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
/
The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector
(USC Thesis Other)
The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE IMPACT OF TRADE LIBERALIZATION ON FIRM PERFORMANCE IN
DEVELOPING COUNTRIES – NEW EVIDENCE FROM PAKISTANI
MANUFACTURING SECTOR
by
Zara Liaqat
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
December 2012
Copyright 2012 Zara Liaqat
ii
Dedication
For my parents & my brother,
without whose affection, support & prayers I could not have achieved this.
And for my husband – my eternal love.
iii
Acknowledgments
I am grateful foremost to Allah for answering my prayers.
This dissertation would not have been possible without the love and support of
my mother, Farasat Shafi, my father, Rai Liaqat Ali, and my brother, Rai Ghulam
Mustafa. I am thankful to them because without their constant encouragement I would
never have survived the Ph.D. program. I truly appreciate the help I received from my
loving husband, Raqib Omer, without whose support completing this dissertation would
have been an unattainable task.
My mentor back in Pakistan, Sir Mumtaz Ahmed, inspired me to pursue an
academic career in Economics and put me on the path to my Ph.D. I am eternally grateful
for his guidance and support. The Department of Economics at The University of
Southern California gave me a chance to live my dream. My sincere thanks and gratitude
to all the faculty and staff for giving me an opportunity to do so.
I am especially grateful to my advisor, Jeff Nugent. Throughout my graduate
school life, he has been a wonderful mentor to me. He has constantly been very
encouraging and his words were vital in motivating me to write this piece. Thank you for
guiding me through my Ph.D. and for helping me to bring it to a successful conclusion.
iv
I wish to acknowledge helpful comments from Carol Wise, Cheng Hsiao,
Caroline Betts, Robert Dekle, and John Strauss, whose invaluable time, knowledge and
assistance were crucial ingredients in completing this dissertation. I am very grateful to
Ann Harrison, Johannes Van Biesebroeck, Kei-Mu Yi, Amit Khandelwal, Ana
Fernandes, Devashish Mitra, Pravin Krishna, K. R. Shanmugam, Alberto Trejos,
and Hildegunn Nordas for their extremely useful suggestions. I would also like to thank
Abid Burki for sharing the Balance Sheet Data of Pakistani Listed and Non-Listed
Companies (BSDPC). My work has immensely benefited from my conversations with
Karrar Hussain, Dimitrios Pipinis, Malgorzata Switek, Hongchun Zhao, Saurabh Singhal,
and Arya Gaduh, and I extend my sincere thanks to them.
Last but not the least, special thanks to all my friends, Sana, Farid, Bareera, Sadaf,
Halla, Sandra, Leena, Younoh, Imamah, Zubia, and Ilham, who have shared my joys and
sorrows over the past five years. Their presence, encouragement and words have not only
helped me complete my degree but have also made it a memorable experience. My
heartfelt thanks to Hira Bashir for her belief in me.
v
Table of Contents
Dedication ............................................................................................................................... ii
Acknowledgments ................................................................................................................ iii
List Of Tables ..................................................................................................................... viii
List Of Figures ..................................................................................................................... xi
Abstract ................................................................................................................................ xiv
Chapter One: Introduction ................................................................................................ 1
Chapter Two: Trade liberalization, competition and productivity growth ......... 12
2.1 Review of Literature ....................................................................................... 12
2.2 Trade Liberalization in Pakistan .................................................................... 20
2.2.1 Some History ...................................................................................... 20
2.2.2 Trade Liberalization – Firm Productivity Link .................................. 22
2.3 Empirical Methodology .................................................................................. 25
2.4 Estimation Results .......................................................................................... 29
2.4.1 Random effects and IV estimates with constant returns to scale ....... 30
2.4.2 Random effects estimates with no restriction on returns to scale ...... 33
2.5 Alternative Measure of Productivity .............................................................. 36
2.6 Conclusion ...................................................................................................... 38
vi
Chapter Three: The End of Multi-Fibre Arrangement and Textile Industry..... 41
3.1 The Multi-Fibre Arrangement ........................................................................ 41
3.2 The T&C sector of Pakistan ........................................................................... 44
3.3 Review of Literature ....................................................................................... 56
3.4 Empirical Methodology .................................................................................. 60
3.5 Industry-level results: Effect on T&C exports ............................................... 67
3.6 Firm-level results: Effect on productivity ...................................................... 68
3.6.1 Adjusted level of quotas and productivity .......................................... 70
3.6.2 Variable trade costs and productivity ................................................. 72
3.6.3 Productivity and other variables ......................................................... 73
3.6.4 Discussion .......................................................................................... 76
3.6.5 Robustness check: Alternative measure of productivity .................... 78
3.7 Firm-level results: Effect on output ............................................................... 80
3.8 Conclusion ...................................................................................................... 81
Chapter Four: The Case of Vertically Integrated Clothing Firms ....................... 84
4.1 Vertical integration of production .................................................................. 84
4.2 Review of Literature ....................................................................................... 89
4.3 Theoretical Framework .................................................................................. 93
4.3.1 Firms and Technology ........................................................................ 94
4.3.2 Households ......................................................................................... 95
4.3.3 Market Clearing .................................................................................. 96
4.3.4 Definition of Equilibrium ................................................................... 96
4.4 Empirical Methodology ................................................................................ 101
4.5 Difference between Integrated and Non-integrated Firms ............................ 104
4.6 Production function estimates - Levinsohn & Petrin ................................... 106
4.7 Productivity and Vertical Integration ........................................................... 109
4.7.1 Adjusted quotas and productivity ..................................................... 109
4.7.2 Productivity and other variables ....................................................... 112
4.7.3 Adjusted quotas and productivity - Across phases ........................... 113
4.7.4 Adjusted quotas and productivity – interaction with Size and
Capital intensity ................................................................................ 114
4.8 Adjusted quotas and other dependent variables ........................................... 114
vii
4.9 Discussion .................................................................................................... 116
4.10 Conclusion .................................................................................................... 118
Bibliography ........................................................................................................... 122
Appendices
Appendix A: Supplementary Figures and Tables ................................................. 129
A.1 Figures for Chapter 2......................................................................... 129
A.2 Tables for Chapter 3 ......................................................................... 143
Appendix B: Review of Olley & Pakes and Levinsohn & Petrin ........................ 147
Appendix C: Appendix to Chapter 4 .................................................................... 152
viii
List Of Tables
2.1 Input and Output Tariffs in Pakistan, 1999-2009 ........................................... 19
2.2 Relationship between Tariffs and Firm Productivity ..................................... 24
2.3 Summary Statistics ......................................................................................... 29
2.4 Random Effects Estimates with Constant Returns to Scale ........................... 31
2.5 Instrumental Variables Estimates with Constant Returns to Scale ................ 32
2.6 Random Effects Estimates with no Restriction on Returns to Scale .............. 35
2.7 Modified TFP and Sensitivity of TFP change to assumptions of perfect
competition (µ=1) and constant returns to scale (β=1) ................................. 40
3.1 Exports of Cotton & Cotton Manufactures in Millions of U.S. Dollars ......... 45
3.2 Exports of Cotton & Cotton Manufactures in Millions of U.S. Dollars ......... 46
3.3 Production & Export of Yarn in Thousands of Kilograms ............................. 47
3.4 Production & Export of Cloth in Million Square Meters ................................ 47
3.5 Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates .............. 49
3.6 Summary Statistics ......................................................................................... 65
ix
3.7 Effect of Elimination of Quota-Restrictions on the Level of Imports
(1993-2003 & across Phases 1, 2 and 3) ........................................................ 69
3.8 Production Function Estimates for Textile and Clothing Firms –
Levinsohn & Petrin ........................................................................................ 70
3.9 Effect of Elimination of Quota Restrictions on Textile Firm Productivity –
Levinsohn & Petrin ........................................................................................ 71
3.10 Effect of Elimination of Quota Restrictions on Clothing Firm
Productivity – Levinsohn & Petrin ................................................................ 72
3.11 Effect of Elimination of Quota Restrictions on Firm Productivity –
Across phases ................................................................................................. 73
3.12 Effect of Elimination of Quota Restrictions on Textile and Clothing Firm
Productivity – Olley & Pakes ......................................................................... 79
3.13 Effect of Elimination of Quota Restrictions on Output .................................. 80
3.14 Foreign Direct Investment in Current U.S. Dollars (1971-2010) ................... 82
4.1 Differences between Integrated and Non-integrated Clothing Producers .... 107
4.2 Differences between Integrated and Non-Integrated Clothing Producers ... 107
4.3 Differences between Integrated and Non-integrated Clothing producers -
Controlling for Firm’s Sales ......................................................................... 108
4.4 Production Function Estimates for Vertically Integrated and
Non-Integrated Clothing Firms .................................................................... 108
4.5 Effect of Elimination of Quota-Restrictions on productivity of Clothing
Firms – Levinsohn & Petrin .......................................................................... 109
x
4.6 Effect of Elimination of Quota-Restrictions on productivity of Clothing
Firms – Olley & Pakes and OLS .................................................................. 110
4.7 Effect of Elimination of Quota-Restrictions on productivity of Clothing
Firms – Across Phases .................................................................................. 111
4.8 Effect of Elimination of Quota-Restrictions on productivity of Clothing
Firms – Interaction of VI with Firm’s Size and Capital Intensity ................ 113
4.9 Effect of Elimination of Quota-Restrictions on Net Profit and Output of
Clothing Firms .............................................................................................. 115
A.2.1 Effect of Elimination of Quota-Restrictions on Firm Productivity –
Levinsohn & Petrin ...................................................................................... 143
A.2.2 Effect of Elimination of Quota-Restrictions on Firm Productivity –
Levinsohn & Petrin ...................................................................................... 144
A.2.3 Effect of Elimination of Quota-Restrictions on Firm Productivity –
Olley & Pakes ............................................................................................... 145
A.2.4 Effect of Elimination of Quota-Restrictions on Firm Productivity –
Olley & Pakes ............................................................................................... 146
xi
List Of Figures
1.1 Mean productivity of textile and clothing firms – Olley & Pakes ................ 3
1.2 Mean productivity of textile and clothing firms – Levinsohn & Petrin ........ 4
2.1 Change in Returns to Scale (1995-2003) – Food & Energy ....................... 14
2.2 Change in Plant Size Distribution (1995-2003) – Textile & Automobile ... 15
3.1 Mean productivity of textile and clothing firms – Levinsohn & Petrin
Productivity Measure .................................................................................. 44
3.2 Level of Imports and Adjusted Quota Base (Examples) ............................. 55
4.1 Vertically Integrated and Non-Integrated Clothing Firms – Productivity .. 87
4.2 Vertically Integrated and Non-Integrated Clothing Firms – Output,
Labor & Raw Materials ............................................................................... 88
4.3 Relative factor costs and relative productivities (Firm 3/Firm 2) ............... 99
4.4 Relative factor costs and relative productivities (Firm 3/Firm 2) ............... 99
4.5 Relative Factor Costs and Relative Productivities (Home/Foreign) ......... 100
A.1.1 Change in Returns to Scale (1995-2003) – Sugar ...................................... 129
A.1.2 Change in Returns to Scale (1995-2003) – Textile .................................... 130
xii
A.1.3 Change in Returns to Scale (1995-2003) – Apparel .................................. 130
A.1.4 Change in Returns to Scale (1995-2003) – Paper & Printing .................... 131
A.1.5 Change in Returns to Scale (1995-2003) – Medicine ................................ 131
A.1.6 Change in Returns to Scale (1995-2003) – Petroleum & Oil ..................... 132
A.1.7 Change in Returns to Scale (1995-2003) – Chemicals .............................. 132
A.1.8 Change in Returns to Scale (1995-2003) – Non-Metallic Materials .......... 133
A.1.9 Change in Returns to Scale (1995-2003) – Automobile ............................ 133
A.1.10 Change in Returns to Scale (1995-2003) – Construction Materials .......... 134
A.1.11 Change in Returns to Scale (1995-2003) – Electronics ............................. 134
A.1.12 Change in Returns to Scale (1995-2003) – Miscellaneous ........................ 135
A.1.13 Change in Plant Size Distribution (1995-2003) – Food ............................. 135
A.1.14 Change in Plant Size Distribution (1995-2003) – Sugar ............................ 136
A.1.15 Change in Plant Size Distribution (1995-2003) – Apparel ........................ 136
A.1.16 Change in Plant Size Distribution (1995-2003) – Paper & Printing .......... 137
A.1.17 Change in Plant Size Distribution (1995-2003) – Medicine ...................... 137
A.1.18 Change in Plant Size Distribution (1995-2003) – Personal Care
Products ...................................................................................................... 138
A.1.19 Change in Plant Size Distribution (1995-2003) – Leather ......................... 138
A.1.20 Change in Plant Size Distribution (1995-2003) – Petroleum & Oil .......... 139
xiii
A.1.21 Change in Plant Size Distribution (1995-2003) – Chemicals .................... 139
A.1.22 Change in Plant Size Distribution (1995-2003) – Metal Tools &
Products ...................................................................................................... 140
A.1.23 Change in Plant Size Distribution (1995-2003) – Construction
Materials ..................................................................................................... 140
A.1.24 Change in Plant Size Distribution (1995-2003) – Electronics ................... 141
A.1.25 Change in Plant Size Distribution (1995-2003) – Energy.......................... 141
A.1.26 Change in Plant Size Distribution (1995-2003) – Non-Metallic
Materials ..................................................................................................... 142
A.1.27 Change in Plant Size Distribution (1995-2003) – Miscellaneous .............. 142
xiv
Abstract
It is generally believed that a rise in foreign competition makes the industrial
sector more efficient. By using a novel firm-level data set from a variety of industries in
Pakistan, this dissertation revisits the productivity-liberalization link, and investigates the
effect of trade liberalization on firm productivity. There is evidence of an increase in
competition following trade liberalization in the literature. In a majority of industries,
there is reduction in the returns to scale, indicating the existence of inflexible capacity
constraints in these industries. Moreover, there is no strong evidence of an improvement
in productivity after trade reforms were introduced in the manufacturing sector of
Pakistan.
A greater part of this dissertation focuses on the textile industry of Pakistan, and
unlike most other studies in the literature which mainly investigate the effect of trade
liberalization reforms in developing countries, this dissertation investigates liberalization
episode in a developed country and its consequence for firms in a developing country.
Furthermore, it highlights sectoral heterogeneity within the manufacturing industry in the
effect of trade reforms. In particular, using a sample of 321 textile and clothing
companies for the years 1992 to 2010, we analyze the effect of quota phase-outs in the
form of the end of Multi-Fibre Arrangement (MFA) on firm-level efficiency. The results
differ for the two industries: MFA expiration lead to an increase in the average
xv
productivity of textile producing firms but a significant reduction in the mean
productivity of clothing producers. We offer several possible explanations for this
outcome, such as, a change in input and product mix, entry by non-exporters in clothing
sector, and sectoral differences in quality ladders. Within the clothing industry, we
compare the productivity of vertically integrated and non-integrated firms to investigate
whether or not efficiency gains associated with a given liberalization episode vary across
firms depending on their organization. The interaction between trade policies and firm
characteristics is a subject of great interest to trade economists at present. Vertical
integration is a firm characteristic that has the potential to affect the impact of trade
policy on firms. Nevertheless, there are currently relatively few studies on this topic.
Therefore, this dissertation addresses a potentially critical missing piece in our
understanding of the impact of trade on firms.
A theoretical framework in relation to vertical integration in the clothing industry
shows that liberalization causes a change in the relative factor cost of the two types of
firms, and consequently, a change in the product range produced by them. One of the
difficulties in comprehending the connection between vertical integration and trade is a
lack of data. The most innovative facet of this dissertation is to present a data set that
includes information on the level of integration within firms, and to merge these data
with a natural experiment resulting from the end of the MFA, in order to explore the
differential impact of trade liberalization on vertically integrated versus non-integrated
firms. This appears to be a fruitful way in which to deal with this subject. The empirical
findings illustrate that a higher trade quota, which represents fewer trade barriers, reduces
xvi
the mean productivity for all clothing firms significantly, though less so for vertically
integrated firms than for non-integrated firms. The greater decline in the efficiency of
non-integrated clothing firms points to the inability of these firms to benefit from tighter
quality control, timely revision of production policies, and greater stability of supplies.
1
Chapter 1
Introduction
Trade creates jobs and lifts people out of poverty. And when that happens, societies stabilize and
grow. And there is nothing like a stable society to fight terrorism and strengthen democracy,
freedom and rule of law.
– Dennis Hastert
During the past few decades, many developing countries turned decisively away
from inward looking policies towards trade liberalization and export promotion. What is
the case for trade liberalization? An early theoretical case was based on the well-known
neoclassical result of Pareto optimality of free trade. In the new trade theory, welfare
gains from trade accrue from economies of scale and expansion of product varieties
available to consumers (Bernard, Jensen, Schott, & Redding, 2007). It is typically
believed that an upsurge in foreign competition makes the industrial sector significantly
more efficient. A large number of empirical analyses at the firm level offer evidence for
aggregate productivity growth driven by trade liberalization. Do efficiency gains
associated with a given liberalization episode vary across different types of firms? This
dissertation aims to study firm-level panel data for evidence that supports this perception.
2
In particular, we analyze the experience of firms in the manufacturing sector of Pakistan
under the trade liberalization reforms introduced over the past two decades. Pakistan has
set out on programs of trade and financial liberalization. There has been a massive
reduction in tariff rates and duties from 2001 to date. Successive governments
consistently strengthened these policies in the subsequent period. Tariffs were reduced to
low levels and import license requirements for many products were eliminated. This
study utilizes an exceptional database which traces annual reports of a representative
sample of manufacturing companies in Pakistan. This source of data is combined with an
exceptional database that traces U.S. trading partners’ performance under the quota
regimes determined by the Multi-Fibre Arrangement (1974-95) and the Agreement on
Textile and Clothing (1995-2005), initially used by Brambilla, Khandelwal, and Schott
(2007). Hence, the dissertation merges micro-level data of firms with the data on quotas
at the industry level in order to answer an essential question which has been the center of
debate in the new trade theory.
Figure 1.1 illustrates the evolution of mean productivity of textile and clothing
firms in our sample. The clothing firms include both vertically integrated as well as non-
integrated firms. Productivity is computed using Olley & Pakes (OP) productivity
measure (which we explain later in detail). For the time period under consideration,
textile firms have a much higher mean productivity than non-integrated clothing firms.
Furthermore, we notice that over the entire sample period, vertically integrated clothing
producers are much more productive than non-integrated clothing producers. Whereas the
average productivity of vertically-integrated clothing firms exhibits an upward trend, we
3
do not see an analogous pattern for non-integrated clothing firms. Instead, the average
productivity of non-integrated clothing firms remains roughly at the same level as at the
start of the period. Figure 1.2 shows mean productivity of the three types of firms
computed using Levinsohn and Petrin (LP) productivity measure. Since there is no
noteworthy change in the observations made earlier, we can claim that our results are not
sensitive to the measure of productivity used.
Figure 1.1: Mean productivity of textile and clothing firms – Olley & Pakes
There are many linkages between trade policy and efficiency gains. Empirical
analyses at the firm level put forward evidence for aggregate productivity growth driven
by the contraction and exit of low-productivity firms and the expansion and entry of high
productivity firms. Increase in competition leads to lower mark-ups of price over
marginal cost. Intra-firm productivity gains will perhaps go together with trade
-2 0 2 4 6 8
Mean productivity
1990 1995 2000 2005 2010
Year
Textile firms
Non-integrated clothing firms
Vertically-integrated clothing firms
Olley & Pakes
4
liberalization if it enlarges the menu of intermediate inputs accessible to domestic firms.
Trade also serves as a channel for technology diffusion. The net effect of liberalization on
productivity depends upon the specifics of the demand shifts that accompany
liberalization, ease of entry or exit, and the nature of competition (Tybout, 1992). Trade
reforms affect the tightness of the link between domestic and world markets, and create
speculation about their own sustainability. Efficient adjustments in capacity are possible
only when trade reforms institute a credible regime.
Arguments derived from increasing
returns are widespread in the literature, but it has been shown that scale economies can
cut both ways (Krugman, 1986).
Figure 1.2: Mean productivity of textile and clothing firms – Levinsohn & Petrin
In chapter 2 of the dissertation, we concisely go over trade reforms introduced in
Pakistan over the past few years. We will then present a simple methodology to link trade
0 5 10 15 20
Mean productivity
1990 1995 2000 2005 2010
Year
Textile firms
Non-integrated clothing firms
Vertically-integrated clothing firms
Levinsohn & Petrin
5
openness and firm productivity. We demonstrate that productivity gains from reducing
overall tariff rates are not obvious from a simple regression of TFP on tariff rates;
reducing input tariffs may perhaps compensate for some of the import competition effects
that occur from lower output tariffs since many firms are affected by both output and
input tariffs (Corden, 1971). If there are any gains from reducing tariffs on intermediate
goods, they are to a great extent counterbalanced by reduction in tariffs on final goods.
Moreover, the productivity gains are much higher from reducing input tariffs than those
from reducing output tariffs.
In the next section, we use econometric methodology of Harrison (1994), who
extended the methodology of Hall (1988) and Domowitz, Hubbard, and Petersen (1988),
in order to measure productivity using a technique that deviates from the assumptions of
perfect competition and constant returns to scale. Hence, we are able to capture the
change in firm’s productivity growth, returns to scale and markup following trade
liberalization in Pakistan. The results from the model imply that there is a rise in
competition following trade liberalization. However, for most of the industries, there is
also a reduction in returns to scale. The reduction in returns to scale may be caused by an
increased exploitation of returns to scale by firms which may have been operating at too
small a scale before the reforms. If the degree of this change is relatively large, it may be
indicative of the existence of rather inflexible capacity constraints in these industries. As
far as the impact of trade openness on firm-level productivity is concerned, there is no
strong evidence of an improvement in productivity after trade reforms were introduced in
the manufacturing sectors of Pakistan. Nonetheless, if returns to scale decrease following
6
liberalization, the assumption of constant returns to scale biases estimates of the growth
rate of productivity upward before liberalization and biases the post-liberalization
estimates downward (Harrison, 1994). Thus, the estimates of change in the growth rate of
productivity post-liberalization are biased downward. Therefore, the main contribution of
chapter 2 is that it uses a novel data set from a developing country in order to revisit an
essential question that has been at the centre of productivity-liberalization debate in the
new trade theory. It offers evidence for the perception that gains from trade liberalization
in the form of improvement in firm performance have been greatly exaggerated.
Moreover, since our data set encompasses a wide range of manufacturing industries, we
are able to demonstrate cross-industry heterogeneity in the effect of trade reforms.
Chapter 3 of the dissertation analyzes the impact of the end of MFA on
productivity of T&C firms. An important development of the Uruguay Round in 1994
was to put an end to the MFA. The MFA imposed restrictions on Textile and Clothing
(T&C) exports to the developed countries by means of bilaterally negotiated quotas on
textile products. The end of the quota system, together with increasing significance of the
industry in its domestic market, leads us to analyze the efficiency issues related to
Pakistan’s textile industry. Accordingly, contrasting other studies in the literature which
primarily examine the effect of trade liberalization reforms in developing countries, this
chapter investigates liberalization episode in a developed country and its consequence for
firms in a developing country. In addition, it highlights sectoral heterogeneity within the
textile manufacturing industry in the effect of MFA expiration.
7
The textile industry consists of various complex processes, starting from the
cultivation of cotton to bleaching, dyeing and printing of the fabric. It primarily consists
of three stages, namely, spinning, weaving and finishing. It is largely based on the
production of yarn using natural or synthetic fibres, and the fabrication of yarn into cloth.
On the other hand, the clothing (or apparel) industry mainly involves the use of fabric for
the production of garments and other products, such as, bed sheets, curtains, and so forth.
In many studies and economic reports, the two industries are lumped together and are
simply referred to as the textile industry. Since one of the objectives of this dissertation is
to highlight the difference in outcomes across the textile and clothing industries, it is
imperative to distinguish between the two.
Using a sample of 321 T&C companies for the years 1992 to 2010, we observe
that in spite of an increase in the adjusted level of quotas taking place simultaneously for
a group of competing developing countries, there is an increase in the level of imports in
the T&C industry of Pakistan. We then analyze the effect of the phasing out of quotas on
firm-level efficiency. In the first step, plant-level productivity is estimated, and then the
change in firm productivity is regressed on the adjusted level of quotas. We use structural
techniques proposed by Olley & Pakes and Levinsohn & Petrin in order to take care of
endogeneity in the estimation of a production function. The results differ for the two
industries: MFA expiration lead to an increase in the average productivity of textile
producing firms but a significant reduction in the mean productivity of clothing
producers. The paper offers several possible explanations as to why this might be the
case, such as, a change in input and product mix, entry by non-exporters in clothing
8
sector, change in the composition of T&C exports, and sectoral differences in quality
ladders. Finally, in order to measure the effect of quotas directly on firm’s output, we
regress output on the adjusted level of quotas and trade costs. In the textile sector, an
increase in the adjusted level of quotas leads to a significant rise in firm’s output.
Nevertheless, this result is not statistically significant for the clothing sector.
In short, the most important contribution of chapter 3 is that it is one of the very
few studies that investigate the effect of liberalization in the form of the phasing out of
quotas on firm-level productivity in textile and clothing industry. Unlike most other
studies in the literature which mainly analyze the impact of trade liberalization in a
developing country, this chapter investigates liberalization episode in a developed
country. It underlines cross-sector disparity in the effect of MFA expiration and that trade
reforms may influence different sectors heterogeneously even within the manufacturing
industry. Obviously, the topic is of concern both in general and for the context of
Pakistan. The textile and clothing industries are vital in many developing countries,
including Pakistan, and the ATC was one of the key negotiated trade reforms for
developing countries in the past 30 years. Consequently, the central question examining
the relationship between these quota phase-outs and firm output and productivity, is of
great significance.
In chapter 4 of the dissertation, we compare the productivity of vertically
integrated and non-integrated firms within a country, allowing both types of firms to
engage in international trade, and analyze how trade liberalization affects efficiency of
these firms differently. In particular, we analyze the experience of vertically integrated
9
and non-integrated clothing firms in Pakistan under the U.S. textile and clothing quotas
and the succeeding end of MFA. The chapter offers a number of interesting findings. A
theoretical framework regarding vertical integration in the clothing industry shows that
liberalization causes a change in the relative factor cost of the two types of firms, and
therefore, a change in the product range produced by each of them. Liberalization in the
home country results in an increase in the product range produced by vertically integrated
firms as well as a rise in the country-wide product range produced. The empirical
findings demonstrate that a higher trade quota, which represents fewer trade barriers,
reduces the mean productivity for all clothing firms significantly, though less so for
vertically integrated firms than for non-integrated firms. This is in contrast to the
conventional wisdom according to which the T&C industry of Pakistan was expected to
perform much better in the quota free regime given that it was apparently constrained by
MFA quotas. The bigger drop in the efficiency of non-integrated clothing firms points to
the inability of these firms to benefit from decreased marketing expenses, stability of
operations, tighter quality control, timely revision of production policies, and guarantee
of supplies. The difference in results across garment firms may possibly be related to the
structure of production, namely, the type of raw materials used by the garment firms after
the end of MFA. Thus, the most important contribution of this chapter is that it highlights
variation in the effect of MFA expiration across firms depending on their organization.
Although a variety of studies look into the efficiency of vertically integrated firms
relative to non-integrated firms, none of these in particular consider the effect of trade
liberalization caused by the phasing out of quotas on firm-level efficiency of these two
10
types of firms. As is frequently emphasized in the new trade theory literature, trade
reforms often influence firms heterogeneously. By merging micro-level data of firms
with data on quotas at the industry level, this study shows that a liberalization episode
may generate divergent changes in productivity levels of different categories of firms
even within the clothing industry.
The abolition of quotas has been the most significant event in the global textile
industry during the past two decades. The textile sector is a key industry in Pakistan. In
conjunction with the cost advantage in terms of nearness to a raw material base in cotton
and man-made fibres, above and beyond the availability of cheap labor, what appears to
be a critical determinant of competitiveness in this industry is the capacity to respond to
changing consumer demands. This requires greater investment in R&D to facilitate
greater flexibility of the production process. The need to invest in cost-saving production
methods plays a bigger role in the clothing industry owing to the characteristics of the
finished good. The conclusion that mean productivity diminished for the clothing firms as
a result of the phasing out of quotas, points to the failure of these firms to shift to a more
efficient input-mix in response to a more competitive world market.
The dissertation is organized in the following manner: the next chapter briefly
discusses trade reforms introduced in Pakistan over the past years, and describes a
methodology that can be used to measure the effect of liberalization on firm efficiency,
returns to scale and competitive behavior. This is followed by chapter 3, which analyzes
the impact of the end of MFA on productivity of T&C firms, followed by the discussion
of results for the empirical model. The last section concludes chapter 3. This leads to
11
chapter 4, which starts by presenting a theoretical discussion based on the model of
vertical integration of Yi (2003), and shows that liberalization causes a change in the
relative factor cost of the two types of firms, and consequently, a change in the product
range produced by each of them. This is followed by a comparison of the productivity of
vertically integrated and non-integrated firms under the U.S. textile and clothing quotas,
and analyzing how trade liberalization affects efficiency of these firms differently. The
last section concludes chapter 4.
12
Chapter 2
Trade liberalization, competition and productivity growth
This is the moment when we must build on the wealth that open markets have created, and share its
benefits more equitably. Trade has been a cornerstone of our growth and global development. But
we will not be able to sustain this growth if it favors the few, and not the many.
– Barack Obama
2.1 Review of Literature
There are many linkages between commercial trade policy and efficiency gains.
Empirical analyses at the firm level offer evidence for aggregate productivity growth
driven by the exit of low-productivity firms and entry of high productivity firms. This
reallocation of resources lifts average industry productivity.
1
Pavcnik (2002) finds that
approximately two-thirds of the 19 percent increase in aggregate productivity following
Chile’s trade liberalization of the late 1970s is because of the relatively greater survival
and growth of high-productivity plants. Increase in product market competition also leads
1
Analogous findings appear from a number of studies of trade liberalization reforms in developing
countries as surveyed in Tybout (2003). The within-industry reallocations of resources established by these
studies take over the across-industry reallocations of resources stressed by old theories of comparative
advantage.
13
to lower mark-ups of price over marginal cost. Together, the fall in mark-ups and rise in
average productivity contribute to lower prices and higher real incomes. Nevertheless, a
varied body of theory puts forward that the direction of change in efficiency hinges
decisively upon model specifics (Goh, 2000; Miyagawa & Ohno, 1995; Rodrik, 1992).
Further effects on intra-firm productivity are somewhat more robust. Intra-firm
productivity gains will possibly go together with trade liberalization if it enlarges the
menu of intermediate inputs accessible to domestic firms. Amiti and Konings (2007) use
Indonesian manufacturing census data from 1991 to 2001 to compare productivity gains
from reducing tariffs on final goods and intermediate inputs. They show that a reduction
in input tariffs leads to productivity gains twice as high as gains from reducing output
tariffs.
International trade may perhaps also serve as a channel for disembodied
technology diffusion if firms learn about products by observing imported varieties, or by
exporting to buyers who offer them blueprints and technical assistance (Grossman &
Helpman, 1991). Baldwin (2005) studies the impact of greater openness at the firm and
aggregate level, focusing on changes in number and type of firms, trade volumes and
prices, and productivity effects. It sets out a basic heterogeneous-firms trade model that is
very much resembling Melitz (2003). The best-known argument linking trade regimes
and productivity is that the returns to entrepreneurial effort amplify with exposure to
foreign competition (Corden, 1974). Another study by Krueger and Tuncer (1982)
estimates the rates of total factor productivity (TFP) growth for two-digit manufacturing
industries in Turkey during 1963-1976. The paper shows that periods of slower
14
productivity growth coincided with periods of stringent trade regimes. These findings are
not confined to developing countries. The effects of a reduction in U.S. trade costs are
inspected by Bernard, Jensen, and Schott (2006).
Figure 2.1: Change in Returns to Scale (1995-2003) – Food & Energy
.5 1 1.5
Ln(average cost)
14 16 18 20 22
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
FOOD
0 .5 1 1.5 2
Ln(average cost)
15 20 25
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
ENERGY
15
Figure 2.2: Change in Plant Size Distribution (1995-2003) – Textile & Automobile
The net effect of liberalization on productivity depends upon the specifics of the
demand shifts that accompany liberalization, ease of entry or exit, and the nature of
competition (Tybout, 1992). Trade reforms also influence the tightness of the link
0 .1 .2 .3 .4 .5
kernel density
14 16 18 20 22 24
Ln(output)
1995 2003
TEXTILE
.05 .1 .15 .2
kernel density
14 16 18 20 22 24
Ln(output)
1995 2003
AUTOMOBILE
16
between domestic and world markets and create speculation about their own
sustainability. These uncertainty effects can affect productivity (Tybout, 1992). Likewise,
when substitution possibilities exist, managers may choose labor-intensive technologies,
albeit capital-intensive technologies would be less expensive to operate if market
conditions were stable (Lambson, 1989). Efficient adjustments in capacity are possible
only when trade reforms institute a credible regime.
Arguments derived from increasing
returns are common in the literature, but it has been shown that scale economies can cut
both ways (Krugman, 1986; Rodrik, 1992; Roberts & Tybout, 1991).
