International Competition and Industrial Evolution: Evidence
from the Impact of Chinese Competition on Mexican
Maquiladoras∗
Hâle Utar†
Luis Bernardo Torres Ruiz‡
This version: December, 2010
Abstract
We analyze effects of South-South competition at the micro-level by employing a new plantlevel panel data set that covers the universe of Mexican export assembly plants (maquiladoras)
from 1990 to 2006. By focusing on the competition in the third, North, market and using
the WTO accession of China as a quasi-natural experiment, our difference in difference approach
reveals a significant effect of intensified Chinese competition on maquiladoras. In particular, competition from China has negative and significant impact on employment growth both through
the intensive and the extensive margin, on the most unskilled labor intensive sectors of those
threatened by competition from China. We do not find a major effect on productivity through
reallocative channels within industries, but quantify significant within plant productivity improvement of maquiladoras attributable to competition from China. The results support field
studies and anecdotal evidence on industrial upgrading in Mexican Maquiladoras in response to
competition with low-wage locations such as China. They thus provide evidence that intensified
global competition can help export processing zones (EPZs) transform from low value-added
assembly plants to high-productivity, high value-added facilities.
JEL Classification: F14; L25; L60
∗
The authors would like to thank Gerardo Durand Alcántara, Director of International Trade Statistics, Administrative Registry and
Prices for access to the confidential data as well as for helpful information on the data-set. We also thank Thibault Fally, Wolfgang Keller,
Hiroshi Mukunoki, seminar participants at the University of Aarhus, ETH-Zurich, and Colorado State University as well as IIOC 2010
and CAED 2010 conference participants for helpful comments.
†
Corresponding Author. Department of Economics, University of Colorado at Boulder, Tel:+1 303 4927869, haleutar@gmail.com
‡
University of Colorado and Banco de México
1
Introduction
Many developing countries with cheap and abundant labor establish export processing zones (EPZs)1
to help increase export, create jobs, and generate foreign exchange. More importantly developing
countries hope that the zones, rather than being a short-run solution to unemployment and a trap
of low-value added facilities with significant vulnerability to shocks, will evolve towards higher valueadded production facilities and eventually gain from foreign direct investment through the transfer of
technologies and skill.
According to the International Labour Office (ILO) (2003) report, the number of countries with EPZs
grew from 25 in 1975 to 116 in 2002, employing 43 million people, of which 30 million are employed in
China. For many countries exports from EPZs account for a sizeable portion of their export earnings,
especially so for Mexico and China.
As global competition for jobs and foreign investment intensifies, it is an important question whether
Maquiladora industries in Mexico, one of the most mature EPZs, can survive intensified competition
with China by increasing productivity and possibly moving to higher value-added processes thereby
strengthening the Mexican host economy.
China’s size, rapid economic growth and trade performance is being felt across the globe. Especially so
in Mexico which has been a main competitor of China in the United States markets for manufactured
products. This competition saw a major shift in favor of China with China’s 2001 accession to the
World Trade Organization (WTO). By 2003 China had surpassed Mexico as the second most important
import supplier to the Unites States, behind Canada. China’s accelerated trade growth due to lower
trade costs in the wake of WTO accession provides us with a natural experiment to analyze the impact
of international competition in general.2 Similarity in export baskets between Chinese and Mexican
1
EPZs as defined by the International Labour Office (ILO) (2003) report are ”industrial zones, with special
incentives to attract foreign investment, in which imported materials undergo some degree of processing before
being exported again.”
2
In connection with entry into the WTO, China granted national status to foreign investors, implemented
Trade-Related Investment Measures (TRIMs) and Trade-Related Aspects of Intellectual Property Rights
(TRIPs), eliminated significant tariff and non-tariff barriers, restricted subsidies and eliminated requirements
based on company performance [39]. So the WTO membership of China can be perceived as lowering trade
costs not only by granting China permanent most favored nation (MFN) status, but also, perhaps more
importantly, by signaling commitment in the international trade and investment arena.
1
manufacturers to the US market makes the competition between Mexico and China even more intense,
and the analysis more revealing.3
Maquiladoras are export assembly plants historically specialized in labor-intensive products such as
apparel, footwear, electronics and toys. Since 1965, long before The North American Free Trade Agreement (NAFTA), favorable duty regulations with the United States have been in place for maquiladoras.
