economic Activity: 67- 112.
[8] Brambilla, I. D. Lederman, and G. Porto. 2012. \Exports, Export Destinations and Skills,"
American Economic Review, Vol. 102, No. 7, pp. 3406-3438.
[9] Brambilla, I. and G. Porto. 2016. \High Income Export Destinations, quality and wages" Journal
of International Economics, Vol. 98, No. 7, pp. 21-35.
[10] Bustos, Paula. 2011. \Trade Liberalization, Exports, and Technology Upgrading: Evidence on the
Impact of MERCOSUR on Argentinian Firms." American Economic Review 101( 1): 304-4
48
[11] Clerides, Sofronis K., Saul Lach, and James R. Tybout. 1998. \Is Learning by Exporting Im-
portant? Micro-dynamic Evidence from Colombia, Mexico, and Morocco." Quarterly Journal of
Economics 113 (3): 903-47.
[12] Dayal-Gulati, Anuradha and Husain, Asim M. 2002 \Centripetal Forces in Chinas Economic Takeo
IMF Working Paper series WP/00/86.
[13] Frazer, Garth. 2013. \Imports, Import Sources and Skill Utilization, mimeo, University of Toronto.
[14] Ge, Suqin and Dennis Tao Yang. 2012. \Changes in Chinas Wage Structure Working paper.
[15] Giri, Rahul, Enrique Seira, and Kensuke Teshima. 2012. \Did Trade Crisis Aect Dierent Ex-
porters Dierently? Case of Mexico." Unpublished Working paper
[16] Giri, Rahul, and Kensuke Teshima. 2013. \Productivity Dierences Between and Within Firms."
Unpublished Working paper
[17] Goldberg, P., Pavcnik, N., 2007. \Distributional Eects of Trade Liberalization in Developing
Countries. Journal of Economic Literature 45 (1), 3982.
[18] Grg, Holger and Halpern, Laszlo and Murakozy , Balazs. 2010 \Why Do within Firm-Product
Export Prices Dier Across Markets?. CEPR Discussion Paper No. DP7708
[19] Gutierrez, Emilio, and Kensuke Teshima. 2011. \Import Competition, Technology and Environ-
mental Performance: Evidence from Mexico." Unpublished Working paper.
[20] Hsieh, Chang-Tai, and Peter J. Klenow. 2009. \Misallocation and Manufacturing TFP in China
and India." The Quarterly Journal of Economics, 124(4): 1403-1448.
[21] Jun Han, Runjuan Liu, Junsen Zhang, 2012. \Globalization and wage inequality: Evidence from
urban China. Journal of International Economics, Volume 87, Issue 2, Pages 288-297
[22] Lu, J., Lu, Y. and Tao, Z., 2014. \Pure Exporter: Theory and Evidence from China. The World
Economy, 37 (9). 1219-1236
49
[23] Mary Amiti, Lisa Cameron. 2012. \Trade Liberalization and the Wage Skill Premium: Evidence
from Indonesia. Journal of International Economics 87:2, 277-287
[24] Manova, Kalina, and Zhiwei Zhang. 2012. \Export Prices across Firms and Destinations." Quar-
terly Journal of Economics 127( 1): 379-436.
[25] Matsuyama, Kiminori. 2007. \Beyond Icebergs: Towards a Theory of Biased Globalization."
Review of Economic Studies 74 (1): 237-53.
[26] Mayer, T. Zignago, S. 2011 \Notes on CEPIIs distances measures: the GeoDist Database CEPII
Working Paper 2011-25
[27] Mayer, Thierry, Marc J Melitz, and Gianmarco IP Ottaviano. 2014. \Market Size, Competition,
and the Product Mix of Exporters." American Economic Review, 104 (2): 495-536
Melitz, Marc J. 2003. "The Impact of Trade on Intra-industry Real locations and Aggregate
Industry Productivity." Econometrica 71 ( 6): 1695-725.
[28] Melitz, Marc J., and Daniel Treer. 2012. \Gains from Trade When Firms Matter." Journal of
Economic Perspectives, 26(2): 91-118.