2
Comparable to the
technique employed by Tybout and Westbrook (1995) in their study on Mexico, Figure
2.1 illustrates average cost curves for two of the industries in our sample over the years
before and after trade liberalization reforms were introduced in Pakistan. Similarly, we
plot average cost curves for the other industries in our data. Only in two industries do we
notice an improvement in returns to scale after liberalization: Food and Automobile.
There seems to be no change in the scale parameter for four industries, and in four other
industries, there is actually a decrease in returns to scale after the reforms.
3
Tybout, Melo, and Corbo (1991) examine changes in industrial sector
performance that accompanied Chile's trade liberalization of the 1970s. A comparison of
pre- and post-liberalization manufacturing data discloses only a modest productivity
improvement on the whole. Conversely, unfavorable macroeconomic conditions might
have masked the affirmative effects of trade reforms. There is some evidence that as
2
Competition from foreign producers can force producers to expand or exit.
3
We show here the average cost curves for only two industries; figures displaying the change in returns to
scale for the remaining industries are presented in Appendix A.
17
protection falls, small plants increase production levels toward minimum efficient scale.
Since labor use falls and output rises as protection is reduced, it is expected that value-
added per efficiency worker and output per efficiency worker both go up.
4
In another
study, Tybout and Westbrook (1995) analyze the impact of Mexico’s trade liberalization
on productivity gains.
5
They find that average costs fell in most industries, with tradeable
goods producers registering the largest reductions. Gains owing to scale economy
exploitation were minor. Figure 2.2 demonstrates the change in firm size distributions for
the textile and automobile industries over the years 1995 and 2003, smoothed using a
bivariate kernel estimator, once again similar to the way Tybout and Westbrook (1995)
illustrate in their study about the Mexican trade liberalization. Likewise, we plot firm size
distributions for the other industries. In thirteen out of twenty industries in our data, we
see an adjustment in plant size distribution that implies scale efficiency gains.
6
Many analysts believed that trade policy openness and higher ratios of trade
volumes were positively correlated with economic growth until Rodriguez and Rodrik
(2000) raised some concerns about the robustness of these results as conclusions
remained sensitive to difficulties in measuring openness, statistically sensitive
specifications and collinearity of protectionist policies with other poorly executed
policies in developing economies. An apprehension is that the link from increased trade
4
This result is in line with the X-efficiency arguments linking trade exposure and productivity.
5
In June 1985, in spite of balance of payments difficulties, Mexico broke away from its inward-looking
development strategies to undertake a remarkable trade liberalization program. Licensing requirements
were significantly scaled back, reference prices were progressively removed, and tariff rates on most
products were reduced (Tybout & Westbrook, 1995).
6
Again, we show here figures for only two industries.
18
to the expansion of higher-productivity firms in developing country results might not be
driven exclusively by changes in trade policy, given that trade liberalization is frequently
part of a broader package of economic reforms. Conversely, comparable patterns of
productivity gains have been found in response to a decline in trade barriers in Canada
and the United States (Trefler, 2004).
7
Trade policy is subject to potential endogeneity
since the government may raise current protection as a reply to lobbying by firms in less
productive industries.
A comparison of adjustment patterns between different three-digit
industries is likely to expose something about the effects of liberalization given that all
industries were exposed to essentially similar measurement errors and changes in
macroeconomic conditions but different industries underwent remarkably different
changes in protection. There is a literature that argues that a selection of industries have
political power to lobby governments for protection (Grossman & Helpman, 1994).
Mobarak and Purbasari (2006) find that political connections do not affect tariff rates in
Indonesia: it is hard for governments in developing countries to offer favors since they
are under the close scrutiny of international organizations.
8
The potential bias is also
diminished as the estimates include fixed effects. If time-varying industry characteristics
could at the same time affect both productivity and tariffs, the bias may persist. Just like
Goldberg and Pavcnik (2005), they use the 1991 levels of tariffs as instruments for
7
The effects of a reduction in U.S. trade costs are inspected by Bernard et al. (2006). Their basic
explanatory variable is a measure of trade costs, including tariff rates and shipping costs at the industry
level. They discover that reduction in trade costs has the utmost influence on plant death for the lowest-
productivity plants.
8
Mobarak, A. M., & Purbasari, D. (2006). Corrupt Protection for Sale to Firms: Evidence from Indonesia.
Unpublished.
19
Table 2.1: Input and Output Tariffs in Pakistan, 1999-2009
Industry Tariff type 1999 2000 2001 2002 2004 2005 2006 2007 2008 2009
Food Input 26.4 26.3 22.4 19.9 18.5 16.3 16.2 15.5 15.5 15.5
Output 26.4 26.6 23.7 21.8 21.4 17.5 17.1 16.8 19.4 19.4
Sugar Input 26.4 26.3 22.3 19.9 18.5 16.2 16.1 15.4 15.4 15.4
Output 32.5 30.9 25.0 22.5 22.3 13.3 13.3 14.1 16.9 16.9
Tobacco Input 26.5 26.4 22.4 20.0 18.5 16.3 16.1 15.5 15.5 15.5
Output 35.0 35.0 30.0 25.0 25.0 18.3 18.3 18.3 25.0 25.0
Leather Input 25.6 25.4 22.1 19.3 17.9 15.7 15.5 14.6 14.6 14.6
Output 19.5 19.7 15.8 14.6 14.6 13.3 13.3 10.5 10.5 10.5
Petroleum Input 26.6 26.5 22.5 20.1 18.6 16.3 16.2 15.6 15.7 15.7
and Oil Output 17.4 17.4 13.9 13.1 12.8 10.9 10.1 6.9 5.4 5.4
Chemicals Input 23.7 23.6 19.8 17.7 16.7 14.6 14.3 12.8 12.4 12.4
Output 19.3 19.1 22.6 15.7 14.0 10.9 10.8 8.6 8.6 8.6
Non-metallic Input 25.9 25.8 21.9 19.6 18.2 15.9 15.7 14.9 14.9 14.9
Materials Output 26.7 26.6 22.6 20.2 18.7 16.4 16.3 15.7 15.8 15.8
Metal Tools and Input 26.2 26.1 22.2 19.7 18.4 16.2 16.1 15.2 15.3 15.3
Products Output 22.7 22.5 18.6 15.0 14.1 13.2 12.9 11.9 11.9 11.9
Automobile Input 26.1 25.9 22.3 19.7 18.2 15.9 15.8 15.1 15.1 15.1
Output 59.5 58.8 44.4 40.8 37.4 33.7 30.7 30.2 30.0 30.1
Furniture Input 24.4 24.4 20.5 18.4 17.2 15.0 14.8 13.6 13.3 13.3
Output 23.9 23.9 18.4 16.4 15.7 12.9 12.9 9.8 9.8 9.8
Construction Input 20.5 20.5 16.8 15.2 14.6 12.8 12.3 9.9 9.0 9.0
Materials Output 32.2 32.1 26.2 23.1 22.7 22.6 22.6 21.3 22.5 22.5
Iron and Steel Input 25.4 25.3 21.5 19.2 17.9 15.6 15.4 14.5 14.4 14.4
Output 29.4 30.1 22.0 18.2 17.4 14.7 14.2 13.5 13.8 13.8
Electronics Input 26.2 26.1 22.1 19.8 18.4 16.2 16.0 15.3 15.3 15.3
Output 21.5 19.7 19.4 17.1 16.5 15.5 14.8 13.9 14.6 14.6
Energy Input 26.4 26.3 22.3 20.0 18.5 16.2 16.1 15.4 15.5 15.5
Output 17.4 17.4 13.9 13.1 12.8 10.9 10.1 6.9 5.4 5.4
Textile Input 25.0 24.9 21.2 18.8 17.6 15.4 15.1 14.1 13.9 13.9
Output 29.8 29.8 24.1 19.6 19.3 14.9 15.6 15.1 15.1 15.1
Apparel Input 25.2 25.1 21.8 19.1 17.7 15.5 15.3 14.3 14.3 14.3
Output 35.0 35.0 30.0 25.0 25.0 24.8 24.8 24.8 24.8 24.8
Paper and Input 22.9 22.8 19.8 17.3 16.1 14.0 13.7 12.1 11.8 11.8
Printing Output 29.1 28.9 24.1 20.7 20.5 19.7 19.7 18.2 18.3 18.3
Medicine Input 26.4 26.3 22.5 19.8 18.5 16.5 16.3 15.4 15.5 15.5
Output 14.1 13.9 12.3 12.4 12.2 11.1 11.1 11.8 11.8 11.8
Personal Care Input 25.4 25.3 21.7 19.2 17.9 15.7 15.5 14.5 14.4 14.4
Products Output 29.4 28.8 24.9 21.9 20.5 17.0 16.1 16.6 20.6 20.6
Agri. & Ind. Input 26.3 26.2 22.3 19.8 18.4 16.1 16.0 15.3 15.3 15.3
Machinery Output 17.7 17.5 15.5 12.5 11.2 9.3 8.6 8.4 8.7 8.7
Miscellaneous Input 26.2 26.1 22.3 19.8 18.4 16.1 16.0 15.2 15.3 15.3
Output 25.0 24.2 18.1 15.9 14.7 12.4 12.8 12.6 13.4 13.4
All Industries Input 25.4 25.3 21.6 19.2 17.8 15.6 15.5 14.5 14.4 14.4
Output 24.9 24.7 20.1 17.4 16.6 14.2 14.0 13.4 13.9 13.9
4otes: The tariff data is taken from the World Trade Organization’s Integrated Data Base (IDB). This database contains
complete information on Most-Favored-Nation (MFN) applied and bound tariffs at the standard Harmonized System
subheading level for all WTO members.
20
changes in tariffs.
9
In this chapter, we use fixed and random effect models as well as
instrumental variables in order to take care of potential bias arising due to endogeneity.
We track a single country through time, eliminating the obscuring country-specific
effects. There is a remarkable variation in trade regimes between sample years which
gives the analysis a laboratory-like flavor, and creates sufficiently great changes in
industrial performance to be detectible.
2.2 Trade Liberalization in Pakistan
In this section, we concisely go over the trade reforms introduced in Pakistan over the
past few years. We then present a methodology to link trade openness and firm
productivity. These estimation results will be used as a basis for the empirical work that
follows.
2.2.1 Some History
Pakistan has been open since 2001. It was closed based on the OPEN90-99 criteria for the
decade due to its tariff barriers (Wacziarg, 2001) which averaged 55% between 1990 and
1999 (World Bank, 2006).
10
In 1988, the initiation of the Structural Adjustment Program
9
The instruments that they use are: 1991 levels of output tariffs, 1991 levels of input tariffs, an interaction
between the 1991 input tariffs and a firm-level indicator equal to one if the firm was an importer in all
years, a dummy indicator for product codes that consisted of at least one nine-digit HS code that was barred
from the commitment to cut bound tariffs to 40 percent, and the share of skilled workers at the five-digit
industry level.
10
Wacziarg (2001) attempted the measurement of a liberalization variable as Sachs and Warner (1995)
classification posed problems on their categorization of open and closed economies (Wacziarg & Welch,
2008). Sachs-Warner dummy for openness classifies an economy as closed if it is closed according to any
one of the following five criteria: (i) its average tariff rate exceeded 40%, (ii) its non-tariff barriers covered
more than 40% of imports, (iii) it had a socialist economic system, (iv) it had a state monopoly of major
21
under an agreement with the IMF introduced a number of reforms. By 1990, significant
measures toward privatization, deregulation and trade liberalization of goods and capital
accounts had been introduced. Pakistan has gradually liberalized its tariffs, reducing the
maximum tariff rate from 225 percent in 1986-87 to 65% in 1995-96. Until the end of
1997, however, the country’s complex and exemption ridden tariff regime continued to
give rise to high levels of effective protection, nurtured inefficient industries, generated a
strong anti-export bias, put a heavy burden on consumers, diverted resources to rent-
seeking activities, and encouraged corruption and smuggling (Wacziarg, 2001). The main
purpose of these protectionist policies was to protect infant industries against foreign
competition. These policies led to wasteful use of resources by encouraging import
substitution. In 1999, Pakistan launched the Economic Revival Program. It has liberalized
its export regime and appreciably reduced its monopoly on exports. At the end of 2001,
Pakistan undertook a major restructuring of its customs tariff, thereby decreasing its
average tariff to 20.4% from 56% in 1993-94 (WTO, 2002). Over the years 2002-06, the
country turned decisively away from inward looking policies towards trade liberalization
and export promotion. At the end of 2004, protection rates had fallen tremendously, and
from 2005 onwards trade was liberalized at an accelerating pace. Successive governments
consistently strengthened these policies in the subsequent period. Pressure from the IMF
and the World Bank group resulted in the controls built up in the earlier years to be
relaxed. The precise turning point is difficult to detect but is most likely to be 2001-02,
when both taxes on trade were lowered and the import approval system began to pull
exports, or (v) its black-market premium exceeded 20% during either the decade of the 1970s or the decade
of the 1980s.
22
apart (Wacziarg, 2001). Change was not instantaneous however; at the beginning of
2004, there was still substantial variation in effective protection.
2.2.2 Trade Liberalization – Firm Productivity Link
To determine the effect of trade liberalization on firm performance, we first need to find a
measure of productivity for the firms in our sample. This measure is then related to an
index of openness using a simple regression equation. Consider a firm with a Cobb-
Douglas production function:
=
, (2.1)
where output in firm i at time t for industry j,
, is a function of labor,
, capital,
, and materials,
. Productivity of firm i is a function of trade policy, denoted by τ.
In the first step, we compute firm’s total factor productivity (TFP), and in the second
step, we identify how productivity is influenced by trade policy. Taking natural logs
denoted by small letters, we can estimate:
=
+
+
+
+
. (2.2)
Total revenue at the firm level,
, is deflated by two-digit industry-level producer price
indices. Although domestic and imported inputs should be adjusted by separate deflators,
the balance sheet data does not provide information on the share of imported inputs.
Consequently all material inputs are also deflated with a two-digit producer price
23
deflator.
11
The production function coefficients for firms in each sector are estimated
separately. These estimates are used to compute the log of measured TFP ( !
) of firm
i at time t for each industry j:
!
=
−
#
−
#
−
#
. (2.3)
In addition, we require measures of output and input tariffs. Following Amiti and
Konings (2007), the output tariff is an average constructed at the two-digit ISIC industry
j. The input tariff for each five-digit industry k is a weighted average of output tariffs,
whereby, the weights are cost shares of each input used:
$%!& '($
= ) *
+,,-
× /&!& '($
,
*ℎ( *
+,,-
=
1 $%!&
+,,-
1 $%!&
+,,-
∙
The weights, *
+,,-
, are the cost shares of industry j in the production of a good in
industry k, based on firm-level data in 1992. These input tariffs are constructed at the
industry level and the cost shares are derived from total input purchases. If the weights
included just the imported inputs, this would bring in endogeneity bias.
12
These input and
11
Amiti and Konings (2007) confirm that domestic and imported input prices normally move together,
provided they are substitutes. Their results are robust to deflating both domestic and imported material
inputs by the same five-digit domestic materials deflators.
12
The balance sheet data does not provide enough information to identify imports and exports by firms in
our sample. In order to derive the weights, the Census of Manufacturing Industries data set is used, in view
of the fact that it provides detailed information on input use as well as whether or not these inputs were
imported from abroad. Since the tariff rates are averages constructed at the two-digit ISIC industry-level, it
seems reasonable to do so.
24
output tariffs are shown in Table 2.1. To measure productivity gains from reducing tariffs
on final goods and intermediate inputs, firm productivity computed using the Cobb-
Douglas production function stated above is regressed on input and output tariff rates
shown in Table 2.1 using OLS:
!
= 3
+ 3
+ 3
+
/&!& '($
+ 3
-
$%!& '($
+ 4
, (2.4)
where 3
is added to control for firm-level heterogeneity. A negative 3
+
(3
-
)
would imply
that a reduction in output (input) tariff raises firm’s productivity. A drop in the price of
imported inputs can compel domestic producers of substitutes to become more
competitive. Cheaper imported inputs can increase productivity via learning, variety, and
quality effects. The tariff data is taken from the World Trade Organization’s Integrated
Data Base (IDB). This database contains complete information on Most-Favored-Nation
(MFN) applied and bound tariffs at the standard Harmonized System subheading level for
all WTO members.
13
Table 2.2: Relationship between Tariffs and Firm Productivity
Dependent variable: ln(tfp
ijt
) (1) (2) (3)
Input tariff
k
t
-0.1914** -0.9757***
(0.1078) (0.2744)
Output tariff
k
t
-0.0673 0.7300***
(0.1012) (0.2630)
Firm fixed effects Yes Yes Yes
Observations 1449 1417 1417
R-squared 0.0066 0.0108 0.0023
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. *** Significant at, or below, 1
percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
13
In addition, it provides data on non-MFN applied tariff regimes which a country grants to its export
partners. This tariff information is sourced from submissions made to IDB for applied tariffs and imports,
and from the Consolidated Tariff Schedules database for bound duties.
25
The estimation results are shown in Table 2.2. Column 1 shows that a drop in
input tariffs of 10 percentage points increases productivity by 1.9 percent. The effect of
output tariff is much smaller but is not significant. If both types of tariffs are included, the
coefficient on input tariffs becomes 9.8 percent and is highly significant. The coefficient
on output tariff, however, becomes significant and positive: a reduction in tariffs on
finished goods actually lowers firm’s productivity. These results show that productivity
gains from reducing overall tariff rates are not very obvious from a simple regression of
TFP on tariff rates. Reducing input tariffs may perhaps compensate for some of the
import competition effects that occur from lower output tariffs since many firms are
affected by both output and input tariffs (Corden, 1971). If there are any gains from
reducing tariffs on intermediate goods, they are to a great extent counterbalanced by
reduction in tariffs on final goods. Also, productivity gains are much higher from
reducing input tariffs than those from reducing output tariffs.
2.3 Empirical Methodology
In this section, we use the econometric methodology of Harrison (1994), who extended
the methodology of Hall (1988) and Domowitz et al. (1988), in order to measure
productivity using a technique that allows for a change in competitive environment as
well as firm’s scale of production as a result of trade liberalization. The production
function for firm i in industry j at time t is given by:
=
5 6
,
,
7 , (2.5)
26
where output,
, is produced using labor,
, capital,
, and materials,
.
is
an industry-specific index of technical progress and
is a firm-specific parameter which
takes into account firm-specific differences in technology. If we assume cournot behavior
by firms, and that the markup only varies across sectors, the first order conditions from
firm’s profit maximization problem and total differentiation of Eq. (2.5) yield:
89
9
:;<
= =
> ?
@
8A
A
B + ?
@
8C
C
B D
+ =
?
@
8E
E
B
+
8F
F
;<
+
8G
:<
G
:<
, (2.6)
where =
is the markup, and ?
, ?
and ?
are shares of labor, capital and materials in
output, respectively. The share of capital in output is unobservable. The sum of factor
shares can be expressed as /= , where may be greater or less than one. In the case of
constant returns, would be equal to 1. Using lower case y, l and m to denote ln(Y/K),
ln(L/K) and ln(M/K), Eq. (2.6) can be written as:
I
= =
[?
I + ?
I]
+ − 1
I/
+ I/
+ I
/
. (2.7)
The final estimating equation uses an interactive slope dummy to account for changes in
competitive behavior following trade reforms in Pakistan. An intercept dummy is also
included to allow for changes in the overall productivity growth by firms. Just as in
Krishna and Mitra (1998), we will also use an interactive dummy to allow the returns to
scale to change with the trade reforms:
I
=
+
M IN
O +
-
M PIN
O +
Q
P +
R
I
+
S
M PI
O +
8F
F
;<
+
8G
:<
G
:<
, (2.8)
where
+
= =
,
R
= − 1
, IN = [?
I + ?
I ], and I =
8E
E
,
27
D = 1 for 2001-03, and D = 0 for 1992-2000.
+
denotes markup before the reforms and
-
is the change in markup after 2001. If trade liberalization leads to greater
international competition, then we would expect mark-ups to fall and
-
to be negative.
If trade liberalization is accompanied by an overall increase in firm productivity growth,
β
3
should be positive.
R
is simply the scale parameter, , minus one. Lastly,
S
denotes
the change in returns to scale subsequent to liberalization. The term
8F
F
;<
can be thought of
as the average rate of productivity growth in industry j, whereas
8G
:<
G
:<
can be written as the
sum of a firm-specific productivity growth, [
, and an error term, &
. For this reason,
8F
F
;<
can be replaced by a constant term,
. The final estimating equation, therefore,
becomes:
I
=
+
+
M IN
O +
-
M PIN
O +
Q
P +
R
I
+
S
M PI
O + [
+ &
. (2.9)
This paper uses balance sheet data collected in the form of a survey. The Balance
Sheet Data of Pakistani Listed and Non-Listed Companies (BSDPC) is a survey of a
representative sample of 654 companies in Pakistan for the years 1992-93 to 2002-2003.
The surveys are conducted by the Centre for Management and Economic Research
(CMER) and they encompass a wide range of topics.
14
The dataset covers almost all large
and medium-sized formal manufacturing enterprises. However, coverage of the industrial
sector is not complete since informal enterprises are excluded and small firms are under-
14
They are carried out in cooperation with the Lahore University of Management Sciences, Pakistan. The
survey is completed by managing directors and accountants of the company.
28
represented. For each company and year, we observe data on sales revenue, input use,
investment, wage bill, and all other costs. The sample includes firms in twenty industries:
Food, Sugar, Tobacco, Textile, Apparel, Leather Products, Electronics, Petroleum & Oil,
Chemicals, Non-Metallic Materials, Iron & Steel, Automobile, Metal tools & Products,
Furniture, Construction Materials, Energy, Paper & Printing, Medicine, Personal Care
Products and Miscellaneous. To estimate Eq. (2.9) using the panel of firms, we need data
on real output, capital stock, labor, raw materials, and their respective shares in real
output. Nominal output deflated by sectoral price deflators gives the real output.
15
Real
labor was found by deflating the total wage bill by the industry wage rate.
16
Materials
were also deflated using two-digit sectoral price deflators.
17
Real capital stock was
calculated by deflating net fixed assets by sectoral investment deflators.
18
Table 2.3
provides summary statistics for the balance sheet data used.
15
The Economic Survey of Pakistan, which is published annually by the Federal Bureau of Statistics,
Pakistan, provides price indices at the industry level for output and intermediate inputs, which are used as
deflators.
16
Real labor is taken to be the total number of employees, and not the number of hours worked, since the
hourly wage rate is not known. Many firms list the number of employees directly so there is no need to
deflate the wage bill by the industry wage rate. This information on wage bill and number of workers is
used to compute the mean industry wage rate by taking the average of the wage rate for firms in a particular
industry in a given year. This is then used to find number of workers for the remaining firms.
17
Ideally, the material inputs should be deflated by separate price indices for each different type of material
used in the production of the final good. However, the balance sheet data only lists the total material
expenditure. Harrison (1994) shows that the estimates based on deflating the material inputs using the
Input-Output table for each sector are not very different from those computed using the two-digit sectoral
price deflators.
18
The investment deflators are also obtained from the Economic Survey of Pakistan.
29
2.4 Estimation Results
Since Eq. (2.9) has a firm specific component, it can be estimated by assuming the
varying intercept terms across firms either as being fixed or random.
19
The overall
conclusion is not sensitive to the specification used; nevertheless, individual estimates do
change. On the basis of Hausman test statistic, we illustrate only the random effects
model results, just as in Krishna and Mitra (1998). In section 2.4.1, the estimates are
based on the assumption of constant returns to scale. We then relax this assumption in the
following section and allow the returns to scale to change in the post-liberalization
period.
Table 2.3: Summary Statistics
Variable Observations Mean Standard Deviation
Sales 5102 1.45E+09 8.38E+09
Cost of goods sold 5040 1.27E+09 7.49E+09
Average cost 4949 2.230019 39.27782
Net fixed assets 5643 5.98E+08 3.04E+09
Manufacturing salaries 3927 4.99E+07 1.45E+08
Long-term Debt 3800 4.76E+08 2.55E+09
Raw materials 3548 5.96E+08 1.96E+09
Wage 5886 114809.4 128830.9
Real Wage 5886 107000 120498.6
Industry Wage 5886 114809.4 110143.5
Real Industry Wage 5886 107000 102728.9
Input tariffs 2677 21.83493 3.304566
Output tariffs 2556 25.51972 7.703608
Energy price index 5949 112.3225 5.824518
Nominal Exchange rate 5949 45.33399 11.58271
ln(Sales) 5119 19.08702 3.045304
ln(Raw materials) 3448 18.85979 1.993501
ln(Fixed assets) 5643 18.42806 2.010811
ln(Number of Employees) 3898 5.356398 1.628081
ln(Real Wage) 5886 11.31265 0.812069
ln(Real Industry Wage) 5886 11.34583 0.732316
ln(Average cost) 4949 0.683118 0.265268
ln(Long-term Debt) 3779 17.93549 2.071613
ln(Energy price index) 5949 4.72004 0.051581
ln(Nominal Exchange rate) 5949 3.779354 0.267729
19
See Hsiao (1990) for a complete discussion of the merits of each methodology.
30
2.4.1 Random effects and IV estimates with constant returns to scale
Table 2.4 shows the results of random effects model which assumes constant returns to
scale but allows the competitive behavior of firms to change after trade liberalization.
+
denotes markup before the reforms and
-
is the change in markup after 2001. A higher
value of
+
implies a greater market power. This coefficient is positive and significant
for all industries, and is higher for industries which were protected more heavily. For
some industries, such as Food, Textile and Apparel, the markup is not extraordinarily
high despite being highly protected. This might be due to the fact that these industries
were also heavily engaged in exports and were forced to act competitively (Harrison,
1994). If trade liberalization leads to greater international competition, then we would
expect mark-ups to fall and
-
to be negative. This is true in all but four industries, and
the results are statistically significant. A fall in markup in exporting sectors, such as
Textile, could be partly due to the appreciation of Pakistani rupee. If trade liberalization
is accompanied by an overall increase in firm productivity growth,
Q
,
should be
positive. This is the case only in nine out of the nineteen industries shown in Table 2.4.
For all industries combined, the coefficient is negative but is very small. The coefficient
is positive and significant only for three industries: Chemicals, Electronics and Non-
Metallic Materials. By and large, the coefficient
Q
is very small and it seems that there
was very little, if any, change in the growth of productivity after opening to trade.
One may argue that a simple regression is likely to produce biased results since
inputs and outputs are simultaneously determined by firms. A standard way of handling
endogeneity is to use instrumental variables (IVs) which are correlated with materials and
31
labor per unit of capital but unrelated to any demand shocks affecting the firm. Following
the literature, we use real wage rate calculated at the sectoral level, nominal exchange
rate, energy price index and firm’s long-term debt as instruments. The instrumental
Table 2.4: Random Effects Estimates with Constant Returns to Scale
Industry β
1
β
2
β
3
R
2
Hausman Test
Food 0.7218*** -0.0665 0.0033 0.5009 0.9641
(0.0661) (0.1141) (0.0026)
Sugar 0.2698** -0.0602 -0.0019 0.0902 0.8374
(0.1255) (0.1902) (0.0041)
Medicine 0.1872*** 0.7329*** 0.0015 0.9405 0.3295
(0.0620) (0.0634) (0.0022)
Textile 0.3503*** 0.2322** 0.0004 0.2528 0.4572
(0.1023) (0.1386) (0.0016)
Apparel 0.3167*** -0.3789*** -0.0005 0.2869 0.9169
(0.1010) (0.1327) (0.0014)
Paper & Printing 0.9597*** -0.8463*** -0.0020 0.6923 0.0000
(0.0935) (0.1190) (0.0029)
Automobile 0.6342*** -0.1468 0.0000 0.3343 0.9815
(0.1093) (0.2372) (0.0054)
Energy 0.1073 -0.2154 0.0018 0.0017 0.9490
(0.1879) (0.3167) (0.0045)
Chemicals 0.6633*** -0.5962*** 0.0030* 0.7637 -
(0.0900) (0.1397) (0.0022)
Electronics 0.5417*** -0.4135*** 0.0046* 0.4502 0.8061
(0.1227) (0.1366) (0.0030)
Furniture 1.6950*** -1.0530*** -0.0041 0.9108 0.9966
(0.2609) (0.4258) (0.0078)
Metal Tools & Products 0.4532 -0.2624 -0.0137 0.1494 -
(0.6279) (0.7106) (0.0192)
Leather Products 0.7705*** -0.2106 -0.0024 0.4218 -
(0.3102) (0.3213) (0.0032)
Personal Care Products 0.3542** 0.0784 -0.0002 0.3419 -
(0.1801) (0.2719) (0.0023)
Construction Materials 0.3830*** 0.0127 0.0004 0.3783 0.9756
(0.1424) (0.2030) (0.0018)
Petroleum & Oil 0.3841** -0.0009 -0.0050** 0.2765 0.9643
(0.1862) (0.1950) (0.0026)
Iron & Steel 0.7948*** -0.2680*** -0.0040 0.8895 0.8458
(0.0647) (0.0891) (0.0061)
Non-metallic Materials 1.0858*** -1.2083*** 0.0060* 0.5930 -
(0.4488) (0.4494) (0.0043)
Miscellaneous 0.6402*** -0.0147 -0.0022 0.6286 0.7690
(0.0946) (0.1207) (0.0047)
All Industries
0.3906***
(0.0567)
-0.1293
(0.1470)
-0.0002
(0.0009)
0.2751 0.7414
4otes: Robust standard errors are given in parentheses. The Hausman test is used to test the hypothesis that fixed and
random effects models give the same estimates. The critical 5 percent value for the χ
2
(2) = 5.99. A higher value
indicates a rejection of the null that the two estimates are the same. *** Significant at, or below, 1 percent. **
Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
32
Table 2.5: Instrumental Variables Estimates with Constant Returns to Scale
Industry β
1
β
2
β
3
R
2
Over-Identification Test
Food 1.2190*** -0.4861 0.0004 0.4569 8.233
(0.4523) (1.2819) (0.0093)
Sugar -0.7975 0.8010 -0.0030 0.0568 1.366
(1.4779) (1.6470) (0.0115)
Medicine 0.5642* -0.5526 -0.0002 0.2916 2.535
(0.3553) (0.4756) (0.0072)
Textile 0.2034* 0.8776* -0.0000 0.1352 10.236
(0.1302) (0.6505) (0.0028)
Apparel 0.2347* -0.6194** -0.0018 0.1156 3.285
(0.1654) (0.3678) (0.0020)
Paper & Printing 1.2503*** -1.2365** 0.0026 0.8229 3.592
(0.2027) (0.5596) (0.0041)
Automobile 0.5423 -1.5836 0.0071 0.0718 0.388
(1.2167) (1.5411) (0.0182)
Energy 0.0589 -0.3940 0.0027 0.0106 5.283
(0.8745) (1.1350) (0.0222)
Chemicals 0.6050*** -0.4326*** 0.0056 0.7300 8.650
(0.0982) (0.1731) (0.0044)
Electronics 0.1924308 2.139 -0.0015 0.2732 2.963
(0.4883) (4.8569) (0.0189)
Leather Products 2.5319*** -2.1697** -0.0020 0.3856 4.165
(0.9690) (1.1574) (0.0195)
Construction Materials -0.6727 0.9534 -0.0045 0.1173 0.349
(0.5590) (1.9460) (0.0125)
Petroleum & Oil -0.3793 1.3915 -0.0089 0.0051 4.077
(0.6563) (3.3422) (0.0080)
Miscellaneous 0.4879*** -0.4801 0.0007 0.4214 2.986
(0.2072) (0.5814) (0.0086)
All Industries
0.0279
(0.1894)
0.0616
(0.5315)
-0.0009
(0.0015)
0.0514 9.087
4otes: Robust standard errors are given in parentheses. The instruments are: log of nominal exchange rate, log price
index for energy, log of debt, log of sectoral wages, D and D interacted with these four variables. The over-
identification test gives the χ
2
-statistic for the hypothesis that the instruments are accepted as valid. The critical 5
percent value of the χ
2
(7) = 14.1. A higher value indicates rejection of the test. *** Significant at, or below, 1 percent.
** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
variables results are shown in Table 2.5.
20
For most of the industries, instrumental
variable estimates are not very different from the random effects model. Almost all
industries display considerable markups prior to trade liberalization and a reduction in
markups following the liberalization episode. Food, Paper & Printing and Leather
Products have a much larger markup now, and the estimates are highly significant. Three
20
We also ran an over-identification test to test for the validity of the instruments used; in all industries, we
cannot reject the hypothesis that the instruments are valid.
33
of the sectors, Sugar, Construction Materials and Petroleum & Oil, nonetheless, have a
negative
+
but they are not statistically significant. Again,
Q
is very small indicating
that liberalization did not affect productivity in most of the sectors. However, the
standard errors are also very large for a large number of instrumental variable estimates.
Since our principal interest is to measure changes in markup, productivity and
returns to scale, in spite of a correlation between inputs and disturbance term, the
direction of bias is not obvious. If this correlation doesn’t change after trade
liberalization, the estimates of changes in markup, productivity and returns to scale will
be unbiased.
21
In the rest of the chapter, we will only present random effects model
results.