Since then, close proximity to the US market and relatively cheap labor made Mexico one of the most
favorable offshore destination for US companies for a long time. In 2006 the Maquiladora industry in
Mexico generated more than 25 billion dollars in foreign exchange, and accounted for 44 percent of
total Mexican manufacturing exports. 94 percent of the Maquiladora export in that year went to the
US.4 Together with enormous growth potential the sector also faces a significant hazard of shrinking
due to its sensitivity to decisions of US firms to source from elsewhere.5
The impact of China’s trade on both developing and developed economies is an important policy
question and has recently also received academic attention. Hanson and Robertson (2008) estimate
the impact of increase in manufacturing export from China on the demand for export from 10 other
developing countries, including Mexico, covering the period between 1995 and 2005. Based on gravity
equation estimates they conclude that the impact is small. On the other hand, firm-level studies
quantify significant impact of Chinese penetration in the developed markets. Bloom et al. (2009) use
a panel of establishments from European countries to test the impact of Chinese imports on the use
of Information Technology equipment and innovation, finding a positive association between the two.
Bugamelli et al. (2010) test the pro-competitive effect of Chinese imports on Italian firms, finding
a significant effect. These studies point out a heterogenous impact of Chinese competition within
industries which leads to reallocation. This could be because firms with relatively less sophisticated
technologies are more exposed to competition from China and probably produce products that are
3
In 2001, the export structure of Mexico is found to be the most similar to China’s export among any other
Latin American countries (Devlin, Estevadeordal and Rodriguez-Clare (2006). The same conclusion is also
derived from Lall and Weiss (2004) and Gallagher and Porzecanski (2007) among others.)
4
Authors’ calculation using the data from Trade Statistics Specialized Technical Committee, formed by
Banco de Mexico, Instituto Nacional Estadı́stica y Geografı́a (INEGI), Servicio de Administración Tributaria
(SAT) and the Secretaria de Economia.
5
Hanson (2002) stresses that the Maquiladora sector with its impressive growth rates experienced in the
1990s (real value added grew at an annual growth rate of 10 %) and its role in Mexico’s export boom is an
important opportunity for Mexico’s economic development. But it also poses a challenge as it is characterized
by footloose industry.
2
more substitutable with Chinese products. In this paper, we analyze the competition in the developed
country with Chinese products for the plants that, in the light of these findings, can be expected to
be impacted the most: relatively labor-intensive, downstream offshore plants.6
Among the most notable empirical studies that examine the relationship between competition and
productivity, Schmitz (2005) identifies the entry of Brazilian iron ore producers to the U.S. market as
the underlying reason for the large increase in productivity of U.S. iron ore manufacturers in the 1980s
providing an evidence for a positive relationship. On the other hand, Aghion, et al. (2005) using data
from the UK, point out a possible non-linear relationship between competition and productivity.7
We investigate the impact of the competition from China on Mexican export assembly plants (maquiladoras); on plants’ growth, entry, exit and productivity using the data from the plant-level survey that
covers the universe of Mexican maquiladoras. The data we use cover the years 1990-2006, a time period
long enough that it allows for proper identification of the effects, if any, of Chinese competition. Our
sample starts in 1990 where China’s share in manufacturing trade in the World was 1.74 % and covers
until 2006 where China’s share became 8.37 % (World Bank). In contrast to previous studies we are
better able to isolate the competition effect, as we focus entirely on export assembly plants in Mexico
that are tied to the US manufacturing sector. We expect competition between Mexican Maquila and
Chinese plants in the US market, and the competition is perhaps more direct because Mexican and
Chinese plants have very similar export baskets as identified by several studies [13][18][19][32]. In
addition, we examine the link between international competition and productivity using the WTO
accession of China as a quasi-natural experiment which allows us to identify the causal impact of intensified competition on the productivity of Mexican plants. This paper also provides a first analysis of
any aspect of Mexican Maquiladoras using comprehensive plant-level panel data and thus contributes
a deeper understanding of this important offshore industry and it sheds light into the future of export
processing zones as global competition intensifies.
Concern is not entirely unjustified that intensified Chinese competition will destroy Maquiladoras as
6
Studies that analyze the impact of intensified competition from China on Mexican Maquiladoras, including
Mollick and Wvalle-Vazquez (2006), and Mendoza (2010) use aggregate data on employment and wages, and
identify Chinese competition as an important factor contributing to Maquiladora slow-down. There are also
recent studies that analyze the impact of Chinese competition on Mexican manufacturing plants which mostly
sell to domestic market, see for example, Iacovone et al. (2010).