[29] Melitz, Marc J., and Stephen J. Redding. 2012. \Heterogeneous Firms and Trade." National
Bureau of Economic Research, Inc NBER Working Papers 18652.
[30] M Dai, M Maitra, M Yu. 2012 \Unexceptional Exporter Performance in China? The role of
Processing Trade The Role of Processing Trade
[31] Ouimet, Paige and Zarutskie, Rebecca. 2014. \Who works for startups? The relation between rm
age, employee age, and growth Journal of Financial Economics 112: 386-407
[32] Pages, Carman, G. Pierre, and S Scarpetta. 2009. \Job Creation in Latin America and the
Caribbean : Recent Trends and the Policy Challenges." Latin American development forum. Wash-
ington, DC: World Bank.
50
[33] Restuccia, Diego, and Richard Rogerson. 2008. \Policy Distortions and Aggregate Productivity
with Heterogeneous Plants." Review of Economic Dynamics, 11(4): 707-720.
[34] Restuccia, Diego, and Richard Rogerson. 2013. \Misallocation and productivity." Review of Eco-
nomic Dynamics, 16(1): 1-10.
[35] Serti, F. and Tomasi, C. 2008. \Firm heterogeneity: do destinations of exports and origins of
imports matter? LEM Working Paper 200814
[36] Syverson, Chad. 2011. \What Determines Productivity?" Journal of Economic Literature, 49(2):
326-365.
[37] Verhoogen, E. A. 2008. \Trade, Quality Upgrading and Wage Inequality in the Mexican Manufac-
turing Sector, Quarterly Journal of Economics, 123(2), 489530.
[38] Wei, Shang-jin and Yi Wu. 2001. \Globalization and Inequality: Evidence from Within China.
NBER Working Paper No. 8611.
[39] Yeaple, Stephen Ross. 2005. \A Simple Model of Firm Heterogeneity, International Trade, and
Wages." Journal of International Economics 65 (1): 1-20.
[40] Zhang, Y., 2010. \Globalization and Firm Demand for Skilled Labor in Chinas Manufacturing
Sector, in Hahn, C. H. and D. Narjoko (eds.), Causes and Consequences of Globalization in East
Asia: What Do the Micro Data Analyses Show? ERIA Research Project Report 2009-2, Jakarta:
ERIA. pp.313-338.
51
Appendix A
Appendix
In this appendix, we further dissect table 2.2 and explore the changes in single plant rms and
multi-plant rms over time.
In table A.1, we are keeping track of the plants that belonged to the single plant rm in the current
year and switched to multi-plant rm in the next year. Here, year by year, we are exploring how
many rms are switching from single plant rms to multi-plant rms and vice versa over time. The
rst row indicates that 540 rms switched from single plant rms to multi-plant rms by adding a
plant from 1998 to 1999 and 332 rms changed their status from multi-plant rms to single plant
rms. It would give us an idea about how many rms are expanding or reducing their size over
time. We notice that from 2003 onward, very few rms tried to change their status from single
plant rms to multi-plant rms as compared to earlier years. Though, from 2000 onward, their
has been a declining trend in changing the type of rm on both sides, but it has been much more
sharper among the expanding rms.
52
Table A.2 provides the further break up of Table A.1. Here we are keeping record of the rms which
changes from single plant ( by nature single industry) rms to multi-plant/multi-industry or multi-
plant/single-industry or vice versa. This table describes whether the rms which expand from single
plant to multi-plant rms, do they expand in the same 4 digit industry or a dierent one. Every
row describes the change in the status of the plant from that year to the next. That is why last row
has zeros in there. So, between 2003 and 2004, we notice that 13 rms changed from single plant to
multi-plant/multi-industry rms and 111 rms changed from multi-plant/multi-industry to single
plant rms. That contributes a net decline of 98 rms in Table 3 for multi-industry/multi-plant
rms.
In Table A.3, we are keeping track of the rms that change their status from single industry to
multi-industry or vice versa. It should be noted that here we have included the single plant rms
as well which are, by nature, single industry type. The numbers in this table are compatible with
Table A.2 and Table A.5.