2.4.2 Random effects estimates with no restriction on returns to scale
Let us now deviate from the simplifying assumption of constant returns to scale and
allow the returns to scale to change after liberalization. As noted above, if β differs from
one, i.e. the firm faces either increasing or decreasing returns to scale, observed
productivity growth reflects both changes in Hicks-neutral technical progress as well as
the impact of scale effects on economic efficiency (Harrison, 1994). If returns to scale
decrease following liberalization, the assumption of constant returns to scale biases
estimates of the growth rate of productivity upward before liberalization and biases the
post-liberalization estimates downward (Krishna & Mitra, 1998). Table 2.6 shows
21
Krishna and Mitra (1998), for instance, run Monte Carlo simulations with different assumptions about the
correlation between the right hand side variables and the error term. In particular, they estimate the bias that
a correlation between the inputs and the error term would create. They could not reject the null hypothesis
that the bias in measuring the change in markup and the growth of productivity was zero.
34
estimates of the random effects model with no restriction on returns to scale. There is a
significant reduction in the estimates of markup compared to the case of constant returns
to scale in a majority of industries, including Food, Sugar, Medicine, Chemicals and
Electronics. For all industries combined, there is a reduction in markup following
liberalization. This is also true for most individual industries except Medicine and Metal
tools & Products. As far as the overall rate of growth of productivity is concerned, the
coefficient is significant and positive for four industries, namely, Food, Electronics,
Furniture and Non-Metallic Materials. The coefficient is, however, again very small and
insignificant for all industries taken together. On the other hand, two of the industries,
Iron & Steel and Metal tools & Products display a 1.2 and 4.1 percentage point reduction
in productivity growth, respectively, and the coefficients are statistically significant.
Another important point to be noted here is that the estimates of rate of growth of
productivity for most of the industries are higher and have a greater significance under no
restriction on returns to scale, compared to productivity growth estimates under the
assumption of constant returns to scale. This can be seen by comparing the coefficients
Q
in Tables 2.4 and 2.6.
The coefficient
R
is a measure of returns to scale. Recall from Eq. (2.9), the
coefficient of dk is given by (β-1), and β = 1 signifies constant returns to scale. This
implies that if
R
is positive (negative), the firm faces increasing (decreasing) returns to
scale. Table 2.6 illustrates that all industries in our sample display decreasing returns to
scale and the negative sign is significant for most of the industries. We also run an F-test
to check whether
R
= 0, and in a majority of industries, the null hypothesis is rejected.
35
Table 2.6: Random Effects Estimates with no Restriction on Returns to Scale
Industry β
1
β
2
β
3
β
4
β
5
R
2
F - test
Food 0.713*** -0.057 0.004** -0.087 -0.026 0.513 0.56
-0.082 -0.122 -0.002 -0.116 -0.154
Sugar 0.252** -0.090 -0.001 -0.157 -0.123 0.102 1.58
-0.123 -0.170 -0.004 -0.125 -0.153
Medicine 0.155*** 0.526*** 0.000 -0.12*** -0.021 0.954 12.27
-0.051 -0.177 -0.002 -0.035 -0.099
Textile 0.342*** 0.081 0.001 -0.072 -0.155* 0.269 0.85
-0.103 -0.159 -0.002 -0.078 -0.097
Apparel 0.232*** -0.29*** 0.001 -0.21*** -0.191* 0.492 17.81
-0.075 -0.116 -0.001 -0.049 -0.135
Paper & Printing 0.963*** -0.92*** 0.002 -0.037 -0.503*** 0.755 0.15
-0.104 -0.107 -0.002 -0.096 -0.125
Automobile 0.554*** -0.057 0.001 -0.17*** 0.128 0.368 13.99
-0.103 -0.204 -0.005 -0.044 -0.335
Energy -0.116 0.113 -0.003 -0.55*** 0.474 0.017 15.59
-0.194 -0.330 -0.005 -0.140 -0.477
Chemicals 0.406*** -0.37*** 0.001 -0.24*** -0.210** 0.867 50.90
-0.055 -0.108 -0.002 -0.034 -0.106
Electronics 0.443*** -0.303** 0.004* -0.17*** -0.571*** 0.558 9.39
-0.120 -0.132 -0.003 -0.055 -0.208
Furniture 0.690*** -0.83*** 0.007* -0.37*** -0.438*** 0.972 50.09
-0.223 -0.287 -0.005 -0.053 -0.081
Metal Tools and Products 0.363 9.571*** -0.041** -3.280 64.694*** 0.475 1.43
-0.601 -0.601 -0.024 -2.741 -2.741
Leather Products 0.566*** -0.046 -0.002 -0.572* 0.468 0.461 2.00
-0.231 -0.280 -0.003 -0.405 -0.452
Personal Care Products 0.318*** -0.047 0.000 -0.29*** -0.078 0.755 25.27
-0.115 -0.167 -0.002 -0.058 -0.113
Construction Materials 0.346** -0.066 0.001 -0.22*** -0.079 0.460 7.40
-0.151 -0.203 -0.002 -0.082 -0.094
Petroleum and Oil 0.590*** -0.339** -0.003 -0.27*** -0.186 0.627 21.81
-0.125 -0.155 -0.002 -0.057 -0.148
Iron & Steel 0.786*** -0.34*** -0.012** -0.07*** -3.022*** 0.917 6.02
-0.070 -0.085 -0.006 -0.027 -0.929
Non-metallic Materials 0.598*** -0.72*** 0.004* -0.46*** 0.429*** 0.933 55.45
-0.189 -0.194 -0.003 -0.061 -0.156
Miscellaneous 0.603*** 0.041 0.000 -0.14*** 0.286** 0.688 22.72
-0.079 -0.118 -0.005 -0.030 -0.134
All Industries 0.331*** -0.212** 0.001 -0.18*** -0.247*** 0.406 40.04
-0.050 -0.103 -0.001 -0.029 -0.067
4otes: Robust standard errors are given in parentheses. The F-test is used to test the hypothesis that the coefficient of
dk is equal to zero, i.e. the firm faces constant returns to scale. *** Significant at, or below, 1 percent. ** Significant at,
or below, 5 percent. * Significant at, or below, 10 percent.
This means that the estimates in Table 2.4 are misleading since they are based on the
assumption of constant returns to scale. Furthermore, a negative sign of
S
indicates that
36
there was a reduction in returns to scale after 2001. This holds true for most of the
sectors, and is consistent with the findings of Tybout et al. (1991) pertaining to the
Chilean experience. According to Devarajan and Rodrik (1991), the reduction in returns
to scale may be caused by an increased exploitation of returns to scale by firms which
may have been operating at too small a scale before the reforms. As Krishna and Mitra
(1998) pointed out, if the degree of this change is relatively large, it may be indicative of
the existence of rather inflexible capacity constraints in these industries. In three
industries, nevertheless, there is a strong evidence of an increase in returns to scale.
A few caveats need to be mentioned here. Firms do not report data on individual
production runs, so output figures reflect aggregation over heterogeneous products. The
resulting aggregation bias either over- or understates productivity change. A second
source of bias is error in measuring explanatory variables. Factor inputs may well be
poorly measured. Measurement of capital, for example, is not adjusted for capacity
utilization by using K*E instead of K, where E is the firm’s energy use, because of
unavailability of data on energy use. This is commonly done in the literature and would
have lessened the problem of measurement error in the stock of capital.
2.5 Alternative Measure of Productivity
In Section 2.4, we concluded that trade openness lead to a fall in markup, returns to scale
and rate of growth of productivity in a majority of the manufacturing industries of
Pakistan. These findings may be sensitive to the estimation techniques used to measure
productivity. In this section, therefore, we present an alternative measure of productivity
37
to quantify change in the rate of growth of productivity. One way of doing this is by
using the Tornquist index number formula, which in our case is given by:
TFP = [ln
− ln
`+
] − [?
ln
− ln
`+
+ ?
ln
− ln
`+
+ 1 − ?
− ?
ln
− ln
`+
], (2.10)
and with the markup factor, µ, and returns to scale parameter, β, just as in Harrison
(1994):
TFP = [ln
− ln
`+
] − =[?
ln
− ln
`+
+ ?
ln
− ln
`+
+ /= − ?
− ?
ln
− ln
`+
]. (2.11)
In order to compute TFP using Eq. (2.11), we use the estimates of = and from Table
2.6. In particular, we need the coefficients
+
and
R
in Eq. (2.9) to estimate Eq. (2.11).
If the data is split into before- and after-reform, it can be used to compute the rate of
growth of productivity over 1992-2000 and 2001-03. These estimates are shown in Table
2.7. The first column shows estimates of productivity growth based on the assumptions of
constant returns to scale and perfect competition. These assumptions are relaxed in the
last column. Under the assumptions of perfect competition and constant returns to scale,
there is a decline in the growth rate of productivity in eight out of eighteen industries
shown in Table 2.7. On the whole, productivity growth rate fell by 0.01 percentage
points. If we relax these assumptions, the difference in the pre- and post-reform rate of
growth of productivity goes down in a majority of industries; ten out of eighteen
industries undergo a fall in productivity growth rate. For instance, the rate of productivity
38
growth in the pre-reform period is 0.05 percent in Apparel industry. In the post-reform
period, it is -0.04 percent; productivity growth rate fell by 0.1 percentage points after
trade liberalization. We note that nearly all results are in line with the estimates of
productivity growth derived earlier in Section 2.4. Also, in fifteen out of the eighteen
industries shown in Table 2.7, under the assumptions of constant returns to scale and
perfect competition, the change in the rate of growth of productivity following
liberalization is over-estimated if it is positive and under-estimated if negative.
2.6 Conclusion
There has been an enormous reduction in tariff rates and duties imposed on the imports of
manufactured goods in Pakistan over the past decade. It is typically presumed that a rise
in foreign competition makes the industrial sector more efficient. In this chapter, we
study firm-level panel data of Pakistani companies in the manufacturing sector for
evidence that supports this perception. The main contribution of this study is that it uses a
novel data set from a developing country in order to revisit a fundamental question that
has been the centre of debate in the new trade theory. It presents evidence for the view
that the gains from trade liberalization in the form of improvement in firm performance
have been significantly overstated. Furthermore, in view of the fact that our data set
encompasses a broad range of manufacturing industries, we are able to reveal cross-
industry heterogeneity in the effect of trade reforms. A comparison of different industries
is expected to depict something about the effects of liberalization given that all industries
were exposed to analogous changes in macro conditions and measurement errors.
39
The initial results show that the productivity gains from reducing overall tariff
rates are not very obvious from a simple regression of TFP on tariff rates; if there are any
gains from reducing tariffs on intermediate goods, they are counterbalanced by a
reduction in tariffs on final goods to a great extent. After reviewing trade liberalization
reforms introduced in Pakistan over the past few years, we use the econometric
methodology of Harrison (1994) in order to measure productivity using a technique that
deviates from the assumptions of perfect competition and constant returns to scale, a
relaxation of estimation constraint that considerably improves the estimates. The results
of the model imply that there is a rise in competition following trade liberalization.
However, for most of the industries, there is also a reduction in the returns to scale. The
reduction in returns to scale may be caused by an increased exploitation of returns to
scale by firms which may have been operating at too small a scale before the reforms. If
the degree of this change is relatively large, it may be indicative of the existence of rather
inflexible capacity constraints in these industries. As far as the impact of trade openness
on firm-level productivity is concerned, there is no strong evidence of an improvement in
productivity after trade reforms were introduced in the manufacturing sectors of Pakistan.
Nonetheless, if returns to scale decrease following liberalization, the assumption of
constant returns to scale biases estimates of the growth rate of productivity upward before
liberalization and biases the post-liberalization estimates downward. Thus, the estimates
of change in the growth rate of productivity following liberalization are biased
downward.
40
Table 2.7: Modified TFP and Sensitivity of TFP change to assumptions of perfect competition (µ=1) and constant
returns to scale (β=1)
Industry µ=1, β=1 µ ≠ 1, β=1 µ ≠ 1, β ≠ 1
Pre- Post- Change Pre- Post- Change Pre- Post- Change
reform reform reform reform reform reform
Food -0.052 0.006 0.058 -0.046 0.008 0.054 -0.036 0.000 0.037
Sugar -0.021 0.038 0.059 -0.020 0.040 0.060 -0.027 0.039 0.066
Medicine -0.006 0.039 0.045 -0.006 0.020 0.026 -0.004 0.011 0.015
Textile 0.000 -0.080 -0.080 0.006 -0.120 -0.127 0.008 -0.120 -0.128
Apparel 0.044 -0.028 -0.072 0.053 -0.041 -0.094 0.051 -0.041 -0.092
Paper & Printing 0.001 0.002 0.002 0.003 0.002 -0.002 0.005 -0.003 -0.008
Automobile 0.001 0.113 0.112 0.036 0.094 0.059 0.036 0.088 0.052
Chemicals -0.114 0.384 0.498 -0.091 0.326 0.417 -0.084 0.282 0.366
Electronics 0.046 0.045 -0.001 0.085 0.025 -0.059 0.089 0.014 -0.075
Furniture 0.220 -0.218 -0.439 0.181 -0.114 -0.295 0.203 -0.190 -0.392
Leather Products 0.030 -0.092 -0.123 0.172 -0.120 -0.292 0.207 -0.116 -0.323
Personal Care -0.022 0.040 0.062 0.026 0.004 -0.023 0.029 0.001 -0.028
Construction Materials 0.000 -0.054 -0.053 -0.002 -0.093 -0.091 0.021 -0.111 -0.131
Petroleum and Oil 0.039 -0.064 -0.104 0.055 -0.094 -0.148 0.055 -0.087 -0.142
Iron & Steel 0.002 -0.155 -0.157 0.005 -0.237 -0.243 0.017 -0.251 -0.268
Metal tools 0.032 0.081 0.050 0.020 0.070 0.050 -0.429 0.266 0.694
Energy -0.450 -0.114 0.336 -0.434 -0.140 0.294 -0.480 -0.156 0.324
Miscellaneous -0.036 0.121 0.157 -0.028 0.086 0.114 -0.029 0.065 0.093
All Industries -0.006 -0.002 -0.009
An apprehension is that the association between trade and expansion of higher-
productivity firms in a developing country might not be driven solely by changes in trade
policy, since trade liberalization is normally a part of a broader package of economic
reforms. Trade policy is also subject to potential endogeneity because the government
may lift current protection as a response to lobbying by firms in less productive
industries. Generalizations about the factors that may explain these changes are difficult
to draw: the pre-existing institutional environment, the extent of political turmoil, the
scope and depth of economic reforms, and the characteristics of concurrent
macroeconomic policies should also be taken into account. Future research should seek to
further identify factors accounting for heterogeneity in the effects of trade reform.
41
Chapter 3
The End of Multi-Fibre Arrangement and Textile Industry
I am in favor of helping the prosperity of all countries because, when we are all prosperous, the
trade with each becomes more valuable to the other.
– William Howard Taft
3.1 The Multi-Fibre Arrangement
The Multi-Fibre Arrangement (MFA) was the outcome of preceding short-term
agreements on the trade of textile and clothing (T&C) products amongst the developed
and developing countries. Signed in 1974, MFA imposed restrictions on exports by T&C
exporters to the developed countries by means of bilaterally negotiated quotas on textile
products. The MFA codified these agreements in a broad system involving almost “1000
different allotments [and] encompassing scores of categories” from 47 countries (Collins,
2003). According to its guidelines “individual quotas were negotiated which set precise
limits on the quantity of T&C which could be exported from one country to another.”
Moreover, T&C products were excluded from multilateral trade negotiations under the
General Agreement on Tariffs and Trade (GATT) and the World Trade Organization
42
(WTO). An important development of the Uruguay Round (1994) was signing of the
Agreement on Textile and Clothing (ATC) which put to an end the MFA. ATC
commenced the practice of integrating T&C products into GATT and WTO. The
integration occurred over a period of ten years and across four phases starting from
January 1
st
, 1995. Importing countries were to include a certain portion of all T&C
products covered by the ATC in each phase. The particular products integrated in each
phase were specific to the importing countries but were determined by two rules
(Brambilla et al., 2007). To begin with, the products retired in each phase had to consist
of goods from all four key textile and clothing segments: Yarn, Fabrics, Made-Up textile
products, and Clothing. Moreover, the selected products had to correspond to an agreed
fraction of each country’s 1990 T&C imports by volume. In Phase I, countries were to
incorporate products representing 16 percent of their 1990 import volumes. At the start of
Phases II and III on January 1
st
, 1998 and January 1
st
, 2002, respectively, further 17 and
18 percent of 1990 export volumes were integrated. Lastly, on January 1
st
, 2005, Phase
IV of the ATC integrated the outstanding 49 percent of export volumes, with all quotas
abolished. Other than removing quotas, the ATC enhanced developing countries’ access
to the developed countries by speeding up quota growth over the four phases through the
“growth-on-growth” provision (Brambilla et al., 2007). In Phase I, quota growth rates
were accelerated 16 percent per year, whereas they were accelerated by 25 and 27 percent
in Phases II and III, respectively.
22
22
A category with a base quota growth rate of 6 percent in 1994, for example, would grow at 6.96 percent
(0.06*1.25) per year during Phase I, 8.7 percent (0.0696*1.25) per year over Phase II, and 11.05
(0.087*1.27) percent per year during Phase III (Brambilla et al., 2007).
43
The first two phases of ATC were not very severe for the producers in developed
countries.
23
The U.S. postponed the removal of quotas on sensitive products until Phase
III. Of the 4,839 ten-digit Harmonized System (HS) product codes that the U.S. retired
over the four phases, 62 percent were retired in 2005.
24
For instance, products like tents
and life jackets were integrated in the ATC. However, they had not been subject to
import quotas. These products were integrated in the first phase (Brambilla et al., 2007).
The expiration of these quotas was expected to bring about a considerable
reallocation of production and exports across countries. This chapter analyzes the impact
of the end of MFA on productivity of T&C firms. In particular, it looks at the case of
T&C industry of Pakistan under the U.S. T&C quotas and the subsequent end of MFA.
The end of quota system, together with the mounting significance of the industry in its
domestic market, leads us to analyze the efficiency issues related to Pakistan’s textile
industry. Thus, unlike most other studies in the literature which mainly investigate the
effect of trade liberalization reforms in developing countries, this chapter investigates a
liberalization episode in a developed country and its consequence for firms in a
developing country. Furthermore, it highlights sectoral heterogeneity within the
manufacturing industry in the effect of MFA expiration.
23
ATC products made up 17.1 billion square meter equivalents (SME) worth of imports in 1990 in the
U.S., but the imports of products truly subject to quotas in that year totaled merely 12.2 billion SMEs
(Textiles and Apparel: Assessment of the Competitiveness of Certain Foreign Suppliers to the U.S. Market,
2004).
24
HS codes are the group of T&C products governed by the ATC and imported by the U.S.
44
Figure 3.1: Mean productivity of textile and clothing firms – Levinsohn & Petrin Productivity Measure
3.2 The T&C sector of Pakistan
The textile sector is an important industry in Pakistan in terms of output, export value,
foreign exchange earnings and employment.
25
Tables 3.1 and 3.2 demonstrate the export
value in millions of U.S. dollars of several cotton and cotton manufactures from 1971 to
2011. Pakistan is the fourth largest producer of cotton in the world and does not have to
rely on other countries for its raw materials. Moreover, labor costs in Pakistan are among
the lowest in the world. T&C make up roughly 74 percent of the total export value.
Tables 3.3 and 3.4 exhibit the production and export of yarn and cloth, respectively.
25
The spinning sector was the most privileged sector by investment. It received 47 percent of the $4 billion
investment in the T&C industry between 1999 and 2003. After China and India, Pakistan has the third-
largest capacity of short-staple spindles for spun yarn in the world (Textiles and Apparel: Assessment of
the Competitiveness of Certain Foreign Suppliers to the U.S. Market, 2004).
0 5 10 15
Mean Productivity
1990 1995 2000 2005 2010
Year
Textile firms
Clothing firms
45
Table 3.1: Exports of Cotton & Cotton Manufactures in Millions of U.S. Dollars
PERIOD
COTTON
YARN
COTTON
CLOTH
TENT & CANVAS
COTTON
BAGS
TOWELS BED WEAR
1971-72 127.5 81.5 1.9 1.2 6.1 0.9
1972-73 200.5 126.8 2.3 4.5 7.0 1.3
1973-74 189.5 143.9 7.5 8.1 16.3 4.9
1974-75 92.3 132.6 22.4 7.2 15.6 5.6
1975-76 145.0 137.5 25.1 7.9 18.3 3.7
1976-77 118.4 162.0 25.0 8.3 14.2 3.3
1977-78 107.0 175.9 26.0 9.0 12.9 4.8
1978-79 197.6 215.7 27.7 11.2 21.2 7.9
1979-80 205.9 244.1 31.7 21.2 26.5 12.0
1980-81 207.0 241.4 65.3 35.7 47.9 20.6
1981-82 196.7 279.5 64.3 31.4 42.9 35.8
1982-83 247.3 281.4 93.8 17.9 39.1 67.5
1983-84 217.6 360.2 64.1 15.5 46.6 53.3
1984-85 260.4 305.9 49.6 12.1 49.7 51.0
1985-86 279.2 314.8 31.1 9.5 67.5 90.1
1986-87 506.1 345.3 23.4 8.1 83.9 123.9
1987-88 541.0 485.4 30.3 12.3 117.4 136.9
1988-89 600.8 464.8 41.1 13.5 140.4 147.9
1989-90 833.7 559.0 28.8 13.4 129.8 190.8
1990-91 1183.0 675.8 79.6 20.5 129.4 246.2
1991-92 1172.5 819.4 51.2 32.4 136.7 284.0
1992-93 1121.5 863.1 39.9 23.7 139.0 351.6
1993-94 1259.3 820.6 29.1 17.3 129.2 285.6
1994-95 1528.1 1081.4 38.2 19.1 144.8 340.2
1995-96 1540.3 1275.9 39.5 24.6 174.1 422.2
1996-97 1411.5 1262.4 36.2 27.6 194.1 456.3
1997-98 1159.5 1250.3 58.1 23.1 200.1 508.8
1998-99 945.2 1115.2 40.8 20.8 177.7 611.0
1999-00 1071.6 1096.2 52.9 19.2 195.6 709.9
2000-01 1076.6 1035.0 50.0 19.0 243.0 734.9
2001-02 942.3 1132.7 47.4 18.2 269.8 918.5
2002-03 928.3 1345.6 73.2 18.2 374.8 1329.0
2003-04 1127.0 1711.7 75 18.0 404 1383
2004-05 1057.0 1863 67 0 520 1450
2005-06 1383.0 2108.0 39.0 13.7 588.0 2038.0
2006-07 1428.0 2027.0 69.0 11.4 611.0 1996.0
2007-08 1,301.0 2,011 71.0 10.4 613.0 1904.0
2008-09 1114.8 1955.3 56.2 8.4 642.9 1735.0
2009-10 1,433.1 1,800.1 61.5 5.3 668.2 1,744.3
2010-11 2,201.4 2,623.2 47.0 10.3 762.3 2,088.9
Source: All Pakistan Textile Mills Association (APTMA)
46
Table 3.2: Exports of Cotton & Cotton Manufactures in Millions of U.S. Dollars
PERIOD OTHER MADE-UPS GARMENTS HOSIERY THREAD
COTTON
MANUFACTURE
TOTAL
EXPORT
1971-72 1.2 3.2 3.2 2.3 229.0 590.7
1972-73 0.9 3.3 6.2 3.2 356.0 817.3
1973-74 4.5 8.6 8.3 5.3 396.9 1026.4
1974-75 3.7 14.4 10.3 5.8 309.9 1039.0
1975-76 5.0 21.5 10.2 4.0 378.2 1136.7
1976-77 3.8 30.4 11.8 4.4 381.6 1100.8
1977-78 2.9 30.4 9.8 7.1 385.8 1311.1
1978-79 5.4 38.1 12.3 5.8 542.9 1709.6
1979-80 3.9 53.9 20.0 7.1 626.3 2364.7
1980-81 11.1 75.2 23.2 10.1 737.5 2957.5
1981-82 9.7 94.2 28.5 7.7 790.7 2490.0
1982-83 19.5 122.7 36.5 12.8 938.5 2707.7
1983-84 22.9 162.4 56.0 8.6 1007.2 2768.1
1984-85 29.2 132.0 42.6 4.8 937.3 2491.2
1985-86 52.2 206.1 54.6 3.8 1108.9 3069.8
1986-87 51.1 355.2 96.6 3.3 1596.9 3686.4
1987-88 64.1 349.9 134.3 3.8 1875.5 4454.6
1988-89 58.8 335.5 166.9 3.0 1972.8 4661.5
1989-90 78.2 393.7 273.7 3.0 2504.2 4954.3
1990-91 108.9 497.1 333.6 3.4 3277.4 6133.1
1991-92 113.5 613.5 425.1 3.7 3652.1 6904.0
1992-93 125.5 617.7 464.1 4.8 3750.9 6813.5
1993-94 129.4 612.2 509.1 4.0 3795.8 6802.5
1994-95 163.5 641.7 688.5 1.9 4647.5 8137.2
1995-96 179.1 648.5 703.4 1.5 5009.1 8707.1
1996-97 208.7 736.4 688.9 1.7 5023.8 8320.3
1997-98 245.8 746.5 696.7 1.8 4890.7 8627.7
1998-99 255.3 651.2 742.1 1.5 4560.8 7779.3
1999-00 307.6 771.7 886.7 1.3 5112.7 8568.6
2000-01 328.2 827.5 910.3 1.0 5225.5 9224.7
2001-02 351.3 882 841.5 - 5404 9123.6
2002-03 359.7 1092.6 1146.6 - 6668.0 11160.2
2003-04 417.0 993 1459 - 7587.7 1231.3.
2004-05 466 1088 1635 0 8146 14391.0
2005-06 418.0 1310 1751 0.3 9649 16451.0
2006-07 514.0 1547.0 1798.0 0.2 10001.6 16976.0
2007-08 537.0 1452.0 1732.3 0.2 9631.9 19052.0
2008-09 480.1 1230.0 1740.8 - 8963.5 17688.0
2009-10 537.2 1,269.3 1,744.3 - 9,263.3 19,290.0
2010-11 625.0 1,773.7 2,305.6 - 12,437.2 24,810.4
Source: All Pakistan Textile Mills Association (APTMA)
47
Table 3.3: Production & Export of Yarn in Thousands of Kilograms
Year Production
Exports
Year Production
Exports
Quantity % of Production Quantity % of Production
1971-72 335,702 130,158 38.77 1991-92 1,188,270 505,863 42.57
1972-73 376,122 184,404 49.03 1992-93 1,234,539 555,294 44.98
1973-74 379,460 100,564 26.50 1993-94 1,498,948 578,648 38.60
1974-75 351,200 78,365 22.31 1994-95 1,413,648 522,091 36.93
1975-76 349,653 112,182 32.08 1995-96 1,505,244 535,889 35.60
1976-77 282,640 64,294 22.75 1996-97 1,530,855 508,188 33.20
1977-78 297,895 59,955 20.13 1997-98 1,540,720 461,919 29.98
1978-79 327,796 97,929 29.87 1998-99 1,547,632 421,481 27.23
1979-80 362,862 99,834 27.51 1999-00 1,678,536 512,971 30.56
1980-81 374,947 95,232 25.40 2000-01 1,729,129 545,134 31.59
1981-82 430,154 95,621 22.23 2001-02 1,818,345 539,500 29.67
1982-83 448,430 134,100 29.90 2002-03 1,924,936 525,130 27.28
1983-84 431,580 101,805 23.59 2003-04 1,938,908 514,279 26.52
1984-85 431,731 125,855 29.15 2004-05 2,290,340 520,782 22.74
1985-86 482,186 157,895 32.75 2005-06 2,216,605 691,492 31.20
1986-87 586,371 259,668 44.28 2006-07 2,727,566 699,259 25.64
1987-88 685,031 210,950 30.79 2007-08 2,809,383 594,936 21.18
1988-89 767,434 291,953 38.04 2008-09 2,862,411 526,246 18.38
1989-90 925,382 374,976 40.52 2009-10 2,880,970 612,413 21.26
1990-91 1,055,228 501,072 47.48 2010-11 3,016,972 549,947 18.23
Source: All Pakistan Textile Mills Association (APTMA)
Table 3.4: Production & Export of Cloth in Million Square Meters
Year Production
Exports
Year Production
Exports
Quantity % of Production Quantity % of Production
1971-72 1350.67 409.81 30.34 1991-92 3238.99 1196.12 36.93
1972-73 1238.11 517.98 41.84 1992-93 3360.00 1127.58 33.56
1973-74 1828.72 353.02 19.30 1993-94 3378.00 1046.79 30.99
1974-75 1827.08 440.81 24.13 1994-95 3100.75 1160.66 37.43
1975-76 1503.36 463.84 30.85 1995-96 3706.00 1323.09 35.70
1976-77 1445.30 416.84 28.84 1996-97 3781.20 1257.43 33.25
1977-78 1573.07 453.47 28.83 1997-98 3913.70 1271.27 32.48
1978-79 1487.10 531.53 35.74 1998-99 4386.79 1355.17 30.89
1979-80 1720.02 545.77 31.73 1999-00 4987.16 1574.88 31.58
1980-81 1834.00 500.90 27.31 2000-01 5591.40 1736.00 31.05
1981-82 2200.44 584.35 26.56 2001-02 5653.09 1957.35 34.62
1982-83 2048.77 605.33 29.55 2002-03 5650.52 2005.38 35.49
1983-84 2165.98 664.38 30.67 2003-04 6833.12 2412.87 35.31
1984-85 2000.00 687.62 34.38 2004-05 6480.67 2751.56 42.46
1985-86 1985.40 727.35 36.63 2005-06 8524.26 2633.98 30.90
1986-87 2009.85 693.42 34.50 2006-07 8694.92 2211.84 25.44
1987-88 2230.82 848.61 38.04 2007-08 9005.44 2035.14 22.60
1988-89 2250.00 845.33 37.57 2008-09 9015.26 1898.54 21.06
1989-90 2734.77 1017.87 37.22 2009-10 8949.77 1753.12 19.59
1990-91 2854.00 1056.53 37.02 2010-11 9018.32 2297.49 25.48
Source: All Pakistan Textile Mills Association (APTMA)
48
The government took steps to ensure competitiveness even prior to the MFA
expiration.
26
The T&C industry in Pakistan is composed primarily of small and medium
enterprises, and a group of vertically integrated mills. Figure 3.1 shows the evolution of
mean productivity of T&C firms in our sample. Productivity is computed using the
Levinsohn and Petrin productivity measure. For the time period under consideration,
textile firms have a much higher mean productivity than clothing firms. Nevertheless, we
do observe an upward trend in the average productivity of both types of firms.
The focus of chapters 3 and 4 is on the exports of T&C products by Pakistan to
the U.S. only. The reason why this is an interesting case to consider is because United
States is the most important trading partner of Pakistan for a sizeable majority of T&C
products. For most of the clothing products exported, the U.S. captures a considerable
proportion of total market share.
27
Moreover, the fill rates for nearly all T&C products are
very close to one hundred, indicating that the quotas imposed by the U.S. were usually
binding.
28
Table 3.5 displays the adjusted quota base, level of imports and fill rates for a
sample of OTEXA (U.S. Office of Textile and Apparel) categories. Let us look at two
examples. Figure 3.2 exhibits the total imports into the U.S. from Pakistan, and the
adjusted quota base from 1984 up to 2004 for two T&C products, one from the textile
26
The private and public sectors together formed the National Textile Institute (Faisalabad) in 1959. The
government proposed the Textile Vision 2005, which involves giving loans to upgrade equipment, interest
rate and tax policy reforms, and promotion of product and market diversification.
27
This was verified using the statistical database of All Pakistan Textile Mills Association (APTMA,
2011).
28
The data set permits us to compute the fill rates for all T&C products. Fill rate is defined in the literature
as total imports as a percentage of adjusted base quota. Even though the adjusted base quotas can exceed
base quotas, the fill rate cannot exceed 100 since it is defined as imports over adjusted base. Evans and
Harrigan (2005) define a binding quota as the one for which the fill rate exceeds 90 percent.
49
and clothing industries each. For Textile and Fabric Finishing as well as Men’s and Boys’
Cut and Sew Suit, Coat, and Overcoat, the actual number of imports closely followed the
adjusted quota base. In the light of the phasing out of MFA, this evidence makes the case
of Pakistan-U.S. trade in T&C industry all the more appealing.