7
For an excellent review of the recent literature on competition and productivity see also Holmes and
Schmitz (2010)
3
they are footloose establishments that can easily be relocated. There is much anecdotal evidence
suggesting that maquiladora plants move operation from Mexico to China: Royal Philips Electronics
moved its operation from Cuidad Juarez to China in 2002. Amases de Juarez moved its operation to
China in 2002. Sanyo Electric Co. closed two of its six Tijuana plants in 2001 and moved to China and
Indonesia [10]. We find evidence partially supporting the frequently stated view by Mexican policy
makers that Chinese competition is forcing maquiladoras to exit low-tech, labor intensive industries
and evolve toward higher value added, technology intensive sectors. At the intensive margin, we
find that employment growth is negatively affected by Chinese competition. More specifically, a one
standard deviation increase in China’s share of the import penetration rate in the U.S. is found to
be associated with a decrease in annual employment growth of 8.7 percentage points. We find that
Chinese competition causes plant exits in more unskilled labor intensive sectors among sectors that
are under direct threat of Chinese competition. We also find that the number of entrants decreases
with intensified Chinese competition. But we do not find that intensified competition from China
improves the productivity of maquiladoras by causing exit of especially low-productivity plants or by
reallocating market shares towards high productivity plants within industries. Instead, we find strong
evidence for significant within plant productivity improvement of maquiladoras due to heightened
competition from China. Auxiliary data supports this finding by showing substantial increases in
the implementation of productivity enhancing management practices and capacity utilization among
maquila plants in the period spanning the WTO accession of China.
Our results lend support to a commonly held view among Maquiladora managers: ”By moving up the
technological ladder, companies say they can afford to pay the relatively high salaries common along the
Mexican border and not relocate to lower-wage countries.” (Lindquist (2004)) Both China and Mexico
liberalized their economies since the 1980s and hope to gain through increasing openness. Although
trade growth was impressive in both countries in the last decades, China’s trade growth was also
fueled by productivity-based economic growth whereas Mexico experienced relatively un-impressive
economic growth performance. Despite official Mexican concerns regarding China’s accession to WTO
[10], our work highlights that long expected productivity growth in Mexico due to export and FDI
may have just begun, ironically, triggered by competition from China.
In the next section we describe the environment of maquiladora industry and the data used. Motivational thoughts are presented in section 3. In section 4 our empirical model is outlined, and results
are interpreted in section 5 followed by robustness checks in section 6. We conclude in section 7.
4
2
Data Overview
2.1
Mexican Maquiladoras
A typical maquiladora plant imports inputs mostly from the United States, processes them, and
then ships them back.8 The maquiladora program started in 1965 with the purpose of reducing
unemployment in the border region; it permits tariff-free transaction of the inputs and the machinery
between ’a maquiladora plant’ and the foreign companies and it also allows 100 % foreign ownership.
Upon the return of the goods, the shipper pays duties only on the value added by manufacture
in Mexico (Gruben (2001)). In order to benefit from the maquiladora program, a plant has to be
registered as a maquiladora plant.9 In general, there are three ways in which a maquiladora can
operate: subcontracting, shelter operation and direct ownership. The subcontracting operation offers
the least amount of control to the foreign firms, since the subcontractor fulfills all of the manufacturing
operations according to an arrangement established with the foreign firm. Shelter operations offer
more control, especially in the production process, but not in the administrative operation of the
maquiladora plant, i.e. legal, accounting, customs, etc. Direct ownership offers the foreign firm the
most control and supervision over manufacturing operations, and is the most common way of maquila
operation. Since its introduction, the maquiladora industry moved from consisting of only low-skilled
labor intensive plants focusing on simple assembly jobs to more advanced manufacturing processes,
like machinery and automotive.10 The government allowed the establishment of maquiladoras in
the interior regions of Mexico. NAFTA also contributed to maquiladoras being allowed to sell their
output domestically, but this option is rarely exercised. The implementation of NAFTA required
Mexico to change certain provisions for the maquiladora industry, such as eliminating certain tariff
benefits. Most importantly, on January of 2001, duty-free imports from non-NAFTA countries were
eliminated because these countries intended to subsequently re-export to another NAFTA country.
These changes were based on the rules of origin that were established under the treaty, where goods
traded between NAFTA countries are allowed duty free treatment only when the goods satisfy a
8
Export Processing Zones (EPZs), similar to the maquila program of Mexico, can be found around the
world, e.g. China, the Philippines, Malaysia, Hungary, Pakistan, Costa Rica, Honduras, among others.
9
The bureaucratic steps necessary for registration were simplified significantly with the 1983 reform.
10
In 1969, 147 companies and 17,000 workers were registered under the Border Industrialization Program.
Among the first companies were RCA (electronics), Convertors (industrial tapes), Sylvania (electronics), Acapulco Fashion (apparel) and Ampex (electronics).