Rest of the analysis has been dedicated to dissecting Table 2.3, which records sharp decline in the
share of the multi-industry rms among the multi-plant rms over time.
To illustrate the importance of the following tables, let's note the decline in the number of multi-
industry/multi-plant rms from 1377 in 2003 to 996 in 2004. This is a decline of 381 rms in one
year. We will use the following tables to account for this change of 381 rms.
Table A.4 provides information about the multi-plant/multi-industry rms which either exited at
the end of the year reported or entered in the beginning of the reported year. It includes both
whether the rm is entering/exiting forever or temporarily between 1998 and 2005. we notice that
53
26 rms exited at the end of 2003 and just 5 rms entered at the start of 2004. So, entry/exit
phenomenon contributes decline of 21 rms in Table 2.3.
Rest of the decline in Table 2.3 can be explained by Table A.5. It records the switching status from
single-industry to multi-industry or vice versa among the multi-plant rms over time. Between
2003 and 2004, 50 rms switched from single-industry to multi-industry and 309 rms switched in
the opposite direction. That contributes a decline of 262 rms in Table 2.3. So, Table A.2, A.3,
A.4 and A.5 can explain the variations in Table 2.3.
Table A.4, A.5, A.6 and A.7 have been calculated to analyze Table A.2 in detail as biggest contri-
bution to the variation in Table 2.5 comes from Table A.3.
In Table A.4 reports the number of rms that switched from single industry to multi-industry by
adding new plants and vice versa by closing plants.
In Table A.7, analyzes the switching behavior between type of industry status for the rms who
don't change the number of plants. Though, it's possible that a rm who has 3 plants in 2003 and
it closes an old plant and opens up a new plant and change the type of industry status.
In Table A.8, we are keeping track of the rms which just switched from single (multi) industry to
multi (single) industry without shutting down or opening up a plant. These are the rms which
have specialized (diversied) their industry line keeping the same plants over years. We notice that
it contributes the biggest decline to multi-industry rms from 2003 to 2004. It indicates that most
of the rms who changed their status from multi-industry to single-industry did so using same
plants between 2003 and 2004.
54
Table A.1: No. of Firms Switching Plant-Type (Single to Multi & Vice versa): 1998-2005
Single-Plant to Multi-Plant Multi-Plant to Single-Plant
1998 540 332
1999 460 419
2000 549 743
2002 408 234
2003 46 265
2004 30 246
2005 0 0
Table A.2: No. of Firms Switching Industry-Type (Single to Multi & Vice versa): 1998-2005
Single-Plant to
Multi-Plant/Single-
Industry
Single-Plant to
Multi-Plant/Multi-
Industry
Multi-Plant/Multi-
Industry to
Single-Plant
Multi-Plant/Single-
Industry to
Single-Plant
1998 314 226 174 158
1999 250 210 189 230
2000 302 247 347 396
2002 275 133 111 123
2003 33 13 106 159
2004 18 12 104 142
2005 0 0 0 0
Table A.3: No. of Firms Switching Industry-Type (Single to Multi & Vice versa): 1998-2005
Single-Industry to Multi-Industry Multi-Industry to Single-Industry
1998 269 256
1999 518 546
2000 326 501
2002 207 300
2003 63 415
2004 117 217
2005 0 0
55
Table A.4: No. of Multi-plant-Multi-industry Firms Entering or Exiting: 1998-2005
Exiting Firms Entering Firms
1998 57 1745
1999 59 75
2000 151 39
2002 42 33
2003 45 54
2004 74 14
2005 808 11
Table A.5: No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005
Single-Industry to Multi-Industry Multi-Industry to Single-Industry
1998 43 82
1999 308 357
2000 79 154
2002 74 189
2003 50 309
2004 105 113
2005 0 0
Table A.6: No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005
Single-Industry to Multi-Industry by
Adding
Multi-Industry to Single-Industry by
Dropping
1998 13 14
1999 29 18
2000 25 26
2002 7 10
2003 0 15
2004 2 17
2005 0 0
56
Table A.7: No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005
Single-Industry to Multi-Industry with
same number of plants
Multi-Industry to Single-Industry with
same number of plants
1998 29 64
1999 277 330
2000 52 123
2002 65 176
2003 50 294
2004 103 96
2005 0 0
Table A.8: No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005
Single-Industry to Multi-Industry without
Adding or Dropping
Multi-Industry to Single-Industry without