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
1993 219 Duck Fabric M2 5500000 5500000 1
1994 219 Duck Fabric M2 5885000 3983780 0.676938
1995 219 Duck Fabric M2 5606114 2842510 0.507038
1996 219 Duck Fabric M2 6818078 6058734 0.888628
1997 219 Duck Fabric M2 8777010 8454310 0.963234
1998 219 Duck Fabric M2 7200397 5611143 0.779283
1999 219 Duck Fabric M2 7758895 3621719 0.466783
2000 219 Duck Fabric M2 8736258 7030377 0.804736
2001 219 Duck Fabric M2 1.08E+07 6753098 0.625608
2002 219 Duck Fabric M2 1.16E+07 10054596 0.87003
2003 219 Duck Fabric M2 1.30E+07 11025657 0.845117
2004 219 Duck Fabric M2 1.67E+07 11393881 0.68291
1993 314 Cotton Poplin & Broadcloth Fabric M2 3529200 3419602 0.968945
1994 314 Cotton Poplin & Broadcloth Fabric M2 4750800 1882077 0.39616
1995 314 Cotton Poplin & Broadcloth Fabric M2 3323319 1206620 0.363077
1996 314 Cotton Poplin & Broadcloth Fabric M2 4958603 2935625 0.592027
1997 314 Cotton Poplin & Broadcloth Fabric M2 6383279 6148264 0.963183
1998 314 Cotton Poplin & Broadcloth Fabric M2 5577228 5577228 1
1999 314 Cotton Poplin & Broadcloth Fabric M2 6944831 4895780 0.704953
2000 314 Cotton Poplin & Broadcloth Fabric M2 6646990 6646990 1
2001 314 Cotton Poplin & Broadcloth Fabric M2 9103492 9103492 1
2002 314 Cotton Poplin & Broadcloth Fabric M2 9619245 9582178 0.996147
2003 314 Cotton Poplin & Broadcloth Fabric M2 1.09E+07 10430209 0.960494
2004 314 Cotton Poplin & Broadcloth Fabric M2 1.23E+07 9637755 0.786177
1991 315 Cotton Print Cloth Fabric M2 5.16E+07 51576942 1
1992 315 Cotton Print Cloth Fabric M2 5.44E+07 54413674 1
1993 315 Cotton Print Cloth Fabric M2 6.06E+07 56601311 0.933711
1994 315 Cotton Print Cloth Fabric M2 6.56E+07 63840951 0.973061
1995 315 Cotton Print Cloth Fabric M2 6.70E+07 62885763 0.938984
1996 315 Cotton Print Cloth Fabric M2 6.25E+07 48527274 0.77664
1997 315 Cotton Print Cloth Fabric M2 8.60E+07 80625620 0.937126
1998 315 Cotton Print Cloth Fabric M2 7.64E+07 76408847 1
1999 315 Cotton Print Cloth Fabric M2 7.11E+07 57271284 0.805458
2000 315 Cotton Print Cloth Fabric M2 7.52E+07 58815757 0.782006
2001 315 Cotton Print Cloth Fabric M2 8.67E+07 78064295 0.90072
2002 315 Cotton Print Cloth Fabric M2 1.17E+08 1.17E+08 1
2003 315 Cotton Print Cloth Fabric M2 1.14E+08 1.06E+08 0.927237
2004 315 Cotton Print Cloth Fabric M2 1.47E+08 78932440 0.537423
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
50
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates (Continued)
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
1994 317/617 MMF Twill And Sateen Fabric M2 2.30E+07 17201696 0.7479
1995 317/617 MMF Twill And Sateen Fabric M2 1.93E+07 12039372 0.622763
1996 317/617 MMF Twill And Sateen Fabric M2 2.66E+07 19048809 0.714866
1997 317/617 MMF Twill And Sateen Fabric M2 3.43E+07 34302672 1
1998 317/617 MMF Twill And Sateen Fabric M2 2.99E+07 29901543 1
1999 317/617 MMF Twill And Sateen Fabric M2 3.31E+07 21604068 0.652369
2000 317/617 MMF Twill And Sateen Fabric M2 3.84E+07 32280324 0.840262
2001 317/617 MMF Twill And Sateen Fabric M2 4.52E+07 33642099 0.744576
2002 317/617 MMF Twill And Sateen Fabric M2 5.70E+07 55857219 0.979842
2003 317/617 MMF Twill And Sateen Fabric M2 5.84E+07 56259003 0.964072
2004 317/617 MMF Twill And Sateen Fabric M2 6.59E+07 56710278 0.860839
1991 331/631 Cotton & MMF Gloves & Mittens DPR 4149613 4149612.9 1
1992 331/631 Cotton & MMF Gloves & Mittens DPR 4298328 4298327.8 1
1993 331/631 Cotton & MMF Gloves & Mittens DPR 5225211 5225211.3 1
1994 331/631 Cotton & MMF Gloves & Mittens DPR 5947642 5925369.9 0.996255
1995 331/631 Cotton & MMF Gloves & Mittens DPR 6430591 6430590.5 1
1996 331/631 Cotton & MMF Gloves & Mittens DPR 7114654 7114654.1 1
1997 331/631 Cotton & MMF Gloves & Mittens DPR 7355412 7355412.1 1
1998 331/631 Cotton & MMF Gloves & Mittens DPR 7784920 7730324.1 0.992987
1999 331/631 Cotton & MMF Gloves & Mittens DPR 9120778 9120778.4 1
2000 331/631 Cotton & MMF Gloves & Mittens DPR 1.06E+07 10561745 1
2001 331/631 Cotton & MMF Gloves & Mittens DPR 1.06E+07 10267923 0.973166
2002 331/631 Cotton & MMF Gloves & Mittens DPR 2747715 1508812 0.549115
2003 331/631 Cotton & MMF Gloves & Mittens DPR 3962053 1456208.9 0.367539
2004 331/631 Cotton & MMF Gloves & Mittens DPR 3716657 1421849.7 0.382561
1992 334/634 Other M&B cotton and MMF coats DOZ 6541200 6541200 1
1993 334/634 Other M&B cotton and MMF coats DOZ 7115729 5373409.5 0.755145
1994 334/634 Other M&B cotton and MMF coats DOZ 7426539 5997514.5 0.807579
1995 334/634 Other M&B cotton and MMF coats DOZ 9241412 6963307.5 0.75349
1996 334/634 Other M&B cotton and MMF coats DOZ 9362300 8715907.5 0.930958
1997 334/634 Other M&B cotton and MMF coats DOZ 9205704 7121214 0.773565
1998 334/634 Other M&B cotton and MMF coats DOZ 1.23E+07 10242740 0.829831
1999 334/634 Other M&B cotton and MMF coats DOZ 1.30E+07 13010882 1
2000 334/634 Other M&B cotton and MMF coats DOZ 1.33E+07 12151176 0.914748
2001 334/634 Other M&B cotton and MMF coats DOZ 2.14E+07 17412737 0.813117
2002 334/634 Other M&B cotton and MMF coats DOZ 2.42E+07 22245428 0.920172
2003 334/634 Other M&B cotton and MMF coats DOZ 2.73E+07 26630447 0.975774
1992 336/636 Cotton & MMF Dresses DOZ 1.00E+07 9381917.6 0.935222
1993 336/636 Cotton & MMF Dresses DOZ 1.21E+07 8639039.7 0.715614
1994 336/636 Cotton & MMF Dresses DOZ 1.54E+07 11835526 0.770508
1995 336/636 Cotton & MMF Dresses DOZ 1.41E+07 13226721 0.939638
1996 336/636 Cotton & MMF Dresses DOZ 1.73E+07 15759919 0.912777
1997 336/636 Cotton & MMF Dresses DOZ 1.73E+07 16131567 0.933601
1998 336/636 Cotton & MMF Dresses DOZ 1.88E+07 17240824 0.915346
1999 336/636 Cotton & MMF Dresses DOZ 1.84E+07 7362984.6 0.399599
2000 336/636 Cotton & MMF Dresses DOZ 2.33E+07 19182251 0.823895
2001 336/636 Cotton & MMF Dresses DOZ 2.56E+07 17012590 0.665267
2002 336/636 Cotton & MMF Dresses DOZ 3.16E+07 26824559 0.847631
2003 336/636 Cotton & MMF Dresses DOZ 2.98E+07 21127582 0.708673
2004 336/636 Cotton & MMF Dresses DOZ 4.11E+07 32319945 0.786017
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
51
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates (Continued)
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
1992 338 M&B Knit Shirts, Cotton DOZ 2.58E+07 25822104 1
1993 338 M&B Knit Shirts, Cotton DOZ 2.45E+07 21908160 0.893081
1994 338 M&B Knit Shirts, Cotton DOZ 2.79E+07 27890238 1
1995 338 M&B Knit Shirts, Cotton DOZ 3.13E+07 31344468 1
1996 338 M&B Knit Shirts, Cotton DOZ 3.17E+07 31693164 1
1997 338 M&B Knit Shirts, Cotton DOZ 3.17E+07 31718982 1
1998 338 M&B Knit Shirts, Cotton DOZ 3.41E+07 33052386 0.970578
1999 338 M&B Knit Shirts, Cotton DOZ 3.68E+07 36774354 1
2000 338 M&B Knit Shirts, Cotton DOZ 4.03E+07 40276782 1
2001 338 M&B Knit Shirts, Cotton DOZ 4.44E+07 44392812 1
2002 338 M&B Knit Shirts, Cotton DOZ 5.17E+07 51688488 1
2003 338 M&B Knit Shirts, Cotton DOZ 5.64E+07 56447706 1
2004 338 M&B Knit Shirts, Cotton DOZ 5.88E+07 58810998 1
1992 339 W&G Knit Shirts/Blouses, Cotton DOZ 5965572 5965572 1
1993 339 W&G Knit Shirts/Blouses, Cotton DOZ 6383160 5891052 0.922905
1994 339 W&G Knit Shirts/Blouses, Cotton DOZ 7121862 7121862 1
1995 339 W&G Knit Shirts/Blouses, Cotton DOZ 6753414 6753414 1
1996 339 W&G Knit Shirts/Blouses, Cotton DOZ 8352198 8352198 1
1997 339 W&G Knit Shirts/Blouses, Cotton DOZ 7526706 7440906 0.988601
1998 339 W&G Knit Shirts/Blouses, Cotton DOZ 9045354 8537808 0.943889
1999 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.07E+07 10733376 1
2000 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.22E+07 12219480 1
2001 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.11E+07 10820190 0.972356
2002 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.59E+07 14536554 0.91195
2003 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.70E+07 16717866 0.982085
2004 339 W&G Knit Shirts/Blouses, Cotton DOZ 1.80E+07 16278546 0.905849
1994 342/642 Cotton & MMF Skirts DOZ 2571174 1685279.4 0.655451
1995 342/642 Cotton & MMF Skirts DOZ 3619448 2781412.8 0.768463
1996 342/642 Cotton & MMF Skirts DOZ 4401907 2625439.6 0.596432
1997 342/642 Cotton & MMF Skirts DOZ 2780534 1119422.1 0.402593
1998 342/642 Cotton & MMF Skirts DOZ 1275127 1275127.1 1
1999 342/642 Cotton & MMF Skirts DOZ 5826571 2450260.3 0.420532
2000 342/642 Cotton & MMF Skirts DOZ 5640335 3453909.4 0.612359
2001 342/642 Cotton & MMF Skirts DOZ 7464006 3887454.7 0.520827
2002 342/642 Cotton & MMF Skirts DOZ 7867513 3826543.5 0.486373
2003 342/642 Cotton & MMF Skirts DOZ 8881696 2981951.9 0.335741
2004 342/642 Cotton & MMF Skirts DOZ 1.13E+07 3799351 0.336536
1992 347/348 Cotton Trousers/Slacks & Shorts DOZ 8402825 8402825.2 1
1993 347/348 Cotton Trousers/Slacks & Shorts DOZ 8251858 8251858.4 1
1994 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.08E+07 9960694.7 0.924569
1995 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.16E+07 9468190.1 0.81285
1996 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.26E+07 12137749 0.963777
1997 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.36E+07 13165104 0.966842
1998 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.50E+07 13742717 0.916263
1999 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.65E+07 1621045.5 0.09812
2000 347/348 Cotton Trousers/Slacks & Shorts DOZ 1.91E+07 19057681 1
2001 347/348 Cotton Trousers/Slacks & Shorts DOZ 2.00E+07 19970932 1
2002 347/348 Cotton Trousers/Slacks & Shorts DOZ 2.42E+07 24176427 1
2003 347/348 Cotton Trousers/Slacks & Shorts DOZ 2.73E+07 27292881 1
2004 347/348 Cotton Trousers/Slacks & Shorts DOZ 2.94E+07 29448628 1
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
52
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates (Continued)
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
1992 351/651 Cotton & MMF Nightwear & Pajamas DOZ 5277116 2973660 0.563501
1993 351/651 Cotton & MMF Nightwear & Pajamas DOZ 9276158 8252167.5 0.889611
1994 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.10E+07 9732690 0.883391
1995 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.14E+07 9906820.5 0.872117
1996 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.26E+07 11851097 0.938869
1997 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.35E+07 13277810 0.985685
1998 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.48E+07 14312109 0.964565
1999 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.63E+07 15460640 0.945955
2000 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.90E+07 19012371 1
2001 351/651 Cotton & MMF Nightwear & Pajamas DOZ 1.99E+07 19932657 1
2002 351/651 Cotton & MMF Nightwear & Pajamas DOZ 3.06E+07 26602512 0.86802
2003 351/651 Cotton & MMF Nightwear & Pajamas DOZ 3.59E+07 35853266 1
2004 351/651 Cotton & MMF Nightwear & Pajamas DOZ 4.05E+07 40474967 1
1992 352/652 Cotton & MMF Underwear etc. DOZ 4645995 4255873.8 0.916031
1993 352/652 Cotton & MMF Underwear etc. DOZ 6092056 4878458.6 0.80079
1994 352/652 Cotton & MMF Underwear etc. DOZ 7180246 4580138.6 0.63788
1995 352/652 Cotton & MMF Underwear etc. DOZ 7483515 7414257.7 0.990745
1996 352/652 Cotton & MMF Underwear etc. DOZ 8091173 7608504.7 0.940346
1997 352/652 Cotton & MMF Underwear etc. DOZ 8413573 7877422.1 0.936276
1998 352/652 Cotton & MMF Underwear etc. DOZ 9970725 9210810.8 0.923786
1999 352/652 Cotton & MMF Underwear etc. DOZ 9412245 8220331.9 0.873366
2000 352/652 Cotton & MMF Underwear etc. DOZ 1.29E+07 12293112 0.953438
2001 352/652 Cotton & MMF Underwear etc. DOZ 1.36E+07 13600364 1
2002 352/652 Cotton & MMF Underwear etc. DOZ 1.72E+07 16746916 0.972925
2003 352/652 Cotton & MMF Underwear etc. DOZ 2.02E+07 19128188 0.944741
2004 352/652 Cotton & MMF Underwear etc. DOZ 2.29E+07 22856951 1
1991 360 Cotton Pillowcases NO 1391385 1391384.7 1
1992 360 Cotton Pillowcases NO 1659218 1574308.8 0.948826
1993 360 Cotton Pillowcases NO 1592996 1592996.4 1
1994 360 Cotton Pillowcases NO 1924688 1902252.6 0.988344
1995 360 Cotton Pillowcases NO 2080972 2073618.9 0.996467
1996 360 Cotton Pillowcases NO 3680633 3378957.3 0.918037
1997 360 Cotton Pillowcases NO 4413190 4187838.6 0.948937
1998 360 Cotton Pillowcases NO 4840267 4840266.6 1
1999 360 Cotton Pillowcases NO 5736731 5736731.4 1
2000 360 Cotton Pillowcases NO 6014405 6014404.8 1
2001 360 Cotton Pillowcases NO 6624866 6624865.8 1
2002 360 Cotton Pillowcases NO 7668605 7668604.8 1
2003 360 Cotton Pillowcases NO 9081286 9081286.2 1
2004 360 Cotton Pillowcases NO 9454337 9454337.1 1
1991 361 Cotton Sheets NO 1.15E+07 11460452 1
1992 361 Cotton Sheets NO 1.30E+07 12950309 1
1993 361 Cotton Sheets NO 1.24E+07 12433444 1
1994 361 Cotton Sheets NO 1.45E+07 14460732 1
1995 361 Cotton Sheets NO 1.56E+07 15634939 1
1996 361 Cotton Sheets NO 2.29E+07 18597389 0.811838
1997 361 Cotton Sheets NO 2.78E+07 25280960 0.909474
1998 361 Cotton Sheets NO 3.44E+07 33095868 0.962913
1999 361 Cotton Sheets NO 3.85E+07 38541344 1
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
53
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates (Continued)
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
2000 361 Cotton Sheets NO 4.04E+07 40406844 1
2001 361 Cotton Sheets NO 4.45E+07 44508136 1
2002 361 Cotton Sheets NO 5.01E+07 50126669 1
2003 361 Cotton Sheets NO 5.62E+07 56164092 1
2004 361 Cotton Sheets NO 6.01E+07 60097534 1
1991 363 Cotton Terry & Other Pile Towels NO 1.17E+07 11689698 1
1992 363 Cotton Terry & Other Pile Towels NO 1.47E+07 14710422 1
1993 363 Cotton Terry & Other Pile Towels NO 1.38E+07 13844720 1
1994 363 Cotton Terry & Other Pile Towels NO 1.54E+07 15357094 1
1995 363 Cotton Terry & Other Pile Towels NO 1.62E+07 16230249 0.998919
1996 363 Cotton Terry & Other Pile Towels NO 1.76E+07 17588729 1
1997 363 Cotton Terry & Other Pile Towels NO 1.86E+07 18594367 1
1998 363 Cotton Terry & Other Pile Towels NO 1.98E+07 19793857 1
1999 363 Cotton Terry & Other Pile Towels NO 2.10E+07 20999250 1
2000 363 Cotton Terry & Other Pile Towels NO 2.25E+07 22521696 1
2001 363 Cotton Terry & Other Pile Towels NO 2.42E+07 24154519 1
2002 363 Cotton Terry & Other Pile Towels NO 2.64E+07 26403174 1
2003 363 Cotton Terry & Other Pile Towels NO 2.88E+07 28834247 1
2004 363 Cotton Terry & Other Pile Towels NO 2.97E+07 29271308 0.984514
1991 369 Shop Towels Only KG 3688660 3688660 1
1992 369 Shop Towels Only KG 4165145 4165144.5 1
1993 369 Shop Towels Only KG 4456703 4456703 1
1994 369 Shop Towels Only KG 4456703 4256052 0.954978
1995 369 Shop Towels Only KG 5155888 4682446 0.908175
1996 369 Shop Towels Only KG 5675884 5675883.5 1
1997 369 Shop Towels Only KG 6144047 6144046.5 1
1998 369 Shop Towels Only KG 6780552 6780552 1
1999 369 Shop Towels Only KG 7363168 733167.5 0.099572
2000 369 Shop Towels Only KG 8096709 8096709 1
2001 369 Shop Towels Only KG 8918523 8918523 1
2002 369 Shop Towels Only KG 1.01E+07 10080558 1
2003 369 Shop Towels Only KG 1.14E+07 11379970 1
2004 369 Shop Towels Only KG 1.21E+07 12131115 1
1991 615 MMF Print Cloth Fabric M2 1.42E+07 14187864 1
1992 615 MMF Print Cloth Fabric M2 1.49E+07 14935279 1
1993 615 MMF Print Cloth Fabric M2 1.76E+07 13794085 0.78531
1994 615 MMF Print Cloth Fabric M2 2.00E+07 13475023 0.673025
1995 615 MMF Print Cloth Fabric M2 1.78E+07 10141540 0.569823
1996 615 MMF Print Cloth Fabric M2 1.94E+07 14184923 0.730959
1997 615 MMF Print Cloth Fabric M2 2.56E+07 22730616 0.889082
1998 615 MMF Print Cloth Fabric M2 2.56E+07 25632933 1
1999 615 MMF Print Cloth Fabric M2 2.87E+07 26963151 0.940312
2000 615 MMF Print Cloth Fabric M2 2.83E+07 26330205 0.929341
2001 615 MMF Print Cloth Fabric M2 3.79E+07 37853501 1
2002 615 MMF Print Cloth Fabric M2 3.83E+07 36837156 0.962278
2003 615 MMF Print Cloth Fabric M2 3.77E+07 27696697 0.735485
2004 615 MMF Print Cloth Fabric M2 4.90E+07 25816627 0.527164
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
54
Table 3.5: Sample OTEXA Categories – Adjusted Base, Imports & Fill Rates (Continued)
Year
MFA
Root
OTEXA Category Description
Native
Units
Adjusted Base
(SME)
Imports
(SME)
Fill Rate
1991 638/639 MMF KN Shirts & Blouses DOZ 1796113 626356.8 0.348729
1992 638/639 MMF KN Shirts & Blouses DOZ 981007.2 981007.2 1
1993 638/639 MMF KN Shirts & Blouses DOZ 1517175 1219419.4 0.803743
1994 638/639 MMF KN Shirts & Blouses DOZ 520253.3 520253.28 1
1995 638/639 MMF KN Shirts & Blouses DOZ 1429216 1429215.8 1
1996 638/639 MMF KN Shirts & Blouses DOZ 976212 976212 1
1997 638/639 MMF KN Shirts & Blouses DOZ 1228789 1228789.4 1
1998 638/639 MMF KN Shirts & Blouses DOZ 2448157 860764.32 0.351597
1999 638/639 MMF KN Shirts & Blouses DOZ 3260684 3252.96 0.000998
2000 638/639 MMF KN Shirts & Blouses DOZ 4240629 4240628.6 1
2001 638/639 MMF KN Shirts & Blouses DOZ 2368803 1903253.8 0.803467
2002 638/639 MMF KN Shirts & Blouses DOZ 4843048 3232677.6 0.667488
2003 638/639 MMF KN Shirts & Blouses DOZ 6378536 6378536.2 1
2004 638/639 MMF KN Shirts & Blouses DOZ 8429119 6944680.8 0.823892
1996 666 MMF Pillowcases ex. Bolsters KG 7456867 7456867.2 1
1997 666 MMF Pillowcases ex. Bolsters KG 1.13E+07 11178763 0.991541
1998 666 MMF Pillowcases ex. Bolsters KG 1.24E+07 12432586 1
1999 666 MMF Pillowcases ex. Bolsters KG 1.47E+07 14678395 1
2000 666 MMF Pillowcases ex. Bolsters KG 1.24E+07 12418099 1
2001 666 MMF Pillowcases ex. Bolsters KG 1.70E+07 17020152 1
2002 666 MMF Pillowcases ex. Bolsters KG 1.56E+07 15551554 1
2003 666 MMF Pillowcases ex. Bolsters KG 1.85E+07 18517118 1
2004 666 MMF Pillowcases ex. Bolsters KG 1.99E+07 19867450 1
1996 666 MMF Sheets KG 3.43E+07 34322674 1
1997 666 MMF sheets KG 5.74E+07 54240566 0.944143
1998 666 MMF Sheets KG 6.68E+07 66772613 1
1999 666 MMF Sheets KG 6.89E+07 68866315 1
2000 666 MMF Sheets KG 7.72E+07 77178125 1
2001 666 MMF Sheets KG 7.25E+07 72539179 1
2002 666 MMF Sheets KG 8.94E+07 89401234 1
2003 666 MMF Sheets KG 9.80E+07 98031859 1
2004 666 MMF Sheets KG 1.02E+08 1.02E+08 1
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
55
Figure 3.2: Level of Imports and Adjusted Quota Base (Examples)
Source: U.S. MFA/ATC Database (Brambilla et al., 2007)
20 40 60 80 100 120
Square meter equivalents ( x 1000000)
1985 1990 1995 2000 2005
Year
Imports
Adjusted quota base
NAICS 313312: Textile and Fabric Finishing
0 10 20 30
Square meter equivalents ( x 1000000)
1985 1990 1995 2000 2005
Year
Imports
Adjusted quota base
NAICS 315222: Men’s and Boys’ Cut and Sew Suit, Coat, and Overcoat
56
3.3 Review of Literature
As observed in chapter 2, a variety of studies look into the efficiency of manufacturing
industries as a result of trade liberalization. These studies focus on liberalization that
primarily comprised of a reduction in tariff rates or a fall in trade costs. There is limited
evidence on the effect of a liberalization regime that mainly consisted of an increase in
the amount of quota, for example, as in the case of the end of MFA. A sizeable number of
studies attempt to examine the impact of MFA expiration on the reallocation of
production and exports across countries. Using a time series of product-level data from
the U.S. on quotas and tariffs that the MFA comprised of, Evans and Harrigan (2005)
analyze how MFA affected the sources and prices of U.S. apparel imports, with a
particular focus on the East Asian exporters during 1990s. They demonstrate that
although a huge fraction of U.S. apparel is imported under binding quotas, there are many
quotas that stay unfilled. In addition, the binding quotas substantially lift import prices,
suggestive of both quality upgrading and rent capture by exporters. Brambilla et al.
(2007) examine China's experience under the U.S. apparel and textile quotas. They
observe that China was somewhat more constrained under these regimes than other
countries and that, as quotas were lifted, China's exports grew noticeably. Diao and
Somwaru (2001) foresee that in the post-quota world, apparel trade will rise twice as
rapidly as textile trade. Production will be reallocated away from high-cost countries that
were producing merely to benefit from the quota. Pakistan, South Korea, Hong Kong,
Taiwan, Vietnam, and several other countries with privileged access to the U.S. and E.U.
markets will gain (Jones, 2003).
57
Lankes (2002) simulates a quota removal model and predicts reallocation of
production to the disadvantage of exporters in developing countries that were protected
by the quota system. Elimination of quotas benefits the consumers as a result of lower
prices (Tyagi, 2003). In 2000, the U.S. abandoned trade restrictions on a list of products
as a result of the African Growth and Opportunity Act (AGOA). Frazer and Biesebroeck
(2007) investigate whether AGOA achieved its desired results. In view of the fact that the
Act was applied selectively to both the countries and products, they approximate the
effect with triple difference-in-difference estimation. This methodology deals with the
‘endogeneity of policy’ critique of the ordinary difference-in-difference estimation.
AGOA had a great and robust impact on apparel imports into the U.S. (Frazer &
Biesebroeck, 2007). It did not give rise to a reduction in exports to Europe, implying that
the AGOA exports were not simply diverted from other destinations.
Ernst, Ferrer and Zult (2005) focus on the exporting developing countries to
illustrate the evolution of trade and employment in T&C until 2005. The forecast of the
gravity model confirms that the phasing out entails noteworthy changes in the global
trade structure. According to the paper, China and Pakistan will be the biggest winners.
There will be ‘slight’ losers but may well be winners if they employ the right policies,
namely, modernization of production, creation of political or enterprise alliances with
leading countries, and the integration into global production systems. Third, there are
loser countries which may perhaps have the capacity to survive in niches by applying
definite restructuring strategies. Mexico and other Central American States come under
58
this category.
29
There is a fourth category of countries that will entirely fail to benefit
from MFA expiration, for instance, small and less developed countries formerly
benefiting from access to the U.S. and E.U. markets (Ernst, Ferrer, & Zult, 2005). Haté et
al. (2005) study the consequences of the abolition of MFA for South Asian countries.
Due to small firm size, low labor productivity, high cost of transportation, lack of
government support and political instability, Nepal is expected to endure losses in the
T&C industry without the protection of quotas. The consequences for Bangladeshi and
Sri Lankan T&C industry are vague. Even though Bangladesh has captured niche markets
and the wages are competitive globally, Bangladesh’s heavy reliance on imported
material is a concern. India and Pakistan are likely to be winners with the expiration of
the MFA. The private sector in Pakistan seems prepared to profit from the domestic raw-
material base in cotton and synthetic fibres, low labor costs, and large-scale investment in
the last number of years. Also, the T&C industry has benefited considerably from
complimentary trade agreements with the U.S. and E.U. since 2001 with regard to the
fight against terrorism.
These studies pertain to the macroeconomic outcomes of the end of MFA, and do
not consider the impact on textile producing firms. Using Bangladeshi garment exporters’
data, Demidova, Kee, and Krishna (2006) model and present evidence for the pattern of
exports and performance of heterogeneous firms in response to variations in trade policy
in diverse product and export destinations. A study by Sasidaran and Shanmugam (2008)
attempts to empirically investigate the implications of the end of MFA on firm efficiency
29
Countries that may gain from niches attributable to their proximity to the E.U. market are Romania,
Turkey, Morocco and Egypt (Ernst, Ferrer and Zult, 2005).
59
in the Indian textile industry. By employing stochastic coefficients frontier approach,
they estimate the overall and input specific efficiency values for a sample of 215 firms
from 1993 to 2006. Results of the analysis illustrate that the average efficiency dropped
over the years. However, their empirical methodology does not utilize the actual number
of quotas imposed by the developed countries on import of T&C products from India,
and instead models the end of MFA by introducing a dummy variable for each of the four
phases. This work, on the other hand, uses an exceptional database initially used by
Brambilla et al. (2007), which traces U.S. trading partners’ exports to the U.S. in addition
to the actual amount of quota under the regimes determined by the MFA and ATC. This
source of data is combined with a unique company-level data set which is a compilation
of annual reports of a representative sample of T&C companies in Pakistan. Hence, the
paper merges micro-level data of firms with the data on quotas at the industry level in
order to scrutinize a fundamental issue in the new trade theory.
A large number of studies that analyze the impact of trade liberalization on firm
performance are repeatedly criticized for endogeneity inherent in either the estimation of
productivity or in the principal regression model used to regress the performance variable
on a proxy for trade liberalization, such as the tariff rate. Hence, the relationship between
openness and performance cannot be taken to imply causality. This is usually the case
because liberalization is more than often a part of a broader package of reforms;
improvement in firm efficiency cannot be traced to trade reforms specifically. Moreover,
even if trade reforms do not come as a part of a package of reforms, there is always a
possibility of lobbying by firms in order to circumvent these reforms whenever they are
60
expected to harm these firms. This is widespread in the case of developing countries.
Because of the regression specification used in the chapter, whereby we regress the
change in firm productivity on the adjusted level of quotas at the six-digit NAICS
industry level, we can rule out the possibility of lobbying by firms. This is because it is
not viable for an individual firm to influence the amount of textile quota at the industry
level.
30
Consequently, the MFA expiration can be thought of as a ‘natural experiment.’
Last but not the least, we use structural techniques proposed by Olley and Pakes, and
Levinsohn and Petrin in order to take care of endogeneity in the estimation of production
functions.
In short, the contribution of this chapter is that it is one of the very few studies
that investigate the effect of liberalization in the form of the phasing out of quotas on
firm-level productivity in the textile and clothing industry. It underlines cross-sector
disparity in the effect of MFA expiration and that trade reforms can influence different
sectors heterogeneously even within the manufacturing industry.
3.4 Empirical Methodology
In this section, we discuss the empirical methodology used to measure the impact of end
of MFA on firm performance in the textile and apparel industries of Pakistan from 1992
to 2010. Let us first take a look at the consequence of end of MFA on the level of T&C
imports into the country. In order to do that, we estimate a simple regression of industry
30
In fact, firms frequently organize together to affect trade policy outcomes, both formally through industry
associations, as well as informally. Since the firms may have collectively organized to affect the overall
level of the initial quota, the initial level of quota may have been endogenous. Therefore, the greater the
degree to which the U.S. is setting these quotas (and the timing of their changes) externally, the more likely
they are to be exogenous. Thus, the changes to these initial quotas under the ATC will not be endogeous.
61
imports at the six-digit NAICS product-level on the post-MFA adjusted level of quotas
and industry trade costs:
log c!/(
=
+
+
log Ide&/'
+
-
log f/g
`+
+ h
+ h
+ 4
, (3.1)
where log c!/(
is the logarithm of total imports of product j at time t,
log Ide&/'
is the logarithm of the adjusted level of quotas, and log f/g
`+
is the
logarithm of industry trade costs at date t-1. h
and h
are the time and industry fixed
effects, respectively, and 4
is the error term. We would expect
+
to be positive if an
increase in the adjusted level of quotas leads to an increase in the level of imports.
Furthermore, we would expect
-
to be negative if imports go down as a result of an
increase in trade costs. Following Bernard et al. (2006), we define industry variable trade
costs as the sum of ad valorem duty and ad valorem freight and insurance rates.
31
The
inclusion of non-tariff barriers (NTBs) such as quotas in the regression equation, unlike
31
Bernard et al. (2006) define variable trade costs (f/g
) for industry j in year t as the sum of ad valorem
duty (I
) and ad valorem freight and insurance (
) rates:
f/g
= I
+
(3.2)
The ad valorem duty rate is duties collected (I&$g
) corresponding to free-on-board customs value of
imports ( /i
):
I
=
8jkl
;<
Gmn
;<
(3.3)
Likewise, the ad valorem freight rate is the markup of the cost-insurance-freight value (o$
) over /i
relative to /i
:
=
pG
;<
Gmn
;<
− 1 (3.4)
The rate for industry j is the weighted average rate across products in j, using the import values from the
source countries as weights. This measure of trade costs has several advantages. It includes information
concerning both trade policy and transportation costs, and it varies across industries and over time.
62
Bernard et al. (2006), is an added advantage of this empirical methodology since NTBs
are a vital source of trade distortions.
32
To determine the effect of trade liberalization on firm performance, we first need
to find a measure of productivity for each firm in our sample. Productivity changes can
be measured using either production functions or cost functions. Nevertheless for
numerous reasons, the two approaches are not equal.
33
This measure is then linked to an
index of openness using a simple regression equation. Ackerberg, Caves, and Frazer
(2005) study the recent literature on the empirical identification of production functions.
An econometric issue facing estimation of production functions is the likelihood that
some of these inputs are unobserved. If the observed inputs are chosen as a function of
these unobserved inputs, then there is an endogeneity problem (Ackerberg et al., 2005).
34
The correlation between unobservable productivity shocks and inputs chosen by the firm
is a recognized difficulty in the estimation of production function. A second endogeneity
problem appears because of sample selection. Firms exit the market when productivity
drops below a particular threshold. Therefore, the OLS estimates of observed input
coefficients will be biased. Two ways of taking care of the endogeneity problem are
32
For a complete discussion of the advantages and disadvantages of this measure, see Bernard et al. (2006).