5
minimum percentage of North American content. Due to complaints from leaders of the maquiladora
industry, the Mexican government revised its regulations of the maquiladora sectors and created a
sectoral promotion program to protect the duty-free status of maquiladora imports and therefore,
allowing the maquiladora program to continue to use non-NAFTA content imports.11 Even after 2001
there is no incentive for a foreign company not to register as being a maquiladora if it is part of a
foreign chain of production re-exporting its goods to the US. This is due to the tax provision (APA)
that allows maquiladoras not to pay income taxes in the same way as the domestic manufacturing
industry (Truett and Truett (2007), Canas and Coronado (2002)).
2.2
Plant-level Data
The maquiladora industry data is from Instituto Nacional Estadstica y Geografa (INEGI). INEGI has
conducted a monthly survey of the universe of plants registered under the maquiladora program until
2007, called the Estadstica de la Industria Maquiladora de Exportacion (EIME).
12
The observation
unit for the industry is a maquiladora establishment, or plant. The data contains firm id’s as well
as plant id’s so that it is possible to identify multi-plant and single-plant firms. INEGI constructed
an annual data set from the monthly surveys, and it is the data set used in the present study. The
annual panel data set covers the period between 1990 and 2006 for eleven manufacturing maquiladora
industries. The majority of the plants are owned by US companies. We can not identify the owner
at the plant level due to confidentiality issues, but we do have aggregate capital investment data
in maquiladora industry which can be used as an ownership proxy. Figure 1 shows the evolution
of capital investment in maquiladora by selective country of ownerships. In 1994, the US share of
capital equipment investment was 92.4 %. The next biggest investor was Japan, with a share of 2.5
%. In 2006 the US share was 88.1 % followed by Canada and Switzerland both having 1.4 % shares
(Source: Banco de Mexico). In terms of sales, maquiladoras’ export to the US was 99.7 % of the total
maquiladora export in 1993. In 2006 94 % of the total maquiladora sales was to the US followed by
11
The sectoral promotion program that allow for each maquiladora sector to continue to have the tariff-free
entry of non-NAFTA components has been established in December 2000.
12
In 2007 a regulatory change was enacted that merges the maquiladora program with an export oriented
program for domestic companies known as the Program for Temporary Imports to Promote Exports (PITEX).
The new program is called Maquiladora Manufacturing Industry and Export Services (IMMEX). As a result,
INEGI stopped reporting maquiladora data after March 2007 and the data has been merged in to the IMMEX
data.
6
Canada with a share of 1.7 % (Source: INEGI). INEGI dropped establishments which did not answer
the questionnaire or did not report output measures from the data set.13 Thus, the final data set
consist of 27,548 plant year observations that consist of 3,769 plants and 1,455 firms (1655 plants on
average per year). In Figure 2, we present the comparison of the nation-wide maquiladora export
value added from INEGI with the aggregate export value added as reported by the plants, we have in
the data-set. Figure 2(a) shows export value added in thousand USD and 2(b) shows growth rates. As
suggested by the graphs, the plant-level data is very comprehensive and closely follows the aggregate
trend.
We have plant-level information for the 9 states: Baja California, Coahuila, Chihuahua, Distrito Federal, Jalisco, Estado de México, Nuevo León, Sonora, and Tamaulipas. Among them Baja California,
Coahuila, Chihuahua, and Tamaulipas are the states where the maquiladora concentration is the highest. For each plant we have information on hours worked and the number of employees by job category,
wages paid by job category as well as plant expenditures/inputs, export sales, and value-added. Plants
also report rental expenditures on different capital items, namely machinery, equipment, buildings and
office space. Our correspondence with INEGI reveals that rental equipment which includes precision
and resistance instruments, rotation bands, forklifts, trucks with special containers (temperature and
toxic waste) etc.. is mostly rented domestically, although the survey does not collect information on
owned imported capital equipment. We proxy capital using rental capital expenditures. Correlation
between rental capital expenditure at the state-level and maquiladora FDI data at the state-level for
the period 1994-2006 (provided by the Secretaria de Economia) is 0.923.14 All nominal values are expressed in thousand 2002 Mexican peso. See Table A-1 for the descriptive statistics. We use separate
industry deflators (industry classification for deflators approximately corresponds to 3-digit SICs) for
each maquiladora sector to deflate revenues and material expenditures. We use energy deflators to
deflate fuel and electricity; a machinery rental deflator to deflate the rental expenditures in machinery
and equipment and a building rental deflator to deflate the building rental expenditures. The deflators are provided by Banco de México. Average expenditure shares of labor, rental capital, materials
and energy are 27.3, 6.5, 63.7 and 2.7 percents respectively. Reflecting the downstream position of
maquiladoras within industries, the share of imported materials constitute by far the most important
13
Every plant operating under the maquiladora program was legally required to answer the questionnaire.