Adding or Dropping
1998 25 60
1999 272 324
2000 41 110
2002 64 172
2003 50 294
2004 103 96
2005 0 0
Table A.9: No. of Firms Switching Industry-Type for Multi-plant rms only (Single to Multi &
Vice versa): 1998-2005
Single-Industry to Multi-Industry when
Dropping Plants
Multi-Industry to Single-Industry when
Adding Plants
1998 1 4
1999 2 9
2000 2 5
2002 2 3
2003 0 0
2004 0 0
2005 0 0
57
Abstract (if available)
Abstract
This dissertation explores the implication of firm's interaction in the international market and its impact on optimal decisions by the firm. ❧ The first Chapter, titled "Export Destination, Skill Utilization and Skill Premium in Chinese Manufacturing sector", joint work with Junjie Xia, analyzes the link between export destination, skill utilization and skill premium. We propose a mechanism behind these links: the difference in quality valuation of the product across exporting destinations and the distribution of level of skill among the skilled workers in the labor market. We test this theory using a cross-section of more than 160,000 single product Chinese Manufacturing firms survey data, of which nearly 22,000 are exporting to more than 200 countries across the world. Contrary to the earlier literature, we explicitly use education as a measure of skill. We find that firms exporting to high-income countries tend to charge a higher price, pay higher average wages to employees, hire more skilled workers as defined by education level, and pay higher skill premium as compared to firms exporting to middle or low-income countries or selling domestically. As in similar recent studies, we did not find exporting per se to significantly impact the proportion of skilled workers or the skill premium in the firm. ❧ The second Chapter "Productivity Differences Between and Within Firms-Case of Taiwan", joint work with Kensuke Teshima and Rahul Giri, highlights the existence of heterogeneity in economic characteristics of plants belonging to the same firm and producing the same product. Using the plant level data for Taiwanese manufacturing firms data from 1998 to 2005, we found some interesting results for multi-plant firms and the exporter premium among these multi-plant firms.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Three essays on macro and labor finance
PDF
Essays on development and health economics
PDF
Three essays on heterogeneous responses to macroeconomic shocks
PDF
Essays in macroeconomics
PDF
Essays on macroeconomics and income distribution
PDF
Economic policy-making in a changing environment: the politics of success in Peru vis-à-vis China
PDF
Essays on the economics of non-profit institutions
PDF
Essays on sovereign debt
PDF
The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector
PDF
Essays in health economics and provider behavior
PDF
Essays on education and institutions in developing countries
PDF
Two essays on financial crisis and banking
PDF
Essays on political economy of privatization
PDF
Essays in development economics
PDF
Essays in environmental economics
PDF
Growth and development in Africa: macroeconomic and microeconomic perspectives
PDF
Essays on competition between multiproduct firms
PDF
Essays on environmental economics: education, employment, and experiments
PDF
The political economy of R&D collaboration: micro- and macro-level implications
PDF
Essays on the new economic history of the Middle East
Asset Metadata
Creator
Khan, Bilal Muhammad
(author)
Core Title
Essays in international economics
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Economics
Publication Date
07/24/2017
Defense Date
07/19/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
export destinations,Exports,firm heterogeneity,Inequality,OAI-PMH Harvest,skill premium,trade,wages
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Imrohoroglu, Ayse (
committee chair
), Aizenman, Joshua (
committee member
), Nugent, Jeffrey B. (
committee member
), Tuzel, Selale (
committee member
)
Creator Email
bkhan@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-409143
Unique identifier
UC11264490
Identifier
etd-KhanBilalM-5596.pdf (filename),usctheses-c40-409143 (legacy record id)
Legacy Identifier
etd-KhanBilalM-5596.pdf
Dmrecord
409143
Document Type
Thesis
Rights
Khan, Bilal Muhammad
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
export destinations
firm heterogeneity
skill premium