For example, changes in the composition of products or importers within industries can bring about
variation in I
and
although the actual statutory tariffs and market transportation costs remain stable.
33
Despite the fact that both methodologies require output data, cost functions exploit cost and factor price
data, while production functions exploit data on physical inputs. Simultaneity between output and the error
term is a problem with cost functions; simultaneity between inputs and the error term is a problem with
production functions (Tybout & Westbrook, 1995). Similarly, measurement error in factor prices or output
can bias cost function estimates, while measurement error in the flow of factor services can bias production
functions (Tybout & Westbrook, 1995).
34
Ackerberg, D. A., Caves, K., & Frazer, G. (2005). Structural Identification of Production Functions.
Unpublished.
63
instrumental variables (IV) and fixed effects estimation (Mundlak, 1961).
35
A group of
contemporary techniques go along the dynamic panel data literature and the methods
introduced by Olley and Pakes (1996), and Levinsohn and Petrin (2003).
36
The OP
methodology allows the error term to have two components: a white noise component
and a time-varying productivity shock. It is derived from dynamic optimization of firms,
whereby it is assumed that unobserved productivity follows a first order Markov process,
and capital is accumulated by means of a deterministic dynamic investment process
(Amiti & Konings, 2007). Profit maximization generates an investment demand function
that is determined by two state variables, capital and productivity. If the investment
demand function is monotonically increasing in productivity, it is feasible to invert the
investment function and get an expression for productivity as a function of capital and
investment (Pakes, 1994). LP adopts a similar approach to solving the endogeneity
problem. Instead of using an investment demand equation, they use an intermediate input
demand function. In the real data, investment is often lumpy. This may not be in line with
the strict monotonicity assumption regarding investment. Also, OP procedure can cause
efficiency loss in a data with a large number of observations with no investment. Given
35
IV estimation involves finding variables that are correlated with observed input choices but uncorrelated
with the unobserved inputs. Fixed-effects estimation is based on the assumption that the unobserved input
or productivity is stable over time (Mundlak, 1961).
36
Biesebroeck (2007) compares the robustness of five commonly used techniques, two non-parametric and
three parametric: (a) index numbers, (b) data envelopment analysis (DEA), (c) stochastic frontiers, (d)
instrumental variables (GMM), and (e) semi-parametric estimation. Parametric estimation (OLS) ignores
the simultaneity of unobserved productivity and input choices, and leads to upwardly biased parameter
estimates. When firms are exposed to idiosyncratic productivity shocks that are not totally transitory, a
semi-parametric estimator (such as OP) will make use of firm’s knowledge about these shocks.
64
that the intermediate input demand normally exhibits a lesser tendency to have zeros, the
strict monotonicity condition is expected to hold in the LP methodology.
In this section, we use the structural techniques proposed by Levinsohn and
Petrin.
37
Consider a firm with a Cobb-Douglas production function, as in Section 2.2:
=
, (3.5)
where output of firm i in six-digit industry j at time t,
, is a function of labor,
,
capital,
, and materials,
. We want to know if productivity of firm i is a function
of trade policy, denoted by . Taking natural logs, denoted by small letters, we get:
=
+
+
+
+
. (3.6)
Productivity is then computed using LP, and the change in firm productivity is regressed
on the change in adjusted level of quotas and trade costs:
∆ !
=
+
+
∆log Ide&/'
+
-
∆log f/g
`+
+ r
+ h
+ h
+ 4
, (3.7)
where r
includes other control variables: a dummy variable for the city in which the
firm is located, size, age and capital intensity of the firm, whether or not the firm is ISO
Certified, whether or not the firm is multinational, and lastly, herfindahl index of the
industry at the six-digit level. Size is measured by the number of workers, and capital-
intensity is the ratio of capital to number of employees.
38
Age of the firm simply
measures the number of years since establishment. Herfindahl index is an indicator of the
37
See Olley and Pakes (1996) and Levinsohn and Petrin (2003) for a complete explanation of the method.
A brief review is also given in Appendix B.
38
In order to avoid the endogeneity problem potentially arising due to the regression of productivity on
number of employees and firm’s capital, we shall use lagged values of size and capital intensity of the firm.
65
extent of competition in the industry. In order to quantify the impact of quotas directly on
firm’s output, we regress output on the level of quotas:
=
+
+
+
-
+
Q
+
R
logIde&/'
+
S
logf/g
`+
+r
+ h
+ h
+ 4
. (3.8)
Table 3.6: Summary Statistics
Variable Observations Mean Standard Deviation
Ln(Sales) 4717 19.25 3.73
Ln(Fixed Assets) 4718 11.50 9.51
Ln(Labor) 4718 16.36 1.93
Ln(Raw Materials) 4718 18.71 3.58
Ln(Net Profit) 4718 12.99 10.32
Ln(Investment) 4813 4.02 7.22
Productivity (Levinsohn and Petrin) 4717 10.55 5.72
Productivity (Olley and Pakes) 4717 1.87 3.04
Age 2895 23.78 16.10
Ln(Age) 2846 2.97 0.82
Ln
2
(Age) 2846 9.48 4.31
Ln(Capital to Labor ratio) 4407 0.73 0.58
Herfindahl Index 4813 0.82 0.62
ISO Certified 4606 0.67 0.47
Multinational 4606 0.10 0.30
Share of Foreign Ownership 4436 0.22 0.41
Exporting firm 4606 0.88 0.33
Importing firm 4606 0.42 0.49
Ln(Cost of Imports) 2385 0.15 0.11
Ln(Adjusted Base New) 3980 29.11 16.11
Ln(Adjusted Base) 2499 16.73 1.13
Ln(Imports) 1544 16.43 2.01
Average Fill Rate 2143 0.81 0.19
The BSDPC data set used in the last chapter contains data for 321 textile and
clothing companies for the years 1992 to 2010.
39
The dataset covers almost all large and
medium-sized formal T&C enterprises in Pakistan. We estimate Eqs. (3.7) and (3.8)
39
The data compiled by CMER only covers the period 1992 to 2003. We updated the dataset to add seven
more years of data on sales revenue, input use, investment, etc. The chapter, therefore, uses data from 1992
to 2010. This was done in order to compute firms’ productivity during the final phase of MFA expiration as
well, since we know that the U.S. postponed removal of quotas on sensitive products until Phase III.
66
using data for nominal output, net fixed assets, total wage bill, raw materials, and their
respective deflated values.
40
Table 3.6 provides summary statistics for the balance sheet
data used. This paper is based on a panel of firms instead of industry data. Accordingly,
we can be fairly specific about the sources of productivity change. It tracks a single
country through time, eliminating the obscuring country-specific effects. There is a
notable variation in trade regimes between sample years which creates sufficiently great
changes in industrial performance to be detectible.
The paper utilizes the data initially used by Brambilla et al. (2007) that traces U.S.
trading partners’ performance under the quota regimes determined by MFA and ATC.
The database is assembled from U.S. trading partners’ Expired Performance Reports,
which were used by the U.S. Office of Textile and Apparel (OTEXA) to supervise
trading partners’ fulfillment with the MFA/ATC quotas. Provided by Ron Foote of the
U.S. Census Bureau, they record imports, base quotas and quota adjustments by OTEXA
category and year for all the countries with which the U.S. negotiated a bilateral quota
arrangement.
41
The U.S. negotiated quotas on 149 three-digit OTEXA specific limit
categories, where each category is an aggregate of roughly 30 HS products. The OTEXA
categories cover four T&C segments: Yarn, Fabric, Made-ups and Clothing. The
negotiated quota for any given category is stated in terms of square meter equivalents
40
The Economic Survey of Pakistan, which is published annually by the Federal Bureau of Statistics,
Pakistan, provides price indices at the two-digit industry level for output and intermediate inputs which are
used as deflators.
41
The base quota is the initially negotiated quota level decided at the beginning of an agreement term.
Adjusted base quotas indicate the use of “flexibilities,” which allowed countries to go over their base quota
in a particular period by borrowing unexploited base quota, across categories within a year and across years
within a category, up to a specified percentage of the receiving category.
67
(SME) of fabric. The data on trade costs is taken from Bernard et al. (2006) which
provides data on free-on-board customs value of imports, ad valorem duty, freight and
insurance rates for the underlying four-digit product-level U.S. import data compiled by
Feenstra (1996).
42
The next section discusses the estimation results.
3.5 Industry-level results: Effect on T&C exports
Let us first look at the effect of the end of MFA on level of T&C imports into the U.S. To
do that, we estimate a simple regression of the logarithm of total industry imports at the
six-digit NAICS product-level on the logarithm of post-MFA adjusted level of quotas and
industry trade costs, taking into account time and industry fixed effects. The results are
depicted in Table 3.7. As expected,
+
is positive and significant. An increase in the
adjusted level of quotas leads to an increase in the level of imports. This is true for the
MFA period as well as post-MFA period. Moreover, the coefficient for textile industry is
higher than that for the clothing industry. In addition,
-
is negative and significant in
most of the cases. As anticipated, the total imports drop as a result of an increase in trade
costs. In this case, the coefficient for textile industry is lower than that for the clothing
industry. These results hold by and large for the entire sample as well as across Phases 1
and 3, and for both industries. However,
+
is negative and significant in Phase 2. Phase 2
was the time period from January 1
st
, 1998 to January 1
st
, 2002. Pakistan underwent
considerable political and economic changes during this period. Due to nuclear tests in
May 1998, followed by military take-over in 1999, various sanctions were imposed on
42
The data on trade costs is available only for the years 1992-2004.
68
Pakistan by a number of countries. Consequently, international trade and foreign direct
investment were particularly adversely affected in this time period. Hence, despite the
increase in quotas, there was a reduction in the level of imports.
In general, we observe that in spite of an increase in the adjusted level of quotas
taking place simultaneously for a group of competing developing countries, there is a rise
in exports by the T&C industry of Pakistan; there is a noticeable evidence for trade
creation, and not trade diversion. For this reason, it would be interesting to investigate
the impact of MFA expiration on T&C firms in Pakistan.
3.6 Firm-level results: Effect on productivity
We estimate the production function coefficients for firms in each sector separately using
a Cobb-Douglas production function and the structural techniques proposed by
Levinsohn and Petrin. These estimates are used to work out the measured TFP of firm i at
time t for each six-digit industry j. The change in firm productivity is then regressed on
the change in adjusted level of quotas, allowing for time and industry fixed effects. Table
3.8 reports production function estimates for T&C firms using LP. Robust standard errors
corrected for clustering at the firm level are stated in parentheses. The regression results
are illustrated in Tables 3.9 and 3.10.
69
Table 3.7: Effect of Elimination of Quota-Restrictions on the Level of Imports (1993-2003 & across Phases 1, 2 and 3)
Entire Sample MFA Period Post-MFA Period
All T&C Textile Clothing All T&C Textile Clothing All T&C Textile Clothing
Adjusted Quota Level 1.048*** 1.030*** 1.075*** 0.958*** 0.926*** 1.021*** 1.064*** 1.049*** 1.082***
(0.00593) (0.00253) (0.0166) (0.00726) (0.00181) (0.0191) (0.00709) (0.0027) (0.0179)
Cost of Imports 0.00748 0.0246 0.483 0.197* -0.0182 0.0956 -0.051 -0.033 0.945
(0.053) (0.0203) (0.693) (0.107) (0.0143) (0.376) (0.0588) (0.0205) (0.832)
Constant -0.901*** -1.490*** 0.457*** -0.424 -1.182*** -0.999*** -1.657***
(0.105) (0.279) (0.122) (0.306) (0.125) (0.0462) (0.299)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
No. of Observations 2397 1763 634 408 310 98 1989 1453 536
No. of firms 315 225 90 225 171 54 306 217 89
Phase 1 Phase 2 Phase 3
All T&C Textile Clothing All T&C Textile Clothing All T&C Textile Clothing
Adjusted Quota Level 0.972*** 0.975*** 0.807*** 1.132*** 1.098*** 1.187*** 1.063*** 1.041*** 1.050***
(0.0109) (0.00425) (0.0261) (0.00977) (0.0038) (0.0259) (0.0112) (0.00318) (0.0397)
Cost of Imports -0.0309 -0.0143 0.581 -0.436 0.341*** -2.257 0.0117 0.0240** 1.980*
(0.0642) (0.0252) (0.725) (0.321) (0.105) (1.384) (0.0264) (0.0112) (1.044)
Constant 0.484*** 3.155*** -1.129*** -1.119*
(0.183) (0.433) (0.194) (0.642)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
No. of Observations 666 488 178 917 663 254 406 302 104
No. of firms 274 199 75 276 199 77 227 166 61
4otes: Robust standard errors corrected for clustering at the industry level in parentheses. This table reports the effect of elimination of quota-restrictions on the level of
imports across phases. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
70
3.6.1 Adjusted level of quotas and productivity
The results vary across the two types of industries: an increase in the adjusted level of
quotas, on average, brings about a significant increase in the productivity of textile firms
and a reduction in the mean productivity of firms in the clothing industry. These
estimation results are derived after controlling for the firm’s size, capital intensity, age,
whether or not the firm is ISO Certified, whether or not the firm is multinational,
herfindahl index of the industry, and finally, the city in which the firm is located.
43
Table 3.8: Production Function Estimates for Textile and Clothing Firms – Levinsohn & Petrin
Textile Clothing
Employment 0.246*** 0.285***
(0.0313) (0.0327)
Fixed Assets 0.0312*** 0.0340**
(0.00805) (0.0152)
Raw Materials 0.125 0.171
(0.116) (0.160)
No. of observations 3274 1443
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. *** Significant at, or below, 1
percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
Furthermore, we run this regression separately for MFA period (1992-1994) and
the post-MFA period (1995-2010), along with each of the four phases individually.
44
Table 3.11 demonstrates estimation results for the four phases. In all the phases, an
43
In addition, when the quotas are completely removed in Phase IV, the adjusted quota base is essentially
equal to infinity. There are a number of possible ways of handling it. For example, we could assume a ‘very
large’ value of the adjusted level of quotas, and vary that value to test if our results are sensitive to this
hypothetical value of the adjusted level of quotas. Another possible way is to predict the adjusted quota
level using the past values of the fill rates. A number of these methods were used in order to prove that the
results are robust to functional form differences.
44
Estimation results for the MFA and post-MFA periods alone are shown in Appendix B.
71
increase in the adjusted level of quotas brings about a significant reduction in clothing
firm’s productivity and an increase in the productivity of firms in textile industry. This is
also true for the post-MFA period as a whole. Only in Phase IV do we observe that the
productivity of clothing firms is positively related to the level of quotas. Nevertheless, the
positive coefficient is not statistically significant. For a majority of the control variables
described above, we do not observe a noticeable change in either the sign or the
magnitude of coefficients.
Table 3.9: Effect of Elimination of Quota Restrictions on Textile Firm Productivity – Levinsohn & Petrin
Variable (1) (2) (3) (4) (5) (6) (7)
Adjusted Quota 0.0238*** 1.277** 1.266** 1.250** 1.192** 1.567* 1.692**
(0.00520) (0.534) (0.530) (0.557) (0.535) (0.875) (0.850)
Cost of Imports -0.126 -0.124 -0.120 -0.122 0.0965 0.0971
(0.225) (0.225) (0.223) (0.237) (0.175) (0.173)
Herfindahl Index 0.0619 0.0602 0.0596 0.0673 0.0924* 0.0971*
(0.0509) (0.0507) (0.0509) (0.0501) (0.0547) (0.0566)
Multinational 0.410* 0.215 0.149 0.0126 0.162
(0.234) (0.206) (0.200) (0.192) (0.261)
ISO Certified 0.830*** 0.827*** 1.020* 0.839
(0.176) (0.169) (0.578) (0.574)
K/L (-1) -0.0333 -0.0709 -0.0696
(0.158) (0.0823) (0.0883)
Size (-1) 0.0474* -0.0246 -0.0273
(0.0282) (0.0203) (0.0198)
Age 0.118 0.117
(0.206) (0.222)
Age
2
0.0262 0.0346
(0.0430) (0.0510)
Constant 11.47*** -12.03 -11.80 -12.47 -11.96 0 0
(0.305) (10.03) (9.973) (10.50) (10.20) (0) (0)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes 4o Yes
No. of Observations 2767 1570 1570 1570 1567 996 996
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
72
Table 3.10: Effect of Elimination of Quota Restrictions on Clothing Firm Productivity – Levinsohn & Petrin
Variable (1) (2) (3) (4) (5) (6)
Adjusted Quota -0.972*** -1.003*** -0.998*** -1.069*** -1.692*** -0.753***
(0.246) (0.248) (0.248) (0.255) (0.327) (0.195)
Cost of Imports (-1) -8.697 -8.787 -8.793 -9.737* -11.70 -11.22
(6.040) (6.041) (6.051) (5.796) (7.886) (8.823)
Herfindahl Index (-1) -0.155** -0.155** -0.192** -0.241*** -0.182**
(0.0719) (0.0720) (0.0765) (0.0879) (0.0782)
Multinational -0.773 -0.749 -4.371*** -2.368
(1.546) (1.572) (1.538) (1.981)
ISO Certified 0.403 1.097 1.719
(1.148) (1.943) (2.174)
K/L (-1) 0.946* 0.969 0.807
(0.563) (0.659) (0.700)
Size (-1) 0.0885* 0.117** 0.0716
(0.0458) (0.0536) (0.0572)
Age 0.669 0.104
(0.490) (0.513)
Age
2
-0.0968 0.277
(0.196) (0.248)
Constant 0 0 14.13*** 16.59*** 26.45*** 0
(0) (0) (4.739) (4.371) (6.223) (0)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes 4o Yes
No. of Observations 503 503 503 502 315 315
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
3.6.2 Variable trade costs and productivity
Although a decrease in trade costs does not seem to have a significant impact on the
textile firms, there is clearly a negative relationship between trade costs and productivity
of garment producers. The productivity of clothing firms goes up, on average, if trade
costs go down, and the estimates are significantly different from zero in a number of
cases. As far as the trade costs coefficient for textile firms is concerned, the estimates
take both positive and negative values, and none of these are statistically significant. The
positive coefficient of trade cost for textile producers may perhaps be indicative of a
73
selection effect for these types of firms, as is highlighted in the literature on new trade
theory. This suggests that as a consequence of a rise in variable trade cost, coupled with
exposure to international competition, only the most productive firms are able to survive.
As a result, an upsurge in trade cost will cause the mean productivity of textile producers
to go up.
Table 3.11: Effect of Elimination of Quota Restrictions on Firm Productivity – Across phases
Variable Phase 1 Phase 2 Phase 3 Phase 4
Textile Clothing Textile Clothing Textile Clothing Textile Clothing
Adjusted Quota 0.862 -0.466 0.845 -1.424** 6.039*** -2.291* 0.0546 0.0200
(0.539) (0.362) (0.673) (0.682) (0.890) (1.230) (0.108) (0.155)
Cost of Imports 0.00578 4.801 -6.675 5.707 2.009 -1.788 - -
(0.289) (3.895) (4.732) (5.436) (6.765) (8.671) - -
Age 0.364 1.524* -0.193 0.00114 2.098 0.0364 4.530 -4.135
(0.449) (0.840) (1.007) (0.283) (2.655) (0.867) (3.088) (3.293)
Age
2
-0.00151 -0.171 0.0873 0.0409 -0.319 0.0312 -0.629 1.279
(0.0870) (0.380) (0.188) (0.240) (0.411) (0.433) (0.457) (0.859)
Size (-1) 0.0271 0.106* 0.0196 0.119 0.0985 0.122*** 0.0348* 0.0370
(0.0328) (0.0642) (0.0284) (0.0737) (0.0645) (0.0372) (0.0188) (0.0293)
K/L (-1) -0.256** 0.618 0.0456 0.163 0.0513 2.254*** -0.252** 0.0254
(0.116) (0.579) (0.121) (0.205) (0.236) (0.779) (0.102) (0.0816)
Herfindahl Index 0.0734 -0.101 0.00460 -0.162* 0.190 0.0704 -0.0418 -0.00690
(0.0600) (0.0741) (0.0477) (0.0857) (0.167) (0.107) (0.0504) (0.0539)
ISO Certified 0.00166 -0.0829 1.099 0.647 0.789 -1.939 1.713 -0.972
(0.241) (5.173) (0.826) (2.137) (0.793) (2.815) (1.071) (8.757)
Multinational -0.217 -0.557 0.429 -0.712 0.154 0.666 -0.528 -5.204
(0.309) (1.888) (0.300) (1.859) (0.256) (3.135) (0.401) (3.246)
Constant 0 0 2.036 15.17 -106.2*** 0 0 0
(0) (0) (14.52) (12.41) (17.02) (0) (0) (0)
Industry Effects Yes Yes Yes Yes Yes Yes Yes Yes
Time Effects Yes Yes Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes Yes Yes Yes
No. of Observations 298 89 405 139 202 61 645 192
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
3.6.3 Productivity and other variables
Let us consider the other control variables. As far as capital intensity of the firm is
concerned, again the two types of firms display disparate results. Higher capital intensity
74
has a significantly positive impact on the productivity of clothing firms but not on the
productivity of textile producers. For most of the different specifications shown in the
table, the coefficient for size is negative for textile firms. However, the only case where it
is significant is when it takes a positive value. On the other hand, it is always positive and
significant for clothing firms. Another intriguing point is that the sign of the herfindahl
index coefficient is positive and significant only for the textile firms; on the other hand, it
is negative and highly significant for the clothing firms. This indicates that higher
concentration in the industry results in lower productivity for clothing firms but not for
the textile firms. One would generally expect that greater degree of concentration in an
industry leads to greater market power for the firms in that industry and, hence, lowers
their productivity growth. This is not the case for textile producers. One possible
explanation for this result is that, although there might be a small number of firms with a
lot of market power, there is intense competition amongst them which forces them to
become more productive in order to capture an even bigger market share. That is why
higher concentration in the textile industry implies that textile producers are, on average,
more productive than if there were a large number of firms capturing an almost identical
market share. While this explanation is plausible, another explanation could be related to
returns to scale. The textile industry is dominated by a few capital intensive firms with
higher returns to scale. With the expansion of quotas, these firms might be capable of
ramping up their output, and productivity, rapidly because of their already large capital
investment. Within the textile industry, sub-industries with more of these large firms
(concentrated sub-industries) will be better able to ramp up output and productivity. On
75
the contrary, the lower returns to scale and lower capital intensity of the clothing industry
may restrict the output and productivity expansion.
45
Textile multinational firms, on average, tend to have a higher productivity
compared to non-multinational textile firms. This is not the case for clothing producers:
multinational clothing firms have a significantly lower mean productivity compared to
non-multinational clothing firms. Older textile firms, which are also likely to be bigger in
size, appear to be much more productive than their younger counterparts.
For most of the above-mentioned control variables, we have seen that the results
are different across the two types of firms. The only case where it is indistinguishable is
in the case of ISO certified T&C firms. ISO certification affects firm efficiency
positively: a firm certified for its quality management system has a higher productivity,
on average, than a firm that is not certified. These estimation results are arrived at after
controlling for industry, time and city fixed effects. The city fixed effects take into
account the fact that some firms are located in more developed areas relative to others.
There may be differences in the infrastructural facilities in different parts of the country;
regional fixed effects take care of these differences. Tables A.2.2 and A.2.4 show the
estimation results for each of the four phases individually. For a majority of the control
variables described above, we do not observe a noticeable change in either the sign or the
magnitude of the coefficients.
45
We thank an anonymous referee for this explanation.
76
3.6.4 Discussion
The above analysis highlights cross-sector variation in the effect of MFA expiration. As
is frequently emphasized in the new trade theory literature, trade reforms often influence
different sectors heterogeneously even within the manufacturing industry. However, what
appears to be exciting is that in our case the outcome differs within what is typically
lumped together as the textile industry. A liberalization episode such as the phasing of
quotas may generate divergent changes in the productivity levels of different categories
of products even within an industry. Quotas weaken the motivation to advance
technologically to capture market share because market shares are predetermined, and in
so doing, hinder productivity growth. MFA expiration will potentially boost competition,
both between and within countries, weakening tendencies toward oligopolies, thereby
resulting in technological advancement and productivity growth. We see this happening
in the textile sector. Pakistan has had a relatively better textile sector historically. The
textile industry is labor intensive and the primary input is cotton. The country has a high
production of cotton and a sizeable labor force that confirms its strong revealed
comparative advantage in the production of textile goods. On the other hand, clothing
industry still faces the challenge of obsolete machinery. Energy outages, workforce
development, product standards, fabric finishing, styles and patterns, customs and port
procedures, and security are other factors that shape productivity growth. One reason why
the TFP of garment firms declines after the end of MFA is competition from foreign
sellers of garments in the Pakistani market; since TFP confounds the effects of efficiency
and market share, if market share declines, this could translate into a decline in measured
77
TFP. Any form of liberalization like this has two opposing effects: the market stealing of
imports lowers sales for the domestic firms and leaves less money available to invest in
productivity improvements, and the higher competition spurs lagging firms to work
harder and improve productivity in order to survive. The balance of these two effects may
perhaps work out differently in the two sectors, for example, because the initial level of
competition may differ. Some theory papers incorporate the asymmetric effects of
liberalization by the firm’s productivity level. If non-exporting firms become exporters,
we may well see a decline in the mean industry productivity because new exporters need
time to adapt to the new environment.
The difference in results across textile and garment firms is related to the structure
of production, namely, the type of raw materials used by garment firms after the end
of MFA. However, the data does not provide information about the types of raw materials
used and it is, therefore, hard to determine if this was the case. Another possible
explanation is a change in product mix, for instance, a shift to the most productive
production lines in textiles, and expansion into new products for which there is still some
learning to do in the garment industry. As the MFA expires, Pakistan is changing the
composition of its textile exports, from a broader category that benefits from the MFA
without much weight of Rules of Origin (RoO), to a narrower category focused on
specific markets that offer Pakistan preferential access through bilateral trade agreements
with strict Rules of Origin. If this is the case, one would expect a fall in productivity as
the mix of inputs utilized by firms would no longer be dictated by rationally choosing the
optimal input-mix given market prices. If the composition of exports has changed in the
78
stated way, one should attempt to decompose the TFP between RoO-affected and non-
RoO-affected exports. One hypothesis is that the country found it harder to compete with
other countries in the garments sector because the production of clothing is relatively
more labor intensive than textiles; the firms in Pakistan could have responded to, say,
competition with China by upgrading the quality of Pakistani textiles but may not have
done so in the garment sector because it is harder to improve quality in that sector. These
cross-sector differences in quality ladders could play a crucial role under these
circumstances.
3.6.5 Robustness check: Alternative measure of productivity
This section provides an alternative measure of productivity to determine whether or not
results derived so far are sensitive to the empirical methodology used to estimate firm
efficiency. The OP methodology can be used to account for the simultaneity between
input choices and productivity shocks, in addition to the sample selection bias. Table 3.12
illustrates the estimation results when the change in firm productivity is regressed on the
change in adjusted level of quotas using OP productivity measure. We note that the
results are not very different from LP regression estimates. As before, an increase in the
adjusted level of quotas brings about a significant reduction in firm’s mean productivity
in the clothing industry but not in the textile sector. Moreover, the sign and magnitude of
most of the control variables’ coefficients remain the same as under LP. However, there
is a change in the sign of some of the coefficients in one or more phases. For example,
the coefficient for the multinational-company dummy variable is not positive across all
79
the four phases for textile firms and negative for clothing firms, as was the case under LP
methodology. Nonetheless, a large number of these estimates are not significant. Trade
costs do not seem to have a noteworthy impact on the productivity of clothing firms, and
trade cost coefficients for both the textile and clothing firms are not statistically
significant.
46
Even so, OP and LP productivity estimates yield almost the same results as
far as the sign and magnitude of coefficients are concerned. OP estimates are,
nevertheless, less significant compared to LP estimates derived earlier.
Table 3.12: Effect of Elimination of Quota Restrictions on Textile and Clothing Firm Productivity – Olley & Pakes
Variable (1) (2) (3) (4) (5) (6)
Textile Clothing
Adjusted Quota 1.087** 1.969*** 2.047*** -1.170*** -1.647*** -0.646
(0.539) (0.752) (0.727) (0.306) (0.353) (0.575)
Cost of Imports (-1) -0.146 0.124 0.103 0.110 -4.953 -5.612
(0.238) (0.223) (0.225) (6.915) (10.55) (11.24)
Herfindahl Index (-1) 0.110* 0.149* 0.162** -0.189 -0.186 -0.113
(0.0604) (0.0760) (0.0800) (0.153) (0.220) (0.224)
Multinational 0.0807 0.0483 0.234 -2.523 -4.167 -3.935
(0.173) (0.139) (0.200) (2.181) (3.770) (3.583)
ISO Certified 0.362** 0.767* 0.583 0.773 1.292 2.066*
(0.152) (0.460) (0.459) (0.618) (0.936) (1.087)
K/L (-1) -0.198 -0.169* -0.187* 0.781 0.214 -0.00531
(0.122) (0.0912) (0.107) (0.486) (0.755) (0.777)
Size (-1) -0.0156 -0.082*** -0.076*** 0.115* 0.108 0.0278
(0.0278) (0.0259) (0.0280) (0.0590) (0.0818) (0.0837)
Age -0.0244 -0.00227 3.653** 3.530*
(0.267) (0.281) (1.841) (2.100)
Age
2
0.0282 0.0372 -0.822** -0.691
(0.0496) (0.0574) (0.365) (0.484)
Constant -17.73* 0 0 11.77* 20.57*** 0
(10.05) (0) (0) (6.369) (7.319) (0)
Industry & Time Effects Yes Yes Yes Yes Yes Yes
City Effects Yes 4o Yes Yes 4o Yes
No. of Observations 1567 996 996 502 315 315
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
46
These differences in signs between LP and OP methodologies across four phases might be owing to the
small sample size when the coefficients are estimated for individual phases. A large number of these results
are not statistically significant and, hence, do not pose a serious threat to our main conclusion, namely, the
differing impact of MFA expiration on the two types of firms.
80
3.7 Firm-level results: Effect on output
In order to measure the effect of quotas directly on firm’s output, we regress output on
the adjusted level of quotas and trade costs. The results are shown in Table 3.13. There
are a number of interesting points to be examined here. First of all, the results vary for
both types of industries. In the textile sector, an increase in the adjusted level of quotas
leads to a significant rise in the firm’s output. For the clothing sector, however, this result
is not statistically significant. Since quotas are measured by quantity and not value, under
a given quota, producers try to manufacture high value products. Consequently, MFA
expiration is expected to bring about a shift to the production of lower-value products.
Table 3.13: Effect of Elimination of Quota Restrictions on Output
Variable (1) (2) (3) (4) (5) (6)
Textile Clothing
Raw Materials 0.264*** 0.285*** 0.132*** 0.0816** 0.0344 0.0102
(0.0629) (0.0741) (0.0503) (0.0393) (0.0286) (0.0269)
Labor 0.0907*** 0.0711*** 0.0597*** 0.114** 0.0315 0.0167
(0.0232) (0.0246) (0.0200) (0.0458) (0.0353) (0.0456)
Fixed Assets 0.0550* 0.0448 0.0764* 0.0936** 0.122** 0.0712
(0.0329) (0.0390) (0.0453) (0.0410) (0.0533) (0.0539)
Adjusted Quota 0.137 1.523** 2.409** 0.494 0.420 0.975*
(0.246) (0.702) (1.049) (0.334) (0.353) (0.509)
Cost of Imports (-1) -0.287 -0.422* 7.774* 11.71*
(0.210) (0.232) (4.265) (6.185)
Multinational 0.379* 0.386 -1.981 -3.538
(0.200) (0.285) (2.074) (3.479)
ISO Certified 0.770*** 0.979 1.709*** 2.512***
(0.191) (0.676) (0.421) (0.896)
Age -0.0135 2.699
(0.249) (1.925)
Age
2
0.0652 -0.493
(0.0574) (0.428)
Constant 8.567* 0 -32.82 6.674 3.331 0
(4.752) (0) (20.00) (6.044) (6.514) (0)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes Yes
No. of Observations 1811 1461 929 648 503 316
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
81
There is a significant reduction in output if trade costs go up in the textile sector.
This, in contrast, is not true for the clothing firms: an increase in the trade costs, on
average, results in an increase in output in the clothing industry and the estimates are
significantly different from zero in nearly all the cases. Another remarkable point is that a
textile multinational firm has, on average, a significantly higher output compared to a
textile firm that is not a multinational company, whereas, the corresponding coefficient
for clothing firms is negative. On average, older textile firms produce lesser output, but
this is not true for clothing firms. Both, the ISO Certified textile as well as clothing, firms
produce a higher output compared to a textile or a clothing firm that is not ISO Certified,
and this finding is statistically significant. To sum up, MFA expiration lead to an increase
in the output of T&C firms in Pakistan. However, for a majority of the specifications that
we consider, this result is statistically significant only for textile firms.