Our data set reveals that plants which did not answer the questionnaire (although legally required) are mostly
located in the interior regions of Mexico where the maquiladora concentration is very small. Further characterization of non-responsive and removed plants is being pursued in correspondence with INEGI.
14
State-level maquiladora FDI data is the most disaggregated FDI data we found.
7
expenditure item. In the data-set we have 11 manufacturing sectors, which we match with the corresponding US industries in order to construct our aggregate variables. Table B-1 presents these 11
industries and corresponding 1997 NAICS codes. The details of the aggregate data construction is
given in the appendix.
3
Motivation
Studies that analyze trade composition of the two main offshore destinations for the US manufacturing
sector, Mexico and China, find significant similarity between their imports in the US market.15 We
expect that China’s recent trade performance accompanied by its accession to WTO has direct and
strong effect on Mexican export assembly plants.
3.1
Comparative Advantage and Compositional Change
Both China and Mexico have a comparative advantage in labor-intensive products compared to the
US. However, China has a comparative advantage in unskilled labor in comparison to Mexico. In 1999,
approximately 13 % of the Latin American population had post-secondary education, compared to 3
% in China (Devlin, Estevadeordal and Rodriguez-Clare (2006)). Factor content theory suggests that
as trade liberalizes in China, industries that disproportionately employ unskilled workers will shrink
in Mexico and the opposite will occur in China. This can happen through the intensive margin, that
export assembly plants operating in Mexico shrink. It can also happen through the extensive margin
that plant exits occur as a result of the competition and/or that heightened competition discourages
entry of new plants. It is also possible that as firms offshore more labor-intensive tasks to China in
response to a fall in trade costs with China, they may deploy to Mexico relatively more skill-intensive
parts or parts in which the relative proximity of an offshore plant is a more important determinant
than labor costs.
Recent survey results point out the heterogeneous structure of maquiladoras and identify three generations in maquiladora sectors. First generation maquilas are characterized by labor-intensive assembly,
second generation maquilas involve more skill-intensive processes such as production of television receivers while third generation maquilas have in-house designs and patent products. (Carillo and Lara
15
See for example Lall and Weiss (2004), Gallagher and Porzecanski (2007), Gallagher, et al. (2008), Devlin,
Estevadeordal and Rodriguez-Clare (2006)
8
(2005)) Even in the traditionally labor-intensive sector, apparel, Bair and Gereffi (2003) identify recent
industrial upgrading from purely garment assembly to full production facilities with cutting rooms,
industrial laundries and finishing plants.
3.2
Competition and Productivity
Product market competition will lead Mexican plants to loose market share in the US market. Typical industrial organization theories with differentiated products (Dixit and Stiglitz (1977), Salop
(1977)) predict a negative relationship between competition and innovation/upgrading, since competition will decrease the rents of innovating/upgrading for innovators upon innovation.16 This is the
Schumpeterian effect that the incentive to innovate decreases as competition increases. However, the
innovation/upgrading decision is also affected by the difference between the pre-innovation and postinnovation rents (Aghion et al. (2005)). If the pre-innovation rent disproportionately decreases due
to intensified competition, then firms upgrade or innovate to be able to survive or ’escape’ from the
competition as much as possible. It is shown in Aghion et al. (2005) that such an escape competition
effect is stronger when the market structure is such that technological differences between firms are
small. Export assembly industries both in China and Mexico are mostly based on labor-intensive
technologies with no large technological gaps between plants, so one may expect to see stronger escape
competition effect on plants’ incentive to upgrade their technologies.17
Another possible channel that can strengthen the escape competition effect is through a parentsubsidiary relationship. Consider two competing offshore destinations. In response to lower trade
costs in one of the offshoring destinations, a parent with a subsidiary in another location would make
a ’credible’ threat of relocating the subsidiary and therefore increases the incentive for the manager of
the subsidiary to put more effort and decrease X-inefficiencies.18 Schmidt (1997) shows in a principal16
Arrow (1962), on the other hand shows that the incentive to do cost-reducing innovation is higher for a
perfectly competitive firm than for a monopolist in the homogeneous product markets under certain assumptions.
17
Indeed, we find the dispersion is quite low among Maquiladora plants as measured by interquartile measure
which is around 0.20 for the overall industry; it is lower for example in comparison to Taiwanese plants as
reported by Aw, Chen and Roberts (1997).
18
Principal-agent problems are especially relevant to our context as we focus on the performances of subsidiaries. Papers analyzing the competition and within firm productivity from a principal-agent problem
perspective also include Hart (1983), Scharfstein (1988), and Hermalin (1992) among others. In Hart (1983)
9
agent setting that the threat of liquidation can decrease managerial slack and decrease X-inefficiencies.