3.8 Conclusion
The most important contribution of this chapter is that it is one of the few studies to
empirically investigate the effect of liberalization in the form of the phasing out of quotas
on firm-level productivity in the textile and clothing industry. The existing studies pertain
to macroeconomic outcomes of the end of MFA, and do not consider the effect on textile
firms. The studies that do attempt to evaluate the impact of lifting a quota at the firm
level do not utilize the actual number of quotas imposed by developed countries on
imports from developing countries. This chapter, on the other hand, uses the database that
traces U.S. trading partners’ exports to the U.S. along with the actual amount of quota
82
under the regimes determined by the MFA. Hence, unlike most other studies in the
literature which mainly investigate the impact of trade liberalization in a developing
country, this chapter investigates liberalization episode in a developed country and its
impact on firms in a developing country. Because of the nature of the data and the
empirical methodology used, it effectively takes care of the endogeneity problem that is
often challenging for the analyses trying to estimate the effect of liberalization on firm
performance.
Table 3.14: Foreign Direct Investment in Current U.S. Dollars (1971-2010)
Foreign Direct Investment
(Current U.S. Dollars)
Year
Foreign Direct Investment
(Current U.S. Dollars)
2010 2022000000 1990 245262963.5
2009 2338000000 1989 210599917.1
2008 5438000000 1988 186491557.3
2007 5590000000 1987 129377643.6
2006 4273000000 1986 105730331.8
2005 2201000000 1985 131389252.2
2004 1118000000 1984 55510169.71
2003 534000000 1983 29457026.66
2002 823000000 1982 63833091.62
2001 383000000 1981 108084748.5
2000 308000000 1980 63632992.78
1999 532000000 1979 58254127.36
1998 506000000 1978 32273192.51
1997 716253125.4 1977 15223204.01
1996 921976182.5 1976 8220530.168
1995 722631560.7 1975 25000000
1994 421024638.5 1974 4000000
1993 348556957.8 1973 -4000000
1992 336479857.1 1972 17000000
1991 258414487 1971 1000000
Source: United Nations Conference on Trade and Development, Foreign Direct Investment Online Database (The
United Nations Statistics Division Statistical Databases)
We observe that MFA expiration lead to an increase in the average productivity of
textile producing firms but a significant reduction in the mean productivity of clothing
and garment producers. The chapter draws attention to cross-sector variation in the
83
impact of MFA expiration and that trade reforms often influence different sectors
heterogeneously even within the manufacturing industry. It proposes various explanations
for this outcome, for example, a change in the product mix, entry by non-exporters in the
clothing sector, cross-sector differences in quality ladders, and so forth.
The competitiveness of T&C industry hinges on numerous factors: labor cost,
production costs (energy, water, production inputs, for example, cotton, polyester, and
chemicals), transport and distribution, and macroeconomic environment (domestic
interest rates, corporate taxes, exchange rate, property rights, and political stability).
Private sector in Pakistan appears to benefit from the domestic raw material base in
cotton and synthetic fibres, low labor costs, and large-scale investment in the last number
of years. The capacity of the T&C industry is expected to augment in the near future.
47
Clearly, the T&C industry has benefited from complimentary trade agreements with the
U.S. and E.U. since 2001 with regard to fight against terrorism. Table 3.14 shows the
foreign direct investment in current U.S. dollars from 1971 to 2010. The government is
promoting diversification in terms of input use and products to lessen the concentration in
low value-added products, and in order to diminish the heavy reliance on cotton. It has
been promoting progress in the weaving sector through implementation of standards and
loan programs to upgrade to auto looms. The declining efficiency of clothing firms points
to the failure of these firms to fight competition. MFA expiration is a chance for them to
trim down their input usage which can help reduce export prices in the world market,
yielding the desired competitive edge over other exporters.
47
For instance, the Kohinoor factory planned to triple its workforce and its capacity of 200 sock-making
machines (Haté et al., 2005).
84
Chapter 4
The Case of Vertically Integrated Clothing Firms
The developing countries must be able to take a more active part in trade negotiations, through
technical assistance and support from the developed countries.
4onetheless, the developing countries must be able to reap the benefits of international trade.
– Anna Lindh
4.1 Vertical integration of production
Vertical integration is the organization of production under which a single business unit
carries on successive stages in the processing or distribution of a product which is sold by
other firms without further processing (Blois, 1972). In line with the Transaction Cost
Economics (TCE) theory founded by Williamson (1975), the internal organization of a
firm needs to be designed in such a way so as to improve incentives and control agency
costs. The literature has identified a range of factors associated with integration. Vertical
integration encourages firm-specific investments and reduces holdup problem whenever
markets are imperfect, and are widespread when it is harder to write long-term contracts
85
between the upstream and downstream firms (Acemoglu, Johnson, & Mitton, 2005).
Firms integrate vertically to erect entry barriers, maintain product quality, assist
investments in specialized assets, and develop coordination (Williamson, 1975).
The advantages must be weighed against the disadvantages which usually consist
of disparities between productive capacities at different stages of production,
governmental pressure, lack of specialization, and lack of direct competitive pressure on
costs of intermediate products (Blois, 1972). Nevertheless, there is only modest micro-
level evidence on productivity-integration relationship and on the way production differs
in vertically integrated firms compared to non-integrated firms. There is even less
evidence on the impact of trade liberalization on productivity of vertically integrated
firms. By contrast, there is a growing literature on international specialization of
production and its impact on firm efficiency. Expansion of international specialization
and the disintegration of production has been an important characteristic of the
international economy (Antràs & Helpman, 2004).
48
However, the focus of this literature
has been on productivity gains associated with the disintegration of production across
different countries.
Figure 4.1 shows the mean productivity of the two types of firms computed using
three different productivity measures. Notice that over the entire sample period, vertically
integrated clothing producers are much more productive than non-integrated clothing
producers if productivity is computed using Levinsohn and Petrin (LP), and Olley and
Pakes (OP) productivity measures. Whereas the average productivity of vertically-
48
Hummels, Ishii, and Yi (2001) demonstrate that international trade has grown faster in components than
in final goods.
86
integrated clothing firms exhibits an upward trend, we do not see an analogous pattern for
non-integrated clothing firms. Instead, the average productivity of non-integrated
clothing firms remains roughly at the same level as at the start of the period. Figure 4.1
also illustrates mean productivity computed from parametric estimation of production
functions of the two types of firms. Although estimation of production function
coefficients may yield biased estimates of productivity (for reasons discussed earlier in
chapter 3), we do observe an upward trend in average productivity of vertically integrated
clothing firms. This is again not true for non-integrated firms.
Let us look at a few other variables. Vertically integrated firms produce, on
average, a slightly higher output than non-integrated producers, as shown in Figure 4.2.
There is an upward trend in the output of vertically integrated firms but we do not notice
a comparable trend for non-integrated firms. As far as labor, raw materials and other
factors are concerned, we do not observe much difference between the two types of
firms.
49
Even though a variety of studies look into the efficiency of vertically integrated
firms relative to non-integrated firms, none of these in particular consider the influence of
trade liberalization caused by the phasing out of quotas on firm-level efficiency of these
two types of firms. We explain how trade liberalization generates a greater incentive for
vertical integration in the production of clothing goods.
49
Figure 4.2, for instance, demonstrates average number of employees and raw materials for the two kinds
of firms.
87
Figure 4.1: Vertically Integrated and Non-Integrated Clothing Firms – Productivity
0 5 10 15 20
Mean productivity
1990 1995 2000 2005 2010
Year
Levinsohn & Petrin
-2 0 2 4 6 8
Mean productivity
1990 1995 2000 2005 2010
Year
Olley & Pakes
-.5 0 .5 1 1.5
Mean productivity
1990 1995 2000 2005 2010
Year
Non-integrated clothing firms
VI clothing firms
OLS
88
Figure 4.2: Vertically Integrated and Non-Integrated Clothing Firms – Output, Labor & Raw Materials
12 14 16 18 20
Ln(output)
1990 1995 2000 2005 2010
Year
Output
15 15.5 16 16.5 17
Ln(labor)
1990 1995 2000 2005 2010
Year
Labor
12 14 16 18 20
Ln(raw materials)
1990 1995 2000 2005 2010
Year
VI clothing firms
Non-integrated clothing firms
Raw Materials
89
4.2 Review of Literature
Although the concern of this chapter is not to determine the causes of vertical integration,
in order to motivate the diverse outcomes in terms of distinct firm performance as a result
of liberalization, it is deemed relevant to review the literature on integration choices of
firms. Acemoglu et al. (2005) obtain cross-country correlations between vertical
integration and financial development, contracting costs, and entry barriers. These
correlations are more or less completely driven by industrial composition. They observe
that countries with less-developed financial markets are significantly more integrated in
industries that are more human capital or technology intensive. Credit market
imperfections often shape the organization of firms. When credit markets are imperfect
and when there is lack of financial development, there are more likely to be larger firms
in a country (Kumar, Rajan, & Zingales, 1999). These firms are prone to produce some of
their own inputs. Therefore, improved financial institutions and credit markets may
perhaps be associated with not as much of vertical integration (Acemoglu et al., 2005).
Grossman, Helpman, and Szeidl (2006) study equilibrium integration choices of firms
that vary in productivity levels, concentrating on the role that industry characteristics,
such as cost of transporting intermediate and final goods, play in determining optimal
integration strategies. Hennessy (1996) suggests that information externalities, arising
from problems in detecting food quality, may possibly be the reason why vertical
coordination is being used to avoid the marketplace. Hart and Moore (1990) identify the
benefits and costs of vertical integration for incentives to carry out relationship-specific
investments.
90
A variety of studies look into the efficiency of vertically integrated firms relative
to non-integrated firms. Babe (1981) makes productivity comparisons of integrated and
non-integrated Canadian telephone companies in order to determine the net effect of
vertical integration on efficiency of telephone operations. Vertical integration results in
lower costs to telephone operations in the form of reduced contracting and selling costs,
system-wide planning of technology, and so forth. Conversely, vertical integration brings
about inefficiency on the part of telephone companies owing to the possibility that
desirable technology available only from unaffiliated suppliers may be foreclosed. He
finds that non-integrated telephone companies achieve productivity gains considerably in
excess of those attained by integrated companies. Despite the possibility of economies of
scale, differences in growth rates of profitable as compared to non-compensatory services
among companies, and initial levels of efficiency, the evidence suggests that vertical
integration seems to be the most important source of this result. Hortaçsu and Syverson
(2007) use a rich plant-level data set of cement and ready-mixed concrete producers to
reflect on the reasons for and results of vertical integration, with particular regard to its
effects on market power. There is modest evidence that vertical foreclosure effects are
quantitatively significant in these industries. This pattern is compatible with an
efficiency-based mechanism: higher productivity producers are more prone to vertically
integrate, and are larger, more probable to grow and survive, and charge lower prices
(Hortaçsu & Syverson, 2007). They also show that integrated producers’ productivity
advantage is owing to the improved logistics coordination afforded by large local
concrete operations, but this advantage is not tied to firms’ vertical structure per se since
91
non-vertical firms with large local concrete operations have equally high productivity
levels. The vertically integrated plants have TFP levels that are 4.3 percent greater on
average than their non-integrated competitors. Contrasting TFP changes among formerly
non-integrated plants shows that the plants that become integrated witness approximately
10 percent faster productivity growth. Hortaçsu and Syverson (2009) make use of the
Longitudinal Business Database to study the productivity of plants in vertically structured
firms.
50
They discover that vertically integrated plants have higher productivity levels
than their non-integrated industry cohorts, but the majority of this distinction reflects the
selection of high-productivity plants into vertical structures rather than a causal effect of
integration on productivity. These productivity differences typically are not related to the
movement of goods along the production chain; vertically integrated firms’ upstream
plants ship an astoundingly little amount to downstream plants in their firm. Forbes and
Lederman (2009) consider how vertical integration affects airline performance. Among
flights departing from a given airport on a given day, airlines that own their regional
affiliates encounter fewer cancellations and shorter delays than those contracting with
affiliated regionals at arms’ length: vertical integration sets out the decision rights within
the organization and allows airlines to more adeptly respond to unanticipated scheduling
issues, however, at the expense of higher wage costs (Forbes & Lederman, 2009).
Tomiura (2007) documents how productivity varies with globalization modes.
Foreign outsourcers and exporters are likely to be less efficient than the firms active in
FDI or in multiple globalization modes but more efficient than domestic firms, even after
50
The Longitudinal Business Database contains most private non-agricultural establishments in the U.S.
92
firm size, factor intensity, and industry are controlled for. Using a panel of longitudinal
data documenting more than 3,500 product introductions in the microcomputer industry,
Rothaermel, Hitt, and Jobe (2006) propose that balancing vertical integration and
strategic outsourcing in pursuit of taper integration augments a firm’s product portfolio
and product success, and in turn adds to competitive advantage and superior
performance.
51
Blois (1972) considers a situation called ‘vertical quasi-integration’ where
firms are gaining the advantages of vertical integration without assuming the rigidity of
ownership. Cohen-Meidan (2009) examines the interaction of trade policy with the
vertical structures of foreign firms exporting goods to the U.S., focusing on antidumping
duties, and shows that the policy has a remarkably mixed impact on vertically integrated
and non-integrated foreign firms. Specifically, non-integrated firms are more prone to
exit the U.S. market than vertically integrated firms following imposition of duties, and
less prone to pass the duties on to the consumers.
There are a large number of studies which attempt to analyze the impact of MFA
expiration on reallocation of production and exports across countries as well as on the
productivity of firms. To my knowledge, there are hardly any analyses that empirically
evaluate the impact of lifting a quota on productivity of different types of firms within a
manufacturing industry. Therefore, the most important contribution of this paper is that it
highlights the variation in the effect of MFA expiration across firms depending on their
organization. Although a variety of studies look into the efficiency of vertically
integrated firms relative to non-integrated firms, none of these considers the effect of
51
Taper integration occurs ‘when firms are backward or forward integrated but rely on outsiders for a
portion of their supplies or distribution’ (Harrigan, 1984).
93
trade liberalization caused by the phasing out of quotas on firm-level efficiency of these
two types of firms. As is frequently emphasized in the new trade theory literature, trade
reforms often influence firms heterogeneously. By merging micro-level data of firms
with the data on quotas at the industry level, this chapter shows that a liberalization
episode may generate divergent changes in the productivity of different categories of
firms even within the clothing industry.
4.3 Theoretical Framework
Subsequent to the above analysis, it would be useful to compare the productivity of
clothing producers that buy their raw materials (either from domestic producers or from
abroad), and that of vertically integrated clothing producers that manufacture their own
yarn and fabric. In this section, we describe how trade liberalization generates an
incentive for vertical integration in the T&C industry. We follow the model by Yi (2003)
which marries a Dornbusch-Fischer-Samuelson Ricardian international trade framework
to a dynamic macroeconomic framework.
52
Trade liberalization triggers a change in the
relative factor cost of the two types of firms, and consequently, a change in the product
range produced by each of them. We consider two special cases of free-trade equilibria
which generate complete specialization in the production of all goods. Consider three
types of firms having the following production functions:
52
Yi (2003) chose a Ricardian framework because Golub and Hsieh (2000), and Eaton and Kortum (2002)
have shown the empirical significance of Ricardian technological differences in rationalizing trade patterns,
and there is insufficient empirical evidence that favors the monopolistic competition model in contrast to
other models. It is desirable to have a trade model in which firms decide whether to use domestic or
imported inputs (Yi, 2003).
94
+
s
=
+
s
+
s
t
+
s
+`t
, (4.1)
-
s
= N
+
s
u
M
-
s
-
s
t
-
s
+`t
O
+`u
, (4.2)
Q
s
=
Q
s
Q
s
v
Q
s
+`v
, (4.3)
where i = H or F (H denoting home production and F denoting foreign production), and
s w [0, 1] indicates product z produced by the firm.
G
s
is the TFP of firm f in the
production of good z in country i, and
G
s
and
G
s
are labor and capital, respectively,
used by firm f in producing output
G
s
. The first type of firm produces raw materials
purchased by the second type, i.e.
+
s
= N
+
s
, where N
+
s
is firm 2’s use of the
output produced by firm 1. The second type of firm combines input produced by the first
type, labor, and capital in a nested Cobb-Douglas production function. In our case, it
suggests that the first type produces textile products, and the second type buys textile
products from type one firm and uses them to produce the final good, for example, ready-
made garments. The third type of firm also produces the final good but, unlike the second
type of firm, it produces its own raw materials.
4.3.1 Firms and Technology
Let us only consider the home industry. Firms maximize profits taking prices as given.
The profit maximization problem for type 1 firm is given by:
Max {
| !
+
s
+
s
− *
+
s
− (
+
s
}
| w. r. t.
+
and
+
,
and that for the type 2 firm is:
95
Max {
| !
-
s
-
s
− !
+
s
N
+
s
− *
-
s
− (
-
s
}
| w. r. t. N
+
,
-
and
-
,
if the firm buys only domestically produced raw materials from type one firms. !
+
s
is
the world price of textile raw materials, !
-
s
is price of the final good, and * and ( are
the wage and rental rates, respectively. If firm 2 buys imported raw materials, it’s profit
maximization problem is given by:
Max {
| !
-
s
-
s
− !
+
s
1 +
N
+
s
− *
-
s
− (
-
s
} w. r. t. N
+
,
-
and
-
| ,
where τ is a measure of trade liberalization, for example, the tariff rate or the price of a
license to a quota. Lastly, the profit maximization problem for type 3 firm is given by:
Max {
| !
-
s
Q
s
− *
Q
s
− (
Q
s
}
| w. r. t.
Q
and
Q
.
4.3.2 Households
The utility maximization problem of households is specified as:
Max )
ln6f
7
€
,
w.r.t. �
f
+ �
M
‚+
− 1 − h
O = *
+ (
+ ƒ
≡ �
,
‚+
= 1 − h
+ c
, ∀ ≥ 1,
where �
and
are the price and output of the final good, respectively, in country i at
time t. f
is consumption,
and
are total capital and labor, respectively, c
is
investment, and ƒ
is the lump-sum transfer of quota license revenue expressed in terms
96
of the home final good. Households own the capital and rent it to firms period by period.
4.3.3 Market Clearing
The market clearing condition for good 1 (textile – raw material) is:
+
s
=
+
‡
s
+
+
ˆ
s
= N
+
‡
s
+ N
+
ˆ
s
,
and the market clearing condition for good 2 (apparel – finished good) is:
-
s
+
Q
s
=
-
‡
s
+
-
ˆ
s
+
Q
‡
s
+
Q
ˆ
s
,
-
s
+
Q
s
= f
+ M
‚+
− 1 − h
O =
, ∀ ≥ 1 and $ = ‰, Š.
The market clearing conditions for labor and capital markets, respectively, are given by:
= ‹
+
s
Is
+
+ ‹
-
s
Is
+
+ ‹
Q
s
Is,
+
= ‹
+
s
Is
+
+ ‹
-
s
Is
+
+ ‹
Q
s
Is.
+
Let the textile raw materials (good 1) be the numeraire, i.e. !
+
s
= 1, and !
-
s
=
�
s
.
4.3.4 Definition of Equilibrium
An equilibrium is a sequence of goods and factor prices, {!
+
s
, !
-
s
, �
, *
and (
},
and quantities {
+
s
,
-
s
,
Q
s
,
+
s
,
-
s
,
Q
s
,
+
s
,
-
s
,
Q
s
,
97
N
+
s
, and
} ∀ ≥ 1, s w [0, 1] for i = H, F such that the first order conditions of firm’s
and household’s maximization problems given above, as well as the market clearing
conditions, are satisfied.
For the sake of simplicity, let us remove the time subscript. The profit
maximization problem of first type of firm can be written as:
Max {
+
, Œ
+
t
+
+`t
− *
+
− (
+
} w. r. t.
+
and
+
.
The productivity,
+
s
, is a function of τ plus all the other factors that influence firm’s
productivity, denoted by Œ. It is affected by a number of factors, such as, worker skills,
energy outages, off-balance sheet transaction costs, corruption, security and
infrastructure. The profit maximization problem for type 2 firm is:
Max � !6N
+
u
[
-
, Œ
-
t
-
+`t
]
+`u
7 − N
+
− *
-
− (
-
Ž w. r. t. N
+
,
-
and
-
,
if the firm buys only domestically produced raw materials. Alternatively, if the firm buys
imported raw materials, it’s profit maximization is given by:
Max � !6N
+
u
[
-
, Œ
-
t
-
+`t
]
+`u
7 − 1 +
N
+
− *
-
− (
-
Ž w. r. t. N
+
,
-
and
-
.
Similarly, we can obtain the profit maximization problem for type 3 firm. We are also
able to derive an expression for the relative productivity,
F
�
�,‘
F
’
�,‘
, by using the first order
conditions for type 2 and type 3 firms (see Appendix C):
s
≡
F
�
“
F
’
“
≡
F
�
�,‘
F
’
�,‘
≅
+ ‚ �
+`t
•“
+`v
(4.4)
98
Hence, the relative productivity is a function of τ as well as all other factors that affect
firm’s productivity. Let us consider two different cases as examples of free-trade
equilibria which generate complete specialization in the production of all goods. Trade
liberalization in the form of a reduction in the price of a license to a quota causes a
reduction in the relative factor cost of firm 2 compared to firm 3. In other words, the
relative factor cost of firm 3 will go up. This is shown in Figure 4.3. The vertical axis
measures the relative productivity and the relative factor cost of firm 3 relative to firm 2.
On the horizontal axis, with no loss of generality, the [0, 1] continuum can be arranged so
that it is diminishing in the productivity of firm 3 relative to firm 2 in the home country; z
= 0 is the good in which firm 3’s productivity (relative to that of firm 2) is the highest.
The “cutoff” s
Q
defines the pattern of production. The arbitrage condition that determines
the cutoff separating production between firms 2 and 3 can be found by equating relative
factor cost to relative productivity. The condition essentially says that the vertically
integrated firm (i.e. firm 3) produces and exports up to the point where its cost advantage
(disadvantage) relative to the non-integrated firm (i.e. firm 2) equals its productivity
disadvantage (advantage). An upward shift in the relative factor cost line will lead to a
reduction in the product range produced by firm 3 and an increase in the product range of
firm 2 if there is no change in relative productivity of the two firms. However, there are
other factors which may also affect relative productivity. As a result, the relative
productivity function may shift up or down. Figure 4.3 shows what happens if the relative
productivity function, A(z), shifts up. In this case, there is an increase in the product range
produced by firm 3 since the cut-off goes up from s
Q
to s
Q
––
.
99
Figure 4.3: Relative factor costs and relative productivities (Firm 3/Firm 2)
Figure 4.4: Relative factor costs and relative productivities (Firm 3/Firm 2)
100
Figure 4.5: Relative Factor Costs and Relative Productivities (Home/Foreign)
Therefore, what happens to the product range would depend not just on the tariff
change but also on all other factors affecting the relative productivity of firms. In Figure
4.4, relative productivity function shifts downward. A reduction in relative productivity
of firm 3 for all values of z, as well as a rise in the relative factor cost (compared to firm
2) will result in an enormous reduction in product range of finished goods produced by
firm 3. In Figure 4.5, the y-axis denotes relative factor costs (home/foreign) and relative
productivities for firm 2 (home/foreign) and for firm 3 (home/foreign). On the horizontal
axis, the continuum is arranged so that it is diminishing in the home country’s
comparative advantage in goods produced by firm 3. Let us also assume that the
comparative advantage ordering of firm 2’s productivity at home is the same as it is for
firm 3. The cut-offs s
-
and s
Q
now define the pattern of production. The middle region of
the continuum engenders the need for vertical integration. In this region, firm 3 in the
home country produces the finished good and exports it to the foreign country. Trade
101
affects the pattern of specialization because it changes the cost of imported inputs. The
range of vertical integration, or goods produced by firm 3, goes up as a result of a
reduction in τ. This is accompanied by an increase in the product range produced by firm
2. Thus, trade liberalization in the home country results in an increase in product range
produced by vertically integrated firms as well as the country-wide product range.
4.4 Empirical Methodology
In this section, we review the empirical methodology used to measure the effect of the
end of MFA on performance of vertically integrated and non-integrated apparel firms
from 1992 to 2010. However, before we begin to explain the methodology, we motivate
the analysis by examining how distinct these firms are in terms of key firm
characteristics, such as output, number of physical inputs used, net profit, etc. This is
imperative because if the two types of firms are not significantly different from each
other with respect to these characteristics, then the divergence in our estimation results
could be the upshot of other factors not directly measurable in the data.
To investigate how dissimilar these firms are in terms of firm characteristics, we
run the following regression:
=
+
+
VI
+ h
+ h
+ 4
, (4.5)
where VI
is an indicator variable denoting a vertically integrated firm, and
are the
different firm characteristics of firm i in year t in industry j at six-digit level, such as,
measures of productivity used, fixed assets, size of the firm, capital intensity, net profit,
102
etc. The coefficient
+
reports the difference across integrated and non-integrated
clothing producers. h
and h
are time and industry fixed effects, respectively, and 4
is
the error term. Robust standard errors are corrected for clustering at the firm level. Since
we are interested in estimating the change in productivity after the end of MFA, our
primary objective is to test if the change in these dependent and independent variables is
significant, and whether or not it is related to the vertical structure of the firm per se.
Therefore, we replace
in Eq.(4.5) by ∆
:
∆
=
+
+
VI
+ h
+ h
+ 4
. (4.6)
Lastly, we run the same regressions but with controls for firm sales:
=
+
+
VI
+
-
log g'g
+ h
+ h
+ 4
, (4.7)
∆
=
+
+
VI
+
-
log g'g
+ h
+ h
+ 4
. (4.8)
To determine the effect of trade liberalization on firm performance, we need to
find a measure of productivity for the firms in our sample. We use OP and LP technique
to compute TFP. The change in firm productivity is then regressed on the adjusted level
of quotas:
∆ !
=
+
+
∆ logIde&/'
+
-
VI
× ∆ logIde&/'
+r
+ h
+ h
+ 4
, (4.9)
where log Ide&/'
is the logarithm of post-MFA adjusted level of quotas of product
j at time t. r
includes other control variables: size, age and capital intensity of the firm,
103
whether or not the firm is ISO Certified, a dummy variable for the city in which the firm
is located, whether or not the firm is multinational and, lastly, a herfindahl index of
concentration of the industry at six-digit level. We expect
+
to be positive if an increase
in the adjusted level of quotas leads to an increase in mean productivity of non-integrated
clothing firms. Also, a negative
-
would signify that the gain in productivity of
vertically integrated firms is less than that in productivity of non-integrated firms (if
+
is
positive and ™
+
™ > ™
-
™), or there is a significant reduction in the mean productivity of
vertically integrated firms (if
+
is positive and ™
+
™ < ™
-
™ ). Industry and year fixed
effects are included in all specifications. Eq.(4.9) is run separately for the different
measures of productivity used.
In addition, to determine whether or not our results differ for more capital
intensive vertically integrated firms or firms that are bigger in size, we run analogous
regressions including interaction terms, i.e. interaction of VI with size and capital
intensity of the firm:
∆ !
=
+
+
∆ logIde&/'
+
-
VI
+
Q
VI
× ∆ log
`+
+r
+ h
+ h
+ 4
, (4.10)
∆ !
=
+
+
∆ logIde&/'
+
-
VI
+
Q
VI
× ∆ log œ
`+
+r
+ h
+ h
+ 4
. (4.11)
To test whether or not vertically integrated clothing firms fared better than non-integrated
firms in terms of other key measures of performance, such as output and net profit, we
104
replace productivity by firms’ output and net profit in Eq.(4.9):
∆ log/&!&
=
+
+
∆ logIde&/'
+
-
VI
× ∆ logIde&/'
+r
+ h
+ h
+ 4
, (4.12)
∆ log% !(/ $
=
+
+
∆ logIde&/'
+
-
VI
× ∆ logIde&/'
+r
+ h
+ h
+ 4
. (4.13)
Control variables now include fixed assets, raw materials, number of employees, variable
trade costs, a dummy variable for the city in which the firm is located, age of the firm,
whether or not the firm is ISO Certified, whether or not the firm is multinational, and the
herfindahl index of industry concentration at six-digit level. The BSDPC includes data
for 90 clothing companies for the years 1992-2010. Once again, we employ the data used
by Brambilla et al. (2007) that traces U.S. trading partners’ performance under the quota
regimes. The data on trade costs is taken from Bernard et al. (2006) which provides data
on free-on-board customs value of imports, ad valorem duty, freight and insurance rates.
4.5 Difference between Integrated and :on-integrate d Firms
Before proceeding to results analyzing the impact of liberalization on firm productivity of
vertically integrated and non-integrated clothing firms, let us first examine the differences
between these two types of firms in terms of output, capital stock, labor, intermediate
inputs, capital intensity and net profit. This can be seen by looking at the firm-level
relationship between these variables and each firm’s integration status by regressing each
105
variable on an indicator for firm’s vertical integration status (denoted by VI
= 1 if a
clothing firm is vertically integrated; VI
= 0 otherwise), and a full set of industry-year
fixed effects. Consequently, the vertical integration dummy coefficient captures mean
difference across integrated and non-integrated producers in the same industry and time
period. This specification is helpful because it compares producers facing identical
industry-level demand and supply conditions (Hortacsu & Syverson, 2007). The results
are shown in Tables 4.1 and 4.2. They report differences in key dependent and
independent variables across integrated and non-integrated clothing producers. The
reported coefficients are for the indicator variable. On average, vertically integrated
producers have higher sales, capital stock, labor and capital intensity than non-integrated
firms but a lower level of raw materials and net profit. On the contrary, none of these
results are statistically significant. On the other hand, coefficients for productivity
measures (both OP as well as LP) are positive and significant in both the tables. Although
the coefficient for parametric estimate of productivity is positive but not significant, as
shown in Table 4.1, it is positive and significant when the change in productivity is
regressed on VI (look at column (3) in Tables 4.1 and 4.2). Table 4.2 also shows that the
change in sales, raw materials, fixed assets and capital intensity is negative for integrated
firms. However, yet again, the coefficients are not statistically significant.
To check the robustness of these results, we add firm’s sales as a control variable.
Table 4.3 demonstrates that the results do not change for all of our measures of
productivity as well as for capital intensity. The capital intensity coefficient is again
positive: vertically integrated firms, on average, have a higher capital intensity than non-
106
integrated clothing firms having similarly sized sales. In other words, the growth in
productivity for vertically integrated firms is greater on average than for non-integrated
firms having equally sized sales, and the coefficients are statistically significant.
4.6 Production function estimates - Levinsohn & Petrin
We estimate production function coefficients separately for vertically integrated and non-
integrated firms using Levinsohn & Petrin semi-parametric estimation. Table 4.4 reports
the production function estimation results for vertically integrated and non-integrated
clothing firms using LP productivity measure. We notice that the non-integrated firms
use more of all inputs per unit of value-added. These estimates are used to work out log
of measured TFP of plant i at time t. The change in firm productivity is then regressed on
the change in adjusted level of quotas for both types of garment producing firms (i.e. firm
2 and firm 3 in the model above), allowing for time and industry fixed effects. This
procedure is then repeated using OP as well as parametric estimate of productivity. The
results are illustrated in Tables 4.5 and 4.6.
107
Table 4.1: Differences between Integrated and Non-integrated Clothing Producers
VARIABLES LP OP OLS Fixed Assets Sales Raw Materials Net Profit Size Capital Intensity
(1) (2) (3) (4) (5) (6) (7) (8) (9)
VI 15.57*** 8.591*** 0.0688 2.154 0.00736 -0.505 -0.383 0.0520 0.120
(0.280) (0.553) (0.438) (1.979) (1.160) (1.041) (2.210) (0.253) (0.119)
Constant 0.402** -1.351*** 0.299 2.631* 19.08*** 18.40*** 21.95*** 16.25*** 0.140*
(0.191) (0.453) (0.418) (1.386) (0.870) (0.671) (1.558) (0.365) (0.0810)
No. of Observations 1255 1255 1255 1255 1255 1255 1255 1255 1254
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. Industry-year fixed effects are included in all specifications. Coefficients for the
industry and year dummies are suppressed. The reported coefficients are those for an indicator variable, VI, denoting that a firm is vertically integrated. *** Significant
at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
Table 4.2: Differences between Integrated and Non-Integrated Clothing Producers
VARIABLES Change in
LP
Change in
OP
Change in
OLS
Change in
Fixed Assets
Change in
Sales
Change in
Raw Materials
Change in
Net Profit
Change in
Size
Change in
Capital Intensity
(1) (2) (3) (4) (5) (6) (7) (8) (9)
VI 0.759** 0.627* 0.602** -0.894 -0.672 -1.435 2.395 0.152 -0.0739
(0.374) (0.358) (0.235) (3.941) (1.157) (1.232) (4.580) (0.489) (0.241)
Constant -0.403 -0.499 -0.259 -0.463 -0.909 -0.709 -0.187 -0.0584 -0.0301
(0.518) (0.565) (0.506) (1.698) (0.975) (0.817) (1.958) (0.477) (0.101)
No. of Observations 1237 1237 1237 1237 1237 1237 1237 1237 1235
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. Industry-year fixed effects are included in all specifications. Coefficients for the
industry and year dummies are suppressed. The reported coefficients are those for an indicator variable, VI, denoting that a firm is vertically integrated. *** Significant
at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
108
Table 4.3: Differences between Integrated and Non-integrated Clothing producers - Controlling for Firm’s Sales
VARIABLES LP Change in LP OP Change in OP OLS Change in OLS Capital
Intensity
Change in
Capital Intensity
(1) (2) (3) (4) (5) (6) (7) (8)
VI 15.57*** 0.784** 8.563*** 0.582** 0.0501 0.572 0.120 -0.0741
(0.276) (0.370) (0.345) (0.271) (0.406) (0.419) (0.104) (0.220)
Sales 0.115* 0.154** 0.484*** 0.452*** 0.399*** 0.377*** 0.0170* 0.0210
(0.0597) (0.0788) (0.0544) (0.0500) (0.0642) (0.0601) (0.0101) (0.0129)
Constant -1.804 -3.304** -10.45*** -8.872*** -7.187*** -7.263*** -0.180 -0.424
(1.177) (1.571) (1.108) (1.068) (1.306) (1.259) (0.206) (0.262)
No. of Observations 1255 1237 1255 1237 1255 1237 1254 1235
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. Industry-year fixed effects are included in all specifications.