In our context, the threat of liquidation can be thought of in terms of relocating the subsidiary (or, in
case of subcontracting, the threat of ending the contract and switching to a lower cost partner). Such
a threat should carry greater weight to maquiladoras, many of which are footloose industries.
In a field study conducted in Reynosa, Sargent and Matthews (2006) identify plant manager or
subsidiary-driven upgrading motives as an important source of technology upgrading in maquiladoras.
”Due to cost pressures, the parent began sourcing from China rather than Reynosa. [...] Given their
rapidly shrinking character, the plant manager first brought the president of the parent to Reynosa.
The plant manager sold this individual on the idea that while it didn’t make sense to continue with
their existing product line, they had a great management team and perhaps the Reynosa plant could
produce for other divisions.....This effort has been successful; six production lines were sent to Reynosa
and they successfully filled up the plant”[41]
We now turn to the empirical model.
4
Empirical Model
Our identification strategy is based on the fact that some of the maquiladora sectors are not affected by
Chinese accession to WTO as much as sectors with a strong Chinese comparative advantage. Across
sector variation in the degree of Chinese competition can be due to structural reasons such as the
transportation costs, or relative skill-intensity of the production processes. The two maquiladora sectors with the lowest share of skilled employees (administrators and technicians) over total employment
are Apparel and Toys with respective sample average ratios 0.16 and 0.17. The sectors with the highest share of skilled employee are Food and Transportation Equipment with respective sample average
ratios 0.25 and 0.24. Various reasons for the variation in the Chinese comparative advantage will be
reflected in the Chinese share of the import-penetration rate. We construct a measure of Chinese
competition for Maquiladoras as the Chinese share of the import penetration for the matched US
and Scharfstein (1988), competition affects the informational structure and changes the possibilities that principal can make inferences about the manager’s action. In Hermalin (1992) competition changes the manager’s
incentive through the income effect.
10
industry, following Bernard, Jensen and Schott (2006).
IM P CHjt =
19
CH
Mjt
Mjt + Qjt − Xjt
(1)
CH
where Mjt
denotes the value of imports of industry j products coming from China to the US at period
t. M , Q and X denote total US imports, US production and US exports respectively. We classify two
main groups based on the median level of the Chinese share of import penetration in the US market,
before China’s WTO accession, in 2000 (4.46 %). We call the above median group HighCHT , where
we expect a high degree of Chinese threat. This group consists of the following sectors: Apparel
(Ecogroup 2), Footwear and Leather (Ecogroup 3), Electronic and Electrical Machinery and Equipment
(Ecogroups 8 and 9), and Toys and Sporting goods (Ecogroup 10). We further decompose the rest of
the sectors by two groups according to the median of the rest of the group, which is 0.97 %. N oCHT
where we expect minimum Chinese presence and threat are sectors with less than 0.97 % Chinese share
of the import-penetration rate in 2000. These are Chemicals, Transportation (Auto Parts) and Food
products.20 Our third group, which is an excluded group in our regressions, consists of Furniture and
Wood products, Metal products and non-electrical machinery parts, and Miscellaneous manufacturing
(Ecogroups 4-7-11).
Although we base our classification on the import-penetration rate, sectors with tiny presence of
Chinese imports are also reflecting the sectors in which Mexico has a comparative advantage due to
transportation costs, relative skill-intensity, and relative capital-intensity.
Across-time variation in the identification strategy comes from China’s accession to WTO, represented
by a dummy variable, I(W T O) that takes 1 after China’s accession to WTO, i.e.
I(W T O) = 1
if YEAR > 2001
=0
otherwise
.
After the WTO accession, the average annual increase in the Chinese share of import penetration rate
across corresponding US sectors was 6 times higher than it was from 1990 to 1991 or 2.6 times higher
than the year before which is a significant jump.
19
An alternative would be the ratio of total imports coming from China to the relevant US industry to total
imports in the US industry as used in Bloom et al. (2009). We use both of them. The results are qualitatively
the same.
20
We also set our threshold level according to 1999 import-penetration levels. Our classification does not
change.
11
Below we present the WTO dummy approach.
4.1
Regression Equations
We constructed a difference in difference approach to investigate the impact of Chinese competition on
the Mexican maquiladoras. We present our approach below using plant tfp as our variable of interest,
although we apply it also to employment growth, plant-exit and plant entry with minor changes.
4.1.1
WTO Dummy
lnT F Pijst = µ0 + µ1 I(HighCHT )j ∗ I(W T O)t + µ3 Xijst +
X
δjI Industryj
j
+
X
YS
δts
Y
(2)
eart ∗ States + ijst
ts
I(HighCHT) is an indicator variable that takes 1 if the plant i at period t belongs to the respective
groups as defined above. Subscripts i, j, s, and t index plant, industry, state and year respectively.