Table 4.4: Production Function Estimates for Vertically Integrated and Non-Integrated Clothing Firms
Vertically Integrated
Clothing
Non-Integrated
Clothing
(1) (2)
Employment 0.169*** 0.355***
(0.0596) (0.0362)
Fixed Assets 0.00113 0.00719
(0.00716) (0.0142)
Raw Materials 0.0553 0.999***
(0.310) (0.313)
No. of observations 490 953
109
Table 4.5: Effect of Elimination of Quota-Restrictions on productivity of Clothing Firms – Levinsohn & Petrin
Variable (1) (2) (3) (4) (5) (6) (7)
Adjusted Quota -0.159** -0.155** -0.155** -0.155* -0.172* -0.144 -0.154
(0.0803) (0.0782) (0.0782) (0.0836) (0.0952) (0.124) (0.136)
VI x AdjQuota 0.0273** 0.0269** 0.0269** 0.0271** 0.0297** 0.0266 0.0270
(0.0128) (0.0127) (0.0127) (0.0125) (0.0131) (0.0182) (0.0192)
Herfindahl Index -0.0314 -0.0302 -0.0296 -0.0578 -0.0691 -0.0501 -0.0505
(0.0557) (0.0556) (0.0556) (0.0572) (0.0577) (0.0519) (0.0527)
Multinational 0.601 0.695 0.682 0.682 1.839 0.390
(0.599) (0.647) (0.654) (0.646) (1.753) (0.661)
ISO Certified -0.468 -0.506 -0.512 -0.979 -0.513
(0.799) (0.794) (0.790) (1.289) (0.453)
K/L (-1) 0.136 0.171 0.171 0.176
(0.131) (0.137) (0.176) (0.183)
Size (-1) 0.0393* 0.0570** 0.0575**
(0.0209) (0.0249) (0.0249)
Age 1.117* 1.103**
(0.590) (0.507)
Age
2
-0.279 -0.282*
(0.228) (0.166)
Constant 8.201** 0 0 0 8.658* 0 6.542
(3.928) (0) (0) (0) (4.854) (0) (7.022)
City Fixed Effects Yes Yes Yes Yes Yes Yes 4o
Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes
No. of Observations 948 948 948 896 896 555 555
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
Coefficients for the industry and year dummies are suppressed. *** Significant at, or below, 1 percent. ** Significant
at, or below, 5 percent. * Significant at, or below, 10 percent.
4.7 Productivity and Vertical Integration
4.7.1 Adjusted quotas and productivity
Table 4.5 exhibits the effect of elimination of quotas on productivity of clothing firms.
An increase in the adjusted level of quotas, on average, brings about a significant
reduction in the productivity of non-integrated clothing firms. This can be deduced from
the negative coefficient of adjusted quota base in almost all specifications shown in Table
4.5. On the other hand, the coefficient of the interaction term, i.e. the interaction of VI
dummy with adjusted level of quota, is positive and statistically significant: a higher trade
110
quota diminishes the mean productivity for all clothing firms significantly, though less so
for vertically integrated firms than for non-integrated firms. The productivity dropped for
vertically integrated firms, just not by as much as for non-integrated firms. These results
persist even when other methodologies are used to compute productivity.
Table 4.6 illustrates the results when we run the same regression using OP semi-
parametric productivity measure and the parametric estimation of productivity. This
shows that a trade liberalization episode causes not just a significant change in the
productivity of these firms, but also that the results differ immensely for the two types of
firms.
Table 4.6: Effect of Elimination of Quota-Restrictions on productivity of Clothing Firms – Olley & Pakes and OLS
Variable (1) (2) (3) (4) (5) (6)
OP OLS
Adjusted Quota -0.449** -0.430** -0.484* -0.159 -0.168 0.00451
(0.179) (0.181) (0.276) (0.150) (0.167) (0.259)
VI x Adjusted Quota 0.051*** 0.052*** 0.063** 0.0349** 0.0376** 0.0219
(0.0184) (0.0180) (0.0280) (0.0139) (0.0148) (0.0226)
Herfindahl Index -0.0152 -0.0848 0.0201 -0.00481 -0.0538 0.120
(0.107) (0.103) (0.142) (0.113) (0.109) (0.148)
Multinational -1.202 -1.692 -2.464 -1.676 -2.011 -2.470
(1.590) (1.988) (3.216) (1.386) (1.765) (2.735)
ISO Certified 0.254 0.767 0.753*** 0.677** 1.136**
(0.477) (0.606) (0.248) (0.269) (0.520)
K/L (-1) 0.201 0.0984 0.0482 -0.120
(0.194) (0.304) (0.181) (0.249)
Size (-1) 0.0748 0.111 0.0674 0.0645
(0.0496) (0.0736) (0.0506) (0.0805)
Age 2.479* 1.879
(1.299) (1.300)
Age
2
-0.579** -0.339
(0.258) (0.256)
Constant 19.87** 16.48* 17.07 6.107 0 0
(9.020) (9.015) (13.81) (7.455) (0) (0)
City Fixed Effects Yes Yes 4o Yes Yes Yes
Industry/Time Effects Yes Yes Yes Yes Yes Yes
No. of Observations 948 896 555 948 896 555
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
Coefficients for the industry and year dummies are suppressed. *** Significant at, or below, 1 percent. ** Significant
at, or below, 5 percent. * Significant at, or below, 10 percent.
111
Let us analyze the implications of these results in terms of the model described
above. Figure 4.3 shows what happens if the relative productivity function, A(z), shifts
up. In this case, there is an increase in the product range produced by firm 3 since the cut-
off goes up from s
Q
to s
Q
––
. Trade liberalization in the form of a reduction in price of a
quota license will cause a reduction in the relative factor cost of firm 2 compared to firm
3. If other factors which affect relative productivity cause the relative productivity
function to shift up, then the product range produced by vertically integrated clothing
producers rises.
Table 4.7: Effect of Elimination of Quota-Restrictions on productivity of Clothing Firms – Across Phases
Variable Phase 1 Phase 2 Phase 3 Phase 4
LP OP LP OP LP OP LP OP
Adjusted Quota -0.104 0.275 -0.0474 0.470 -0.0435 0.278 -0.0302 -0.35***
(0.0960) (0.938) (0.0596) (0.364) (0.160) (0.231) (0.0384) (0.0545)
VI x AdjQuota 1.002*** 0.480*** 1.021*** 0.519*** 0.984*** 0.617*** 0.346*** 0.189***
(0.0221) (0.0636) (0.0269) (0.0840) (0.0355) (0.0532) (0.0100) (0.0129)
Age 0.844 2.021 0.135 4.114 0.448 -2.42*** 1.191 5.629***
(0.538) (2.125) (0.136) (3.236) (0.318) (0.756) (1.308) (1.831)
Age
2
-0.153 -0.424 -0.0328 -0.768 -0.101 0.371** -0.225 -0.99***
(0.120) (0.495) (0.0431) (0.647) (0.0983) (0.182) (0.224) (0.312)
Size (-1) 0.0763** 0.121 0.0836 -0.0668 0.0318 -0.353* 0.0336 0.161*
(0.0339) (0.155) (0.0573) (0.127) (0.0399) (0.207) (0.0268) (0.0850)
K/L (-1) 0.315 0.385 0.0248 -0.374 0.0729 0.228 -0.00778 0.0114
(0.257) (0.983) (0.0823) (0.504) (0.0926) (0.625) (0.0745) (0.415)
Herfindahl Index -0.146* 0.273 -0.182* -0.645 -0.0323 -0.147 -0.0405 -0.174
(0.0778) (0.354) (0.0982) (0.421) (0.0983) (0.212) (0.0528) (0.216)
ISO Certified 0.0269 -0.531 -0.216 1.370 -0.244 0.623 -1.08*** 0.838*
(0.146) (0.887) (0.159) (1.012) (0.175) (0.476) (0.406) (0.430)
Multinational 0.176 -1.920 0.142 0.279 0.298 0.773 0.249 0.687
(0.371) (3.206) (0.185) (1.291) (0.506) (0.999) (0.260) (0.647)
Constant 0 0 -0.696 0 -0.358 0 0 0
(0) (0) (0.801) (0) (3.287) (0) (0) (0)
Industry/Time
Effects
Yes Yes Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes Yes Yes Yes
No. of Observations 107 107 160 160 66 66 192 192
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. (-1) denotes lagged variables.
*** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at, or below, 10 percent.
112
4.7.2 Productivity and other variables
Let us look at other control variables. As far as size of the firm is concerned, for all
specifications shown in the table, the coefficient of size is positive and statistically
significant. Although capital intensity does not have a significant impact on productivity
of clothing producers, nevertheless higher capital intensity is associated with a higher
productivity level. Furthermore, older firms are more productive on average than their
younger counterparts. The coefficient for multinational dummy takes both positive and
negative values and is never significant. The sign of herfindahl index coefficient is
negative and significant for most of the specifications. This means that higher
concentration in the industry leads to lower productivity of firms. One would normally
expect that a greater degree of concentration in an industry leads to greater market power
for firms in that industry and, hence, lower their productivity growth. This is the case
here.
ISO Certification does not affect firm efficiency significantly and takes both
positive and negative values. The only case where it does have a significant impact is
when it takes a positive value: a firm certified for its quality management system has a
higher productivity, on average, than a firm that is not certified (look at Table 4.6). The
city fixed effects take into account the fact that some firms are located in more developed
areas relative to other firms. There may also be differences in infrastructural facilities in
different parts of the country; regional fixed effects take care of these differences.
113
Table 4.8: Effect of Elimination of Quota-Restrictions on productivity of Clothing Firms - Interaction of VI with Firm’s
Size and Capital Intensity
Variable (1) (2) (3) (4) (5) (6) (7) (8)
Interaction of VI with Firm’s Size
Interaction of VI with Firm’s Capital
Intensity
Adjusted Quota -0.0163 -0.0104 -0.0030 -0.105 0.0970 -0.0673 -0.0553 -0.203
(0.0754) (0.0780) (0.0792) (0.122) (0.0677) (0.102) (0.0640) (0.166)
VI 13.1*** 13.2*** 13.0*** 13.1*** 15.3*** 14.3*** 14.8*** 14.3***
(0.996) (0.970) (0.981) (1.145) (0.563) (0.664) (0.602) (1.060)
Size (-1) -0.0080 -0.0086 0.00505 -0.0041 0.059** 0.08*** 0.0792*
(0.0165) (0.0165) (0.0280) (0.0456) (0.0231) (0.0282) (0.0437)
VI x Size (-1) 0.108** 0.108** 0.118** 0.138**
(0.0476) (0.0472) (0.0469) (0.0611)
K/L (-1) 0.293 0.524 0.148** 0.0545 -0.178* 0.0452
(0.209) (0.383) (0.0698) (0.0559) (0.0916) (0.0732)
VI x K/L (-1) 0.0479 0.521 0.917** 0.817
(0.292) (0.399) (0.432) (0.549)
Cost of Imports -2.771 -2.804 -2.870 -1.670 -2.982 -2.231 -2.557
(2.378) (2.410) (2.334) (2.801) (2.571) (3.130) (2.896)
Herfindahl Index -0.0397 -0.0417 -0.0400 -0.0512 -0.0598 -0.0312 -0.0672
(0.0792) (0.0795) (0.0797) (0.0710) (0.0853) (0.0671) (0.0759)
Multinational -0.0689 -0.119 -0.0734 0.139 -0.160 0.0765 0.184
(0.182) (0.171) (0.181) (0.394) (0.185) (0.210) (0.276)
ISO Certified 0.176 0.167 -0.0242 0.171 -0.179 -0.0354
(0.174) (0.174) (0.131) (0.174) (0.184) (0.123)
Age 0.693* 0.226 0.655*
(0.406) (0.318) (0.394)
Age
2
-0.163* -0.0372 -0.161*
(0.0975) (0.0767) (0.0935)
Constant 1.036 0.719 0 1.244 0 0 0.364 2.343
(1.497) (1.588) (0) (2.563) (0) (0) (1.371) (2.942)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes
City Fixed Effects Yes Yes Yes 4o Yes Yes Yes 4o
No. of Observations 566 566 565 347 896 565 347 347
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. VI is an indicator variable
denoting a vertically integrated clothing firm. (-1) denotes lagged variables. Coefficients for the industry and year
dummies are suppressed. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. * Significant at,
or below, 10 percent.
4.7.3 Adjusted quotas and productivity - Across phases
Table 4.7 demonstrates the effect of elimination of quota restrictions on productivity
across phases. Although a majority of the results described above stand across individual
phases as well, some of the findings differ across phases and for the two types of firms.
An increase in the adjusted level of quotas brings about a reduction in the mean
114
productivity of clothing firms that are not vertically integrated. This is true for all phases
under LP but only for Phase IV under OP. In Phases I, II and III, the sign of the
coefficient for level of quotas under OP is positive but insignificant. On the other hand, in
all the four phases, a growth in the adjusted level of quotas diminishes the mean
productivity of vertically integrated firms not by as much as for non-integrated firms, and
the coefficients are statistically significant. These estimation results are arrived at after
controlling for industry, time and city fixed effects.
4.7.4 Adjusted quotas and productivity – interaction with Size and Capital
intensity
Table 4.8 displays the results we get by including the interaction of VI with firm’s size
and capital intensity as right hand side variables. Although there is no strong evidence
that more capital intensive integrated firms perform better than the less capital intensive
ones, we do observe that larger vertically integrated firms, on average, outdo smaller
integrated firms. As far as other control variables are concerned, there is no noteworthy
change in either the sign or magnitude of their coefficients.
4.8 Adjusted quotas and other dependent variables
To test whether or not vertically integrated clothing firms fared better than non-integrated
firms in terms of other key measures of performance, such as output and net profit, we
replace productivity by firms’ output and net profit in Eq.(4.9). The results are depicted
in Table 4.9.
115
Table 4.9: Effect of Elimination of Quota-Restrictions on Net Profit and Output of Clothing Firms
Variable (1) (2) (3) (4) (5) (6)
4et Profit Output
Raw Materials 0.319*** 0.109* 0.127* 0.231*** 0.0273 0.0559
(0.0970) (0.0592) (0.0705) (0.0721) (0.0308) (0.0360)
Employment 0.388*** 0.162** 0.112 0.121*** 0.107* 0.141**
(0.0599) (0.0667) (0.0934) (0.0359) (0.0627) (0.0582)
Fixed Assets -0.933*** -0.686*** -0.822*** 0.0221* 0.0943** 0.116**
(0.0342) (0.129) (0.103) (0.0122) (0.0478) (0.0543)
Adjusted Quota -0.123 0.409 1.175*** -0.351 1.003** 1.048**
(0.327) (0.326) (0.300) (0.295) (0.419) (0.440)
VI x Adjusted Quota 0.0106 0.161*** 0.235*** 0.0204 0.0979 0.0335
(0.0289) (0.0497) (0.0530) (0.0277) (0.0975) (0.112)
Cost of Imports 0.893 6.189 13.40** 14.47***
(5.695) (7.041) (5.825) (4.680)
Herfindahl Index 0.178 0.205
(0.151) (0.197)
Multinational 0.599 1.180 -2.642 -2.127
(1.186) (1.335) (2.609) (2.677)
ISO Certified 1.417*** 1.583 2.343*** 1.334*
(0.497) (1.048) (0.847) (0.787)
Age 0.949 1.877 2.160
(0.867) (1.460) (1.325)
Age
2
-0.136 -0.302 -0.438*
(0.202) (0.326) (0.261)
Constant 19.07 5.577 -7.558 29.52** 0 -10.57
(15.19) (6.300) (5.445) (13.79) (0) (6.850)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes
City Fixed Effects Yes Yes 4o Yes Yes 4o
No. of Observations 948 617 374 948 374 374
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. Coefficients for the industry and
year dummies are suppressed. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. *
Significant at, or below, 10 percent.
Yet again, the results vary widely across the two kinds of firms. An increase in
the adjusted level of quotas leads to a significant increase in the mean output of non-
integrated firms. While once again it does lead to a higher output for integrated firms, the
positive coefficient of the interaction term in this case is not significant. This is not the
case if we replace output on the left hand side by net profit: the impact on net profit of all
clothing firms, both vertically integrated as well as non-integrated, is positive in all the
specifications used. This can be seen from columns (2) and (3) in Table 4.9. Moreover,
116
the coefficient of the interaction term (VI x Adjusted Quota) remains highly significant.
This analysis makes the interpretation of estimation results derived earlier all the more
fascinating.
4.9 Discussion
To sum up, what we have seen is that trade liberalization resulted in a fall in the mean
productivity of clothing firms in the T&C sector of Pakistan. However, vertical
integration of production is linked to a lesser decline in productivity. As mentioned in the
literature review above, there could be a large number of possible reasons why that might
hold. The most cited benefits to a firm through vertical integration are decreased
marketing expenses, stability of operations, tighter quality control, timely revision of
production policies, guarantee of supplies, improved inventory control, and the ability to
charge lower prices on final products. Vertically integrated firms respond to competition
by upgrading the quality of their clothing products. On the other hand, it is harder to
upgrade quality quickly for non-integrated firms. These differences in quality ladders
could play a crucial role under these circumstances. Suppliers usually control new
technology in the technologically advanced industries and internalizing these
technological capabilities through vertical integration promises access to the knowledge
required to build a portfolio of products based on highly developed technology (Afuah,
2001). Hortaçsu and Syverson (2007) show that integrated producers’ productivity
advantage is owing to the improved logistics coordination, but this advantage is not tied
to firms’ vertical structure per se since non-vertical firms have equally high productivity
117
levels. As long as intangible inputs such as managerial talent or marketing know-how are
complementary to the physical inputs in the process of making a vertically linked set of
products, equilibrium assignment allocates the top intangible inputs to various plants in
separate but often vertically related product categories. Textile quality, product standards,
fabric finishing, styles, and patterns are other factors that shape competitiveness.
Some theory papers incorporate the asymmetric effects of liberalization by the
productivity level of a firm. If non-exporting firms become exporters, we may see a
decline in mean industry productivity because new exporters need time to adapt to the
new environment. Another proposition is a change in product mix, for instance, a shift to
the most productive lines in vertically integrated firms and expansion into new
products for which there is still some learning to do in non-integrated garment industry.
Vertical integration sets out decision rights within the organization and lets firms more
adeptly respond to not just the changes in consumer demands but also to adjustments in
the cost of their input-mix.
One hypothesis is that as the MFA expired, the T&C industry in Pakistan may
have possibly changed the composition of its exports, from a broader category that
benefitted from MFA without much weight of Rules of Origin (RoO), to a narrower
category focused on specific markets that offer Pakistan preferential access through
bilateral trade agreements with strict Rules of Origin. If this is the case, one would
expect a greater fall in the productivity of non-integrated firms as the mix of inputs
utilized by them would no longer be dictated by rationally choosing the optimal input-
mix given market prices, but rather by what the RoO says. Therefore, the disparity in
118
results across garment firms may perhaps be related to the structure of production,
namely, the kind of raw materials used by garment firms after the end of MFA. However,
we do not have information about the types of raw materials used (for example, whether
or not they are imported), and consequently, it is hard to determine if this was the case.
4.10 Conclusion
Expansion of international specialization and disintegration of production has been an
important characteristic of the global economy. Although a variety of studies look into
productivity gains associated with the disintegration of production across different
countries, none of these in particular consider the effect of trade liberalization on the
efficiency of vertically integrated firms relative to non-integrated firms within a country.
In this chapter, we compare the productivity of these two types of firms, allowing both
types of firms to engage in international trade, and analyze how a given liberalization
episode affects their productivity. In particular, we analyze the experience of clothing
firms in Pakistan under the U.S. textile and clothing quotas and the subsequent end of
MFA.
The interaction between trade policies and firm characteristics is a subject of great
interest to trade economists at present. As is recurrently emphasized in the new trade
theory literature, trade reforms often influence firms heterogeneously even within the
same manufacturing industry. By merging micro-level data of firms with data on quotas
at the industry level, this chapter shows that a liberalization episode may engender
divergent changes in productivity levels of different types of firms even within the
119
clothing industry. Vertical integration is a firm characteristic that has the potential to
affect the impact of trade policy on firms. A theoretical background to vertical integration
in the clothing industry illustrates that trade liberalization causes a change in the relative
factor cost of the two types of firms, and thus, a change in the product range produced by
each of them. The empirical findings confirm that vertical integration of production is
linked to a smaller reduction in productivity after trade reforms were introduced in the
clothing industry of Pakistan. Increases in the adjusted level of quotas bring about a
significant reduction in the mean productivity of clothing firms that are not vertically
integrated. The most cited benefits to a firm through vertical integration are tighter
quality control, timely revision of production policies, and greater stability in supplies.
Vertical integration sets out decision rights within the organization and lets firms more
proficiently respond to changes in consumer demands. The declining efficiency of non-
integrated clothing firms points to the inability of these firms to benefit from stability of
operations, investment in specialized assets, and decreased marketing expenses. MFA
expiration is a chance for them to trim down their input usage which can help yield a
competitive edge over other clothing exporters.
The elimination of quotas has been the most important event in the global textile
and garment industry during the past two decades. The textile sector is a key industry in
Pakistan in terms of output, export value, foreign exchange earnings and employment.
T&C make up almost 75 percent of the total export value. Pakistan is the fourth largest
producer of cotton in the world and does not have to count on other countries for its raw
materials. Furthermore, labor costs in Pakistan are among the lowest in the world. Along
120
with the cost advantage in terms of proximity to a raw material base in cotton and man-
made fibres, as well as the availability of cheap labor, what appears to be a crucial
determinant of competitiveness in this industry is the ability to respond to rapidly
changing consumer demands. This, in turn, requires greater investment in research and
development in order to ensure greater mobility and adaptability of the production
process to changes in fashion trends. Although the need to invest in cost-saving
production methods is vital for the textile industry as well, it plays a greater role in the
clothing industry owing to the nature of the finished good and its global price sensitivity.
The sectoral heterogeneity in the effect of MFA expiration further corroborates this
notion. The finding that mean productivity fell for the clothing firms, and more so for the
non-integrated clothing firms, as a result of the phasing out of quotas, points to the
inability of these firms to shift to a more efficient composition of inputs as well as the
product range of output produced in response to a more competitive world market. For
example, according to a report by the World Bank’s Poverty Reduction and Economic
Management Sector Unit, compared to its competitors, Pakistani garment industry labor
is cheaper but the least productive: limited training in productivity, design, and other
product related skills are the major constraints to raising productivity, and clothing firms
have been unable to tailor products particularly for their customers, deliver fast and
within multiple fashion cycles in one season (PAKISTAN Growth and Export
Competitiveness, 2006). Even though several institutions for training and skills upgrade
are present, in general, the country has an insufficient number of institutes that offer
support services to garment firms. Moreover, the industry faces the challenge of obsolete
121
machinery and export concentration in low value-added products. According to the
report, higher efficiency at the firm level is necessitated in order to compensate for the
time costs associated with higher distance to the U.S. market.
122
Bibliography
Acemoglu, D., Johnson, S., & Mitton, T. (2005). Integration: Finance, Contracts, and
Regulation. 4BER Working Paper 11424 .
Afuah, A. (2001). Dynamic boundaries of the firm: Are firms better off being vertically
integrated in the face of a technological change? Academy of Management Journal,
44, 1211–1228.
All Pakistan Textile Mills Association (APTMA). (2011). Retrieved from Pakistan Textile
Statistics: http://www.aptma.org.pk/Pak_Textile_Statistics/repo.asp
Amiti, M., & Konings, J. (2007). Trade Liberalization, Intermediate Inputs, and
Productivity: Evidence from Indonesia. American Economic Review, 97, 1611–
1637.
Antràs, P., & Helpman, E. (2004). Global sourcing. Journal of Political Economy, 112,
552–580.
Babe, R. E. (1981). Vertical Integration and Productivity: Canadian Telecommunications.
Journal of Economic Issues, 15.1, 1-31.
Baldwin, R. E. (2005). Heterogeneous Firms and Trade: Testable and Untestable
Properties of the Melitz Model. 4BER Working Paper 4o. 11471 .
Bernard, A. B., Jensen, J. B., & Schott, P. K. (2006). Trade Costs, Firms, and
Productivity. Journal of Monetary Economics, 53(5), 917–37.
Bernard, A. B., Jensen, J. B., Schott, P. K., & Redding, S. J. (2007). Firms in
International Trade. Journal of Economic Perspectives, 21(3), 105-130.
Biesebroeck, J. V. (2007). Robustness of Productivity Estimates. Journal of Industrial
Economics, 55(3), 529-569.
Blois, K. J. (1972). Vertical Quasi-Integration. The Journal of Industrial Economics,
20.3, 253-72.
123
Brambilla, I., Khandelwal, A., & Schott, P. (2007). China's Experience under the
Multifiber Arrangement (MFA) and the Agreement on Textiles and Clothing
(ATC). 4BER Working Papers 13346 .
Cohen-Meidan, M. (2009). Vertical Integration and Trade Protection: The Case of
Antidumping Duties. SIEPR Discussion Paper 4o. 08-34 .
Collins, J. L. (2003). Threads: Gender, Labor, and Power in the Global Apparel
Industry. Chicago: University of Chicago Press.
Corden, M. W. (1971). The Theory of Protection. Oxford: Oxford University Press.
Corden, M. W. (1974). Trade Policy and Economic Welfare. Oxford: Clarendon Press.
Demidova, S., Kee, H. L., & Krishna, K. (2006). Do Trade Policy Differences Induce
Sorting? Theory and Evidence from Bangladeshi Apparel Exporters. 4BER
Working Paper 4o. 12725 .
Devarajan, S., & Rodrik, D. (1991). Pro-competitive effects of trade reform: results from
a CGE model of Cameroon. European Economic Review, 35, 1157–1184.
Diao, X., & Somwaru, A. (2001). Impact of the MFA Phase-Out on the World Economy:
An Intertemporal Global General Equilibrium Analysis. International Food Policy
Research Institute, Trade and Macroeconomics Division.
Domowitz, I., Hubbard, R., & Petersen, B. (1988). Market structure and cyclical
fluctuations in US manufacturing. The Review of Economics and Statistics, 70(1),
55-66.
Eaton, J., & Kortum, S. (2002). Technology, Geography, and Trade. Econometrica, 70,
1741–79.
Ernst, C., Ferrer, A. H., & Zult, D. (2005). The end of the Multi-Fibre Arrangement and
its implication for trade and employment. Employment Strategy Papers,
International Labour Organization (ILO), Employment Analysis Unit, Employment
Strategy Department.
Evans, C. L., & Harrigan, J. (2005). Tight Clothing: How the MFA Affects Asian
Apparel Exports. 4BER Working Paper 10250, 14 , 367-390.
Feenstra, R. C. (1996). U.S. Imports, 1972–1994: Data and concordances. 4BER
Working Paper 5515.
124
Forbes, S., & Lederman, M. (2010). Does Vertical Integration Affect Firm Performance?
Evidence from the Airline Industry. The RA4D Journal of Economics, 41 (4), 765-
90.
Frazer, G., & Biesebroeck, J. V. (2007). Trade Growth under the African Growth and
Opportunity Act. 4BER Working Papers 13222 .
Goh, A. (2000). Opportunity Cost, Trade Policies and the Efficiency of Firms. Journal of
Development Economics, 62, 363-83.
Goldberg, P. K., & Pavcnik, N. (2005). Trade, Wages, and the Political Economy of
Trade Protection: Evidence From the Colombian Trade Reforms. Journal of
International Economics, 66(1), 75–105.
Golub, S. S., & Hsieh, C. T. (2000). Classical Ricardian Theory of Comparative
Advantage Revisited. Review of International Economics, 8, 221–34.
Grossman, G. M., & Helpman, E. (1991). Innovation and Growth in the Global
Economy. Cambridge, MA: MIT Press.
Grossman, G. M., & Helpman, E. (1994). Protection for Sale. American Economic
Review, 84(4), 833–50.
Grossman, G. M., Helpman, E., & Szeidl, A. (2006). Optimal Integration Strategies for
the Multinational Firm. Journal of International Economics, 70, 216-38.
Hall, R. (1988). The relation between price and marginal cost in the US industry. Journal
of Political Economy, 96(5), 921–947.
Harrigan, K. R. (1984). Formulating vertical integration strategies. Academy of
Management Review, 9, 638–652.
Harrison, A. E. (1994). Productivity, imperfect competition and trade reform: Theory and
Evidence. Journal of International Economics, 36, 53–73.
Hart, O., & Moore, J. (1990). Property Rights and the Nature of the Firm. Journal of
Political Economy, 98, 1119-1158.
Hastert, D. (n.d.). Middle East Gate Exhibition Services. Retrieved May 30, 2012, from
http://www.megexhibition.com/middle_east_gate_exhibition_stands_int.htm
125
Haté, A., Khanal, S., Larsen, J., Smart, P., Soria, R., & Zanni, D. (2005). The Expiration
of the Multi-Fibre Arrangement: An Analysis of the Consequences for South Asia.
Public Affairs 860: Public Affairs Workshop, International Issues. University of
Wisconsin-Madison: Robert M. La Follette School of Public Affairs.
Hennessy, D. A. (1996). Information Asymmetry as a Reason for Food Industry Vertical
Integration. American Journal of Agricultural Economics, 78(4), 1034-043.
Hortacsu, A., & Syverson, C. (2007). Cementing Relationships: Vertical Integration,
Foreclosure, Productivity, and Prices. Journal of Political Economy, 115(2), 250-
301.
Hortacsu, A., & Syverson, C. (2009). Why do Firms Own Production Chains? US Census
Bureau Center for Economic Studies Paper 4o. CES 09 -31.
Hsiao, C. (1990). Analysis of Panel Data. Cambridge University Press.
Hummels, D., Ishii, J., & Yi, K.-M. (2001). The Nature and Growth of Vertical
Specialization in World Trade. Journal of International Economics, 54, 75–96.
(2000). International Trade and Development: Data on Import Tariff and 4TBs. World
Bank, Washington, D.C. Retrieved from
www1.worldbank.org/wbiep/trade/TR_data.html
Jones, L. E. (2003). Textiles and Apparel: Assessment of the Competitiveness of Certain
Foreign Suppliers to the U.S. Market. U.S. International Trade Commission
Investigation, U.S. Association of Importers of Textile and Apparel.
Krishna, P., & Mitra, D. (1998). Trade Liberalization, Market Discipline and Productivity
Growth: New Evidence from India. Journal of Development Economics, 56(2), 447-
462.
Krueger, A. O., & Tuncer, B. (1982). Growth of Factor Productivity in Turkish
Manufacturing Industries. Journal of Development Economics, 11(3), 307-25.
Krugman, P. (1986). New Trade Theories and the Less-Developed Countries. In G.
Calvo, Debt, Stabilization and Development: Essays in Honor of Carlos Diaz-
Alejandro. New York: Basil Blackwell.
Kumar, K., Rajan, R., & Zingales, L. (1999). What Determines Firm Size? 4BER
Working Paper 4o. 7208 .
Lambson, V. E. (1989). Industry Evolution with Sunk Costs and Uncertain Market
Conditions. SSRI Working Paper 8904.
126
Lankes, H. P. (2002). Market Access for Developing Countries. Finance and
Development, 39(3).
Levinsohn, J., & Petrin, A. (2003). Estimating Production Functions using Inputs to
control for Unobservables. Review of Economic Studies, 70, 317-341.
Lindh, A. (n.d.). BrainyQuote.com. Retrieved May 30, 2012, from
http://www.brainyquote.com/quotes/quotes/a/annalindh319761.html
Lindh, A. (n.d.). BrainyQuote.com. Retrieved May 30, 2012, from
http://www.brainyquote.com/quotes/quotes/a/annalindh319749.html
Melitz, M. (2003). The impact of trade on intraindustry reallocations and aggregate
industry productivity. Econometrica, 71, 1695-1725.
Miyagawa, K., & Ohno, Y. (1995). Closing the Technology Gap under Protection.
American Economic Review, 85, 755-70.
Mobarak, A. M., & Purbasari, D. (2006). Corrupt Protection for Sale to Firms: Evidence
from Indonesia. Unpublished.
Mundlak, Y. (1961). Empirical Production Function Free of Management Bias. Journal
of Farm Economics, 44-56.
Obama, B. (2009). Words on a Journey: The Great Speeches of Barack Obama.
Rockville, Maryland: Arc Manor.
Olley, G. S., & Pakes, A. (1996). The Dynamics of Productivity in the
Telecommunications Equipment Industry. Econometrica, 64(6), 1263–1297.