Xijst is a vector of plant-level controls : multi-plant dummy, age dummies, entrant dummy (takes 1
if the plant enters in that period), and exit dummy (takes 1 if the plant does not participate the next
period).21
The plant-level total factor productivity measure is calculated separately for each industry allowing
differing technologies as described in the Appendix. We add interactive state-by-year fixed effects
to control for aggregate shocks that may affect productivity across all sectors but may vary across
different states for example due to local labor market conditions. We also control for industry fixed
factors such as differing technology that may affect productivity. We cluster the standard errors
by each industry in each year to account for correlation of shocks within each industry-year cell as
suggested by Moulton (1990).
In this specification we separate the variation in productivity due to the WTO accession of China from
other sources by exploiting not only the variation of productivity before and after the WTO accession
of China, but also across plants that are exposed to Chinese competition with differing degrees. Our
21
Since INEGI does not ask what year a plant was established, we constructed three age dummies according
to the number of years that plants have been in the sample since 1990 as follows: young (Age Dummy1: 1 − 4
years), mid-age (Age Dummy2: 5 − 9 years), and old (Age Dummy3: >= 10 years). Age Dummy 3 is excluded
from the regressions.
12
difference in difference estimates of the effect of Chinese competition are represented by µ1 , it indicates
the productivity differential for sectors with heavy presence of Chinese imports in the corresponding
US market compared to the rest of the maquiladora sectors. If Chinese competition makes plants more
productive, say through upgrading of production techniques, by cutting out slack and getting rid of
X-inefficiencies in management and organization, the coefficient µ1 should be positive. We further use
our second categorization as follows.
lnT F Pijst = ν0 + ν1 I(N oCHT )j ∗ I(W T O)t + ν2 Xijst +
X
δjI Industryj
j
+
X
YS
δts
Y
(3)
eart ∗ States + ijst
ts
lnT F Pijst = η0 + η1 I(HighCHT )j ∗ I(W T O)t + η2 I(N oCHT )j ∗ I(W T O)t
X
X
YS
+η3 Xijst +
δjI Industryj +
δts
Y eart ∗ States + ijst
(4)
ts
j
In equation 3, our difference in difference estimate is ν1 , which indicates the productivity differential
between the sectors which are not under the dominance of China compared to the rest of the the
sectors. Finally, in equation 4, we use both of the groups (HighCHT, and NoCHT) and estimate the
productivity differential of these groups in comparison to sectors which experience a moderate threat
from China. Our regression model identifies the impact of Chinese competition on within productivity,
as our dependent variable is un-weighted productivity. But aggregate productivity will also be affected
by reallocation at the extensive margin, that is, through entry and exit of plants. So we include entry
and exit dummies to capture these effects. Since competition could interact with entry and exit we
also include interaction between entry and exit dummies with the WTO dummy.
In our specifications in 2-4, we do not consider intensified competition from China as a gradual change.
Say, if the intensified competition helps to decrease X-inefficiencies and so leads to an increase in plant
productivity, the effect is likely to be gradual. On the other hand, if our WTO dummy captures,
for example, the US recession between 2001 and 2003 also assuming that somehow the recession
disproportionately affected industries that are the most receptive to the competition from China then
one might expect a temporary change in the dependent variable. One way to investigate year by year
change is to interact our group dummies with year dummies.
lnT F Pijst = µ0 +
X
µ1t Y eart ∗ I(HighCHT )j + µ2 Xijst +
t
X
δjI Industryj +
j
X
ts
13
YS
δts
Y
eart ∗ States + ijst
(5)
In this specification, µ1t , will give the productivity differential between the plants that are exposed to
high level of Chinese competition with others at each year t.
4.1.2
Import-Penetration Rate
Following the literature, we also use the Chinese share of import-penetration rates as our measure
of the Chinese competition [6] [8]. In explaining this approach, employment growth is used below as
the variable of interest although we apply it also to plant tfp, plant-exit and plant entry with minor
changes. Consider the following specification:
∆lnEijst = α0 + α1 Xijst + α2 Zjt + α3 IM P CHjt + α4 IM P CHjt ∗ xijst +
X
YS
δts
Y eart ∗ States + ui + ijst
(6)
ts
where ∆lnEijst = lnEijst+1 − lnEijst and Eijst refers to total employment as measured by head
count at plant i in industry j located in state s at year t. We allow for unobserved heterogeneity ui
which may be correlated with regressors and estimate equation 6 using OLS. Vector X includes time
varying plant-level controls that are found to be important in determining firms’ growth: these are
size dummies, plant tfp, a multi-plant dummy, and age dummies.22 Vector Z includes time varying
industry-wide controls; these are industry aggregate variables for the matched US industries that may
affect the demand for a particular maquiladora sector: import-penetration rate of the corresponding US
industry calculated without the imports from China, the matched US industry hourly wages relative
to the corresponding measure in the Maquiladora sector, and the production index of the matched US
industries to control for the sector specific business cycles.23 We then interact our Chinese competition
measure with several variables of interests xijst (productivity, skill-intensity, capital-intensity); to see
if trade between the US and China has a disproportionate effect on any particular type of exportassembly plants in Mexico.