Pakes, A. (1994). Dynamic structural models, problems and prospects. In C. Sims (Ed.),
Advances in Econometrics. Sixth World Congress, Cambridge University Press.
(2006). PAKISTA4 Growth and Export Competitiveness. World Bank, Poverty Reduction
and Economic Management Sector Unit, South Asia Region.
(2002). Pakistan: Trade Policy Review. World Trade Organization.
Pavcnik, N. (2002). Trade Liberalization, Exit, and Productivity Improvement: Evidence
from Chilean Plants. Review of Economic Studies, 245–276.
Roberts, M., & Tybout, J. (1991). Size Rationalization and Trade Exposure in
Developing Countries. In R. Baldwin, Empirical Studies of Commercial Policy.
Chicago: U. Chicago Press for NBER.
127
Rodrik, D. (1992). Closing the Technology Gap: Does Trade Liberalization Really Help?
In G. Helleiner, Trade Policy, Industrialization and Development: 4e w
Perspectives. Oxford: Clarendon Press.
Rodrik, D., & Rodriguez, F. (2000). Trade Policy and Economic Growth: A Skeptics
Guide to the Cross-National Evidence. In B. Bernanke, & K. Rogoff, 4BER
Macroeconomics Annual 2000. Cambridge, MA: MIT Press.
Rothaermel, F. T., Hitt, M. A., & Jobe, L. A. (2006). Balancing Vertical Integration and
Strategic Outsourcing: Effects on Product Portfolio, Product Success, and Firm
Performance. Strategic Management Journal, 27, 1033-056.
Sachs, J. D., & Warner, A. (1995). Economic Reform and the Process of Global
Integration. Brookings Papers on Economic Activity, 1-118.
Sasidaran, G., & Shanmugam, K. R. (2008). Impact of trade liberalization on the
efficiency of textile firms in India. ISAS working Paper, 52.
Taft, W. H. (n.d.). William Howard Taft. Retrieved May 30, 2012, from
http://www.allthingswilliam.com/presidents/taft.html
(2004). Textiles and Apparel: Assessment of the Competitiveness of Certain Foreign
Suppliers to the U.S. Market. Investigation No. 332-448, U.S. International Trade
Commission.
Tomiura, E. (2007). Foreign Outsourcing, Exporting, and FDI: A Productivity
Comparison at the Firm Level. Journal of International Economics, 72, 113-27.
(1996–2002). Trade Policy Reviews. World Trade Organization.
Trefler, D. (2004). The Long and Short of the Canada - U.S. Free Trade Agreement.
American Economic Review, 94(4), 870–95.
Tyagi, R. (2003). Apparel Globalization: The Big Picture. Natural Fibers Information
Center. Austin: University of Texas at Austin.
Tybout, J. (1992). Linking Trade and Productivity: New Research Directions. World
Bank Economic Review, 6, 189-212.
Tybout, J. (2003). Plant- and Firm-Level Evidence on the 'New' Trade Theories. In E. K.
Choi, & J. Harrigan, Handbook of International Trade. Oxford: Basil Blackwell.
128
Tybout, J., & Westbrook, M. D. (1995). Trade liberalization and the dimensions of
efficiency change in Mexican manufacturing industries. Journal of International
Economics, 39, 53-78.
Tybout, J., Melo, J., & Corbo, V. (1991). The Effects of Trade Reforms on Scale and
Technical Efficiency: New Evidence from Chile. Journal of International
Economics, 31, 231-59.
U4SD Statistical Databases . (n.d.). Retrieved from The United Nations Statistics
Division:
http://unstats.un.org/unsd/databases.htm
Wacziarg, R. (2001). Measuring the Dynamic Gains from Trade. World Bank Economic
Review, 15(3), 393-429.
Wacziarg, R., & Welch, K. H. (2008). Trade Liberalization and Growth: New Evidence.
The World Bank Economic Review, 22(2), 187–231. doi:10.1093/wber/lhn007
Williamson, O. (1975). Markets and Hierarchies: Analysis and Antitrust Implications.
New York: Free Press.
(2006). World Development Indicators. Washington D.C.: The World Bank.
World Trade Organization. (2009). Retrieved from Pakistan: The Tariff database,
Integrated Database:
http://www.wto.org/english/thewto_e/countries_e/pakistan_e.htm
Yi, K.-M. (2003). Can Vertical Specialization Explain the Growth of World Trade?
Journal of Political Economy, 111(1), 52-102.
129
Appendix A
Supplementary Figures and Tables
A.1 Figures for Chapter 2
Figure A.1.1: Change in Returns to Scale (1995-2003) – Sugar
0 2 4 6 8
Ln(average cost)
10 15 20 25
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
SUGAR
130
Figure A.1.2: Change in Returns to Scale (1995-2003) – Textile
Figure A.1.3: Change in Returns to Scale (1995-2003) – Apparel
0 2 4 6 8
Ln(average cost)
14 16 18 20 22 24
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
TEXTILE
.4 .6 .8 1 1.2 1.4
Ln(average cost)
0 5 10 15 20 25
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
APPAREL
131
Figure A.1.4: Change in Returns to Scale (1995-2003) – Paper & Printing
Figure A.1.5: Change in Returns to Scale (1995-2003) – Medicine
0 .5 1 1.5 2
Ln(average cost)
14 16 18 20 22
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
PAPER & PRINTING
.4 .6 .8 1 1.2
Ln(average cost)
14 16 18 20 22
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
MEDICINE
132
Figure A.1.6: Change in Returns to Scale (1995-2003) – Petroleum & Oil
Figure A.1.7: Change in Returns to Scale (1995-2003) – Chemicals
.4 .5 .6 .7 .8 .9
Ln(average cost)
16 18 20 22 24 26
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
PETROLEUM & OIL
.5 1 1.5 2
Ln(average cost)
10 15 20 25
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
CHEMICALS
133
Figure A.1.8: Change in Returns to Scale (1995-2003) – Non-Metallic Materials
Figure A.1.9: Change in Returns to Scale (1995-2003) – Automobile
.2 .4 .6 .8 1 1.2
Ln(average cost)
14 16 18 20 22
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
NON-METALLIC MATERIALS
.6 .8 1 1.2 1.4
Ln(average cost)
14 16 18 20 22 24
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
AUTOMOBILE
134
Figure A.1.10: Change in Returns to Scale (1995-2003) – Construction Materials
Figure A.1.11: Change in Returns to Scale (1995-2003) – Electronics
0 .5 1 1.5 2
Ln(average cost)
12 14 16 18 20 22
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
CONSTRUCTION MATERIALS
.5 1 1.5 2 2.5
Ln(average cost)
14 16 18 20 22 24
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
ELECTRONICS
135
Figure A.1.12: Change in Returns to Scale (1995-2003) – Miscellaneous
Figure A.1.13: Change in Plant Size Distribution (1995-2003) – Food
0 .5 1 1.5 2
Ln(average cost)
10 15 20 25
Ln(output)
95% CI Predicted Ln(AC) in 1995
1995 Predicted Ln(AC) in 2003
2003
MISCELLANEOUS
.05 .1 .15 .2 .25
kernel density
14 16 18 20 22
Ln(output)
1995 2003
FOOD
136
Figure A.1.14: Change in Plant Size Distribution (1995-2003) – Sugar
Figure A.1.15: Change in Plant Size Distribution (1995-2003) – Apparel
0 .2 .4 .6 .8 1
kernel density
10 15 20 25
Ln(output)
1995 2003
SUGAR
0 .1 .2 .3 .4 .5
kernel density
0 5 10 15 20 25
Ln(output)
1995 2003
APPAREL
137
Figure A.1.16: Change in Plant Size Distribution (1995-2003) – Paper & Printing
Figure A.1.17: Change in Plant Size Distribution (1995-2003) – Medicine
.05 .1 .15 .2 .25
kernel density
14 16 18 20 22
Ln(output)
1995 2003
PAPER & PRINTING
0 .1 .2 .3 .4
kernel density
14 16 18 20 22
Ln(output)
1995 2003
MEDICINE
138
Figure A.1.18: Change in Plant Size Distribution (1995-2003) – Personal Care Products
Figure A.1.19: Change in Plant Size Distribution (1995-2003) – Leather
0 .05 .1 .15 .2 .25
kernel density
16 18 20 22 24
Ln(output)
1995 2003
PERSONAL CARE PRODUCTS
.1 .15 .2 .25 .3
kernel density
19 19.5 20 20.5 21 21.5
Ln(output)
1995 2003
LEATHER
139
Figure A.1.20: Change in Plant Size Distribution (1995-2003) – Petroleum & Oil
Figure A.1.21: Change in Plant Size Distribution (1995-2003) – Chemicals
.08 .1 .12 .14 .16
kernel density
16 18 20 22 24 26
Ln(output)
1995 2003
PETROLEUM & OIL
0 .05 .1 .15 .2
kernel density
10 15 20 25
Ln(output)
1995 2003
CHEMICALS
140
Figure A.1.22: Change in Plant Size Distribution (1995-2003) – Metal Tools & Products
Figure A.1.23: Change in Plant Size Distribution (1995-2003) – Construction Materials
.05 .1 .15 .2 .25
kernel density
12 14 16 18 20
Ln(output)
1995 2003
METAL TOOLS & PRODUCTS
0 .5 1 1.5
kernel density
12 14 16 18 20 22
Ln(output)
1995 2003
CONSTRUCTION
141
Figure A.1.24: Change in Plant Size Distribution (1995-2003) – Electronics
Figure A.1.25: Change in Plant Size Distribution (1995-2003) – Energy
0 .05 .1 .15 .2
kernel density
14 16 18 20 22 24
Ln(output)
1995 2003
ELECTRONICS
0 .1 .2 .3
kernel density
15 20 25
Ln(output)
1995 2003
ENERGY
142
Figure A.1.26: Change in Plant Size Distribution (1995-2003) – Non-Metallic Materials
Figure A.1.27: Change in Plant Size Distribution (1995-2003) – Miscellaneous
.15 .2 .25 .3 .35
kernel density
14 16 18 20 22
Ln(output)
1995 2003
NON-METALLIC MATERIALS
0 .05 .1 .15
kernel density
10 15 20 25
Ln(output)
1995 2003
MISCELLANEOUS
143
A.2 Tables for Chapter 3
Table A.2.1: Effect of Elimination of Quota-Restrictions on Firm Productivity – Levinsohn & Petrin
Entire Sample Post-MFA Period
All T&C Textile Clothing All T&C Textile Clothing
Adjusted Quota -0.222 1.692** -0.753*** 0.0749 1.744** -0.259**
(0.314) (0.850) (0.195) (0.229) (0.860) (0.121)
Cost of Imports -0.0895 0.0971 -11.22 0.0171 0.131 -1.653
(0.164) (0.173) (8.823) (0.158) (0.177) (2.582)
Age 0.581** 0.117 0.104 0.732** 0.232 0.323
(0.270) (0.222) (0.513) (0.364) (0.366) (0.667)
Age
2
-0.148 0.0346 0.277 -0.206* 0.0136 0.101
(0.0967) (0.0510) (0.248) (0.113) (0.0736) (0.233)
Size (-1) -0.0161 -0.0273 0.0716 -0.00878 -0.0210 0.0827*
(0.0179) (0.0198) (0.0572) (0.0193) (0.0225) (0.0492)
K/L (-1) 0.0432 -0.0696 0.807 0.0335 -0.0589 0.887
(0.138) (0.0883) (0.700) (0.119) (0.0913) (0.576)
Herfindahl Index (-1) 0.0197 0.0971* -0.182** 0.0235 0.112* -0.175**
(0.0443) (0.0566) (0.0782) (0.0485) (0.0650) (0.0764)
ISO Certified 0.128 0.839 1.719 0.152 0.830 2.469
(0.578) (0.574) (2.174) (0.599) (0.576) (2.346)
Multinational 0.453 0.162 -2.368 0.557 0.200 -1.263
(0.350) (0.261) (1.981) (0.366) (0.261) (1.845)
Constant 3.495 0 0 -1.596 0 0
(5.668) (0) (0) (4.349) (0) (0)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes Yes
No. of Observations 1343 996 315 1222 905 289
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. This table reports the effect of
elimination of quota-restrictions on productivity of T&C firms across phases using the Levinsohn & Petrin productivity
measure. (-1) denotes lagged variables. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. *
Significant at, or below, 10 percent.
144
Table A.2.2: Effect of Elimination of Quota-Restrictions on Firm Productivity – Levinsohn & Petrin
Phase 1 Phase 2 Phase 3 Phase 4
Textile Clothing Textile Clothing Textile Clothing Textile Clothing
Adjusted 0.862 -0.466 0.845 -1.42** 6.039*** -2.291* 0.0546 0.0200
Quota (0.539) (0.362) (0.673) (0.682) (0.890) (1.230) (0.108) (0.155)
Cost of 0.00578 4.801 -6.675 5.707 2.009 -1.788 - -
Imports (0.289) (3.895) (4.732) (5.436) (6.765) (8.671) - -
Age 0.364 1.524* -0.193 0.00114 2.098 0.0364 4.530 -4.135
(0.449) (0.840) (1.007) (0.283) (2.655) (0.867) (3.088) (3.293)
Age
2
-0.0015 -0.171 0.0873 0.0409 -0.319 0.0312 -0.629 1.279
(0.0870) (0.380) (0.188) (0.240) (0.411) (0.433) (0.457) (0.859)
Size (-1) 0.0271 0.106* 0.0196 0.119 0.0985 0.122*** 0.0348* 0.0370
(0.0328) (0.0642) (0.0284) (0.0737) (0.0645) (0.0372) (0.0188) (0.0293)
K/L (-1) -0.26** 0.618 0.0456 0.163 0.0513 2.254*** -0.25** 0.0254
(0.116) (0.579) (0.121) (0.205) (0.236) (0.779) (0.102) (0.0816)
Herfindahl 0.0734 -0.101 0.00460 -0.162* 0.190 0.0704 -0.0418 -0.0069
Index (0.0600) (0.0741) (0.0477) (0.0857) (0.167) (0.107) (0.0504) (0.0539)
ISO Certified 0.00166 -0.0829 1.099 0.647 0.789 -1.939 1.713 -0.972
(0.241) (5.173) (0.826) (2.137) (0.793) (2.815) (1.071) (8.757)
Multinational -0.217 -0.557 0.429 -0.712 0.154 0.666 -0.528 -5.204
(0.309) (1.888) (0.300) (1.859) (0.256) (3.135) (0.401) (3.246)
Constant 0 0 2.036 15.17 -106.2** 0 0 0
(0) (0) (14.52) (12.41) (17.02) (0) (0) (0)
Time effects Yes Yes Yes Yes Yes Yes Yes Yes
City effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry
effects
Yes Yes Yes Yes Yes Yes Yes Yes
No. of
Observations
298 89 405 139 202 61 645 192
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. This table reports the effect of
elimination of quota-restrictions on productivity of T&C firms across phases using the Levinsohn & Petrin productivity
measure. (-1) denotes lagged variables. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. *
Significant at, or below, 10 percent.
145
Table A.2.3: Effect of Elimination of Quota-Restrictions on Firm Productivity – Olley & Pakes
Entire Sample Post-MFA Period
All T&C Textile Clothing All T&C Textile Clothing
Adjusted Quota Level 0.405 2.047*** -0.646 0.429 2.167*** -0.905*
(0.687) (0.727) (0.575) (0.631) (0.674) (0.531)
Cost of Imports -0.0806 0.103 -5.612 -0.0439 0.150 0.232
(0.252) (0.225) (11.24) (0.249) (0.228) (6.999)
Age 1.543* -0.00227 3.530* 2.131 0.00451 4.250
(0.876) (0.281) (2.100) (1.368) (0.455) (2.676)
Age
2
-0.350* 0.0372 -0.691 -0.459* 0.0336 -0.850
(0.184) (0.0574) (0.484) (0.266) (0.0863) (0.575)
Size (-1) -0.0198 -0.0762*** 0.0278 -0.00107 -0.0706** 0.0789
(0.0311) (0.0280) (0.0837) (0.0333) (0.0315) (0.0907)
K/L (-1) 0.0210 -0.187* -0.00531 0.0245 -0.182* 0.307
(0.170) (0.107) (0.777) (0.163) (0.109) (0.591)
Herfindahl Index 0.0934 0.162** -0.113 0.114 0.198** -0.0713
(0.0732) (0.0800) (0.224) (0.0816) (0.0901) (0.239)
ISO Certified 0.873* 0.583 2.066* 0.791 0.524 1.627
(0.523) (0.459) (1.087) (0.518) (0.464) (1.052)
Multinational -0.132 0.234 -3.935 0.128 0.319 -2.393
(0.575) (0.200) (3.583) (0.457) (0.202) (2.978)
Constant -10.56 0 0 -11.67 -37.69*** 1.194
(12.48) (0) (0) (11.37) (12.79) (12.02)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Time Fixed Effects Yes Yes Yes Yes Yes Yes
City Effects Yes Yes Yes Yes Yes Yes
No. of Observations 1343 996 315 1222 905 289
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. This table reports the effect of
elimination of quota-restrictions on productivity of T&C firms across phases using the Olley & Pakes productivity
measure. (-1) denotes lagged variables. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. *
Significant at, or below, 10 percent.
146
Table A.2.4: Effect of Elimination of Quota-Restrictions on Firm Productivity – Olley & Pakes
Phase 1 Phase 2 Phase 3 Phase 4
Textile Clothing Textile Clothing Textile Clothing Textile Clothing
Adjusted 0.318 -0.802 0.0418 -2.2*** 7.748*** -2.554** -0.0962 0.0629
Quota (0.370) (0.697) (0.604) (0.505) (1.063) (1.016) (0.116) (0.153)
Cost of -0.172 4.016 -12.66 8.659 8.095 3.495 - -
Imports (0.353) (13.84) (8.310) (26.99) (9.156) (31.78) - -
Age -1.160** 2.400 1.680 6.992* 3.139 0.515 3.103 -4.971
(0.481) (2.066) (1.252) (4.251) (2.606) (2.807) (3.210) (5.376)
Age
2
0.279*** -0.502 -0.269 -1.392* -0.501 -0.257 -0.407 1.070
(0.104) (0.493) (0.230) (0.818) (0.401) (0.624) (0.470) (0.952)
Size (-1) -0.00760 0.155 -0.0176 0.0888 -0.0398 -0.207 0.0362 0.323**
(0.0419) (0.160) (0.0281) (0.165) (0.0880) (0.259) (0.0346) (0.151)
K/L (-1) -0.168 0.0208 -0.0760 0.562 -0.159 0.553 -0.194 0.631
(0.160) (1.442) (0.116) (0.925) (0.258) (2.497) (0.130) (0.576)
Herfindahl 0.0170 0.194 0.0817 -0.360 0.201 -0.217 -0.0231 0.524*
Index (0.0801) (0.331) (0.0648) (0.367) (0.216) (0.407) (0.0951) (0.288)
ISO Certified -0.355* -0.200 0.963 0.646 0.605 -0.444 0.885* 1.079
(0.207) (2.433) (0.776) (1.482) (0.727) (1.528) (0.523) (3.864)
Multinational -0.470** -2.306 0.804** 1.762 0.285 1.815 -0.392 -2.769
(0.237) (3.703) (0.359) (1.806) (0.249) (2.647) (0.242) (1.868)
Constant 0 2.531 0 0 -147.4** 44.80*** 0 0
(0) (13.56) (0) (0) (21.45) (13.82) (0) (0)
Time effects Yes Yes Yes Yes Yes Yes Yes Yes
City effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry
effects
Yes Yes Yes Yes Yes Yes Yes Yes
No. of
Observations
298 89 405 139 202 61 645 192
4otes: Robust standard errors corrected for clustering at the firm level in parentheses. This table reports the effect of
elimination of quota-restrictions on productivity of T&C firms across phases using the Olley & Pakes productivity
measure. (-1) denotes lagged variables. *** Significant at, or below, 1 percent. ** Significant at, or below, 5 percent. *
Significant at, or below, 10 percent.
147
Appendix B
Review of Olley & Pakes and Levinsohn & Petrin
This section provides a review of the techniques of Olley & Pakes and Levinsohn &
Petrin. Consider the following Cobb-Douglas production function:
=
+
+
+ �
+ w
. B. 1
is the log of output,
is the log of capital input, and
is the log of labor input. The
OP methodology allows the error term to have two components, a white noise component
and a time-varying productivity shock. There are two terms in this equation that are
unobservable to the econometrician, �
and w
. w
represents shocks that are not
observable by firms before making their input decisions. On the contrary, �
represents
shocks that are potentially expected by firms when they make input decisions. w
can
also represent measurement error in the output variable. We will refer to �
as the
‘productivity shock’ of firm i in period t (Ackerberg et al., 2005). It is assumed that �
follows a first order Markov process and capital is accumulated by means of a
deterministic dynamic investment process:
!�
‚+
™ c
= !�
‚+
™�
,
where c
is firm i’s information set at t. Current and past realizations of �, i.e. (�
, ...,
148
�
) are assumed to be a part of c
. OP assumes that labor is a non-dynamic input. This
investment adds to future capital stock deterministically:
= Ÿ
`+
, $
`+
.
In view of the fact that
is decided at t −1, the above assumptions entail that it must be
uncorrelated with the unexpected innovation in �
between t −1 and t. This orthogonality
will be used to form a moment to spot
. Unlike capital,
is decided at t and,
consequently, correlated with the innovation component of �
. Considering the firm’s
dynamic decision of investment level, $
, OP state conditions under which a firm’s
optimal investment level is a strictly increasing function of their current productivity, �
,
i.e.
$
=
�
,
. B. 2
Profit maximization generates an investment demand function that is determined by two
state variables, capital and productivity. The reason f is indexed by t is the assumption
that variables such as input prices, are allowed to vary across time but not across firms
(Ackerberg et al., 2005). If the investment demand function is monotonically increasing
in productivity, it is feasible to invert the investment function and get an expression for
productivity as a function of capital and investment (Pakes, 1994):
�
=
`+
$
,
. B. 3
149
The heart of OP is to make use of this inverse function to control for �
in the
production function:
=
+
+
`+
$
,
+ w
. B. 4
The first stage of OP is to estimate this equation. f is the solution to a complex dynamic
programming problem. To avoid the computationally demanding assumptions, OP treats
`+
non-parametrically (Ackerberg et al., 2005). Given this non-parametric treatment,
direct estimation of (B.4) does not identify
, as
is collinear with the non-parametric
function. Nevertheless, one does find an estimate of the labor coefficient
, and of the
composite term
+
`+
($
,
), which we denote by Φ
¤
. By the timing
assumptions regarding capital, we can write:
�
= ¥[�
™c
`+
] + ¦
= ¥[�
™�
`+
] + ¦
,
where ¦
is orthogonal to
, i.e. ¥[¦
™
] = 0. This is the moment which OP uses to
identify the capital coefficient. To operationalize this process by GMM, given a guess at
the capital coefficient
, one can ‘invert’ out the �
’s in all periods:
�
= Φ
¤
−
.
Given these �
’s, one can compute ¦
’s in all periods by non-parametrically
regressing �
’s on �
`+
’s and taking the residual, i.e.
¦
= �
− ψ
¤
�
`+
,
where ψ
¤
�
`+
are predicted values from the non-parametric regression. Treating
the regression of �
on �
`+
non-parametrically allows for �
to follow an arbitrary
first-order Markov process. These ¦
’s can subsequently be used to establish:
150
1
ƒ
1
¨
) ) ¦
∙
In a GMM procedure,
would be estimated by setting this empirical analogue as close
as possible to zero (Ackerberg et al., 2005). LP adopts a similar approach to solving the
endogeneity problem. Instead of using an investment demand equation, they use an
intermediate input demand function to invert out �
. In the real data, investment is often
lumpy. This may not be in line with the strict monotonicity assumption regarding
investment. Also, OP procedure can cause efficiency loss in a data with zero investment.
Given that the intermediate input demands normally exhibit a lesser tendency to have
zeros, the strict monotonicity condition is expected to hold in the LP methodology. LP
considers the following production function:
=
+
+
+ �
+ w
,
where
is an intermediate input, such as electricity. LP considers the following
intermediate input demand function:
=
�
,
. B. 5
First, the intermediate input at t is chosen as a function of �
. Secondly,
is also taken
to be a ‘perfectly variable’ input. If
was chosen before
, then it would influence the
firm’s optimal choice of
. Under the assumption that intermediate input demand (B.5)
is monotonic in �
, we can invert:
�
=
`+
,
. B. 6
151
And substitute this in the production function to get:
=
+
+
+
`+
,
+ w
. B. 7
The first step of the LP estimates
using the above equation, treating
`+
non-
parametrically. Once more,
and
are not identified as
and
are collinear with
the non-parametric term. One also obtains an estimate of the composite term, in this case
+
+
`+
(
,
). In the second stage, there is one more parameter to
estimate,
. LP uses the same moment condition as OP to identify the capital coefficient
(Ackerberg et al., 2005). ¦
,
can be constructed as the residual from a non-
parametric regression of �
,
= Φ
¤
−
−
on 6�
`+
,
=
Φ
¤
`+
−
`+
−
`+
. They include a further moment to identify
, i.e. the
condition that ¦
,
is orthogonal to
`+
. This results in the following moment
condition:
¥[¦
,
™
`+
] = 0
¦
is not orthogonal to
because �
is observed at the time that
is chosen, and ¦
should be uncorrelated with
`+
, as
`+
was chosen at − 1 (Ackerberg et al.,
2005).
152
Appendix C
Appendix to Chapter 4
From the first order conditions of type 2 firm, we get:
( =
-
, Œ
?
@
’
’
B
+`t
× !N
+
u
1 − ¬
[
-
, Œ
-
t
-
+`t
]
`u
, C.1
* =
-
, Œ
1 − ?
@
’
’
B
t
× !N
+
u
1 − ¬
[
-
, Œ
-
t
-
+`t
]
`u
, C.2
1 = !¬N
+
u`+
[
-
, Œ
-
t
-
+`t
]
+`u
, C.3
or if the firm buys imported raw materials:
1 + = !¬N
+
u`+
[
-
, Œ
-
t
-
+`t
]
+`u
. C.4
From the first order conditions of type 3 firm, we obtain:
( = !
Q
, Œ
3 @
�
�
B
+`v
, C.5
* = !
Q
, Œ
1 − 3
@
�
�
B
v
. C.6
We can find an expression for relative productivity,
F
�
�,‘
F
’
�,‘
, by using Eqs. (C.1) – (C.6):
* = !
Q
, Œ
1 − 3
@
�
�
B
v
=
-
, Œ
1 − ?
@
’
’
B
t
× !N
+
u
1 − ¬
[
-
, Œ
-
t
-
+`t
]
`u
,
153
( = !
Q
, Œ
3 ®
Q
Q
¯
+`v
=
-
, Œ
?
®
-
-
¯
+`t
× !N
+
u
1 − ¬
[
-
, Œ
-
t
-
+`t
]
`u
,
1 + = !¬N
+
u`+
[
-
, Œ
-
t
-
+`t
]
+`u
.
Let � =
°
±
F
’
�,‘
’
²
’
±³²
be the ratio of the intermediate inputs purchased from type 1 firm
to the other inputs used by type 2 firm. Then the relative productivity can be written as:
Q
, Œ
-
, Œ
=
1 − ¬
1 − ?
�
u
1 − 3
∙
Using the above results, we get:
� = �
u
× �
+`u
=
Q
, Œ
-
, Œ
×
1 − 3
1 − ¬
1 − ?
×
!¬
1 +
,
Q
, Œ
-
, Œ
=
�1 +
1 − ¬
1 − ?
!¬1 − 3
∙
In Equilibrium, � can be approximated by
u
+`u
, the respective factor shares, i.e.
� =
N
+
-
, Œ
-
t
-
+`t
≅
¬
1 − ¬
∙
Therefore,
Q
, Œ
-
, Œ
≅
1 +
1 − ?
!1 − 3
,
s
≡
Q
s
-
s
≅
1 +
1 − ?
!s
1 − 3
∙
Furthermore,
154
´ , Œ
´
= ®
1 − ?
1 − 3
¯ µ
!s
− 1 +
´�s
´
!
-
s
¶ = ®
1 − ?
1 − 3
¯ ·
1
!s
−
1 +
!
-
s
∙
´�s
´
¸,
´ , Œ
´
¹
º
»
º
¼ > 0 $
´�s
´
<
!s
1 +
= 0 if
´�s
´
=
!s
1 +
< 0 $
´�s
´
>
!s
1 +
∙ C. 7
|
The change in relative productivity of type 3 firm to type 2 firm depends, among other
factors, on the change in the price of finished good as a result of trade liberalization.
Abstract (if available)
Abstract
It is generally believed that a rise in foreign competition makes the industrial sector more efficient. By using a novel firm-level data set from a variety of industries in Pakistan, this dissertation revisits the productivity-liberalization link, and investigates the effect of trade liberalization on firm productivity. There is evidence of an increase in competition following trade liberalization in the literature. In a majority of industries, there is reduction in the returns to scale, indicating the existence of inflexible capacity constraints in these industries. Moreover, there is no strong evidence of an improvement in productivity after trade reforms were introduced in the manufacturing sector of Pakistan. ❧ A greater part of this dissertation focuses on the textile industry of Pakistan, and unlike most other studies in the literature which mainly investigate the effect of trade liberalization reforms in developing countries, this dissertation investigates liberalization episode in a developed country and its consequence for firms in a developing country. Furthermore, it highlights sectoral heterogeneity within the manufacturing industry in the effect of trade reforms. In particular, using a sample of 321 textile and clothing companies for the years 1992 to 2010, we analyze the effect of quota phase-outs in the form of the end of Multi-Fibre Arrangement (MFA) on firm-level efficiency. The results differ for the two industries: MFA expiration lead to an increase in the average productivity of textile producing firms but a significant reduction in the mean productivity of clothing producers. We offer several possible explanations for this outcome, such as, a change in input and product mix, entry by non-exporters in clothing sector, and sectoral differences in quality ladders. Within the clothing industry, we compare the productivity of vertically integrated and non-integrated firms to investigate whether or not efficiency gains associated with a given liberalization episode vary across firms depending on their organization. The interaction between trade policies and firm characteristics is a subject of great interest to trade economists at present. Vertical integration is a firm characteristic that has the potential to affect the impact of trade policy on firms. Nevertheless, there are currently relatively few studies on this topic. Therefore, this dissertation addresses a potentially critical missing piece in our understanding of the impact of trade on firms. ❧ A theoretical framework in relation to vertical integration in the clothing industry shows that liberalization causes a change in the relative factor cost of the two types of firms, and consequently, a change in the product range produced by them. One of the difficulties in comprehending the connection between vertical integration and trade is a lack of data. The most innovative facet of this dissertation is to present a data set that includes information on the level of integration within firms, and to merge these data with a natural experiment resulting from the end of the MFA, in order to explore the differential impact of trade liberalization on vertically integrated versus non-integrated firms. This appears to be a fruitful way in which to deal with this subject. The empirical findings illustrate that a higher trade quota, which represents fewer trade barriers, reduces the mean productivity for all clothing firms significantly, though less so for vertically integrated firms than for non-integrated firms. The greater decline in the efficiency of non-integrated clothing firms points to the inability of these firms to benefit from tighter quality control, timely revision of production policies, and greater stability of supplies.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data
PDF
Financial crises and trade policy in developing countries
PDF
The murky risk of trade protectionism in an interconnected and uncertain global economy: a state, industrial, and regional analysis
PDF
Essays in panel data analysis
PDF
Empirical essays on trade liberalization and export diversification
PDF
Essays on business cycle volatility and global trade
PDF
The impact of economic shocks on firm behavior: insights from three studies
PDF
Essays on health and aging with focus on the spillover of human capital
PDF
The political economy of Brazilian trade policy: domestic determinants, world and regional strategies
PDF
Developing countries in the digital era: state and business interactions for industrial upgrading
PDF
Essays on effects of reputation and judicial transparency
PDF
Essays in macroeconomics
PDF
The political economy of implementation: intellectual property rights protection across the world
PDF
Essays on the estimation and inference of heterogeneous treatment effects
PDF
Trade, training, employment, and wages: evidence from the U.S. manufacturing industry
PDF
Growth, trade and structural change in low income industrializing economies
PDF
Political determinants and economic effects of HIV/AIDS: a push for the multisectoral approach
PDF
Two essays in econometrics: large N T properties of IV, GMM, MLE and least square model selection/averaging
PDF
Empirical essays on industrial organization
PDF
Essays on the politics of the American welfare state
Asset Metadata
Creator
Liaqat, Zara
(author)
Core Title
The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Economics
Publication Date
09/18/2012
Defense Date
09/04/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
multi-fibre arrangement,OAI-PMH Harvest,Pakistan ❧,productivity,trade liberalization
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Nugent, Jeffrey B. (
committee chair
), Hsiao, Cheng (
committee member
), Wise, Carol (
committee member
)
Creator Email
liaqat@usc.edu,zara.25@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-97264
Unique identifier
UC11289386
Identifier
usctheses-c3-97264 (legacy record id)
Legacy Identifier
etd-LiaqatZara-1203.pdf
Dmrecord
97264
Document Type
Dissertation
Rights
Liaqat, Zara
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
multi-fibre arrangement
Pakistan ❧
productivity
trade liberalization