Our industry level variables including the Chinese share of imports are variables for the third market,
that is for the US industries, not for Maquiladora industries, but there would still be an endogeneity
problem if unobserved factors that affect the variable of interest in maquiladoras also affect the Chinese
share of import penetration in the US industry. If, for example, Chinese products decrease market
shares in those sectors where Mexican maquiladoras experience productivity increase, then we may
22
We constructed five dummies for plant size, measured by the number of employees, for each of the ranges
1-50, 51-100, 101-500, 501-1000 and 1000+. We exclude the smallest size category from the regressions.
23
Details of these data are given in the appendix.
14
expect a downward bias in OLS estimates of productivity. If Chinese products increase market shares
in those sectors where Mexican maquiladoras shrink then we expect upward bias in OLS estimates of
employment growth. The difference in difference strategy described above gets around this potential
problem as it is based on exogenous change in the degree of competition due to China’s accession
to WTO. However, in order to see the robustness of our results to alternative specifications, we also
use import-penetration rates as a direct measure of intensified competition from China. To do that,
we instrumented the Chinese share of import penetration rate with the real exchange rate between
China and the US interacted with the 1999 Chinese share of import penetration of the corresponding
US NAICS for each Maquiladora sector. The real exchange rate between China and the US must be
exogenous from the perspective of Mexican plants. By interacting it with the cross-sectional shares
before China’s accession to the WTO, we get the cross-industry variation in the degree of Chinese
comparative advantage. Another instrument we use is the worldwide Chinese imports (exports from
China) as a share in total world imports interacted with the 1999 Chinese import shares over all imports
of the corresponding US NAICS for each Maquiladora sector. We also use the import penetration rate
without Chinese imports, defined below, as an industry-level control variable.
IM Pjt =
CH
Mjt − Mjt
Mjt + Qjt − Xjt
(7)
In order to instrument the import penetration rate calculated without Chinese imports, IM P , we use
the industry specific exchange rate for the US industry where the weights for each trading partner’s
currency are lagged share of imports of that particular trading partner. We also use MFN tariff average
of matched US industries and lagged values of import-penetration rates constructed without Mexican
and Chinese imports.24 We now turn to the results.
5
Results
5.1
Employment Growth
In Table 1 we present the WTO dummy estimation results for employment growth. In column 1,
the coefficients of the WTO dummy is negative and significant, indicating a significant decline in
employment growth by about 28 % in all maquiladora sectors after 2001. Obviously this decline is not
entirely attributable to intensified competition with China. Our difference in difference estimates of
24
The sources of these data are stated in the Appendix.
15
the effect of the intensified Chinese competition is measured by the coefficient of HighCHT ∗ W T O
which is found to be negative and significant at 10 % level (column 2). The coefficient indicates
that employment growth on average declined 8 % more after the WTO accession of China in those
plants which are threatened the most by Chinese competition. An alternative difference in difference
estimate is given by the coefficient of N oCHT ∗ W T O, which is found to be positive and significant.
The coefficient of N oCHT ∗ W T O in column 3 indicates that after the WTO accession of China,
employment growth is on average 8.4 % higher in plants that do not face significant competition
from China in comparison to the rest of the plants. In column 3, both of our interaction terms have
the right signs, that is, in comparison to plants that face moderate competition the plants that face
the toughest competition experienced higher decline in employment growth on average and plants
that face the least competition experienced a lower decline; but the differences are not statistically
significant. We then use stricter classification by defining industries under most threat as those in
the third quartile of the Chinese share of import penetration rate as opposed to those above median,
(High3qCHT : Footwear, Apparel and Toys), in column (4). The sectors defined by the third quartile
are also the most unskilled labor-intensive sectors among the ones that are under threat of Chinese
competition.25 We found the coefficient of the interaction term negative and significant at the 1 %
level indicating a drop in employment growth attributable to competition from China.
Our plant-level coefficients in all of our regressions are significant and they all have the expected signs.
Employment growth increases with productivity, decreases with age, and decreases with size.26 We
also find that on average employment growth is higher in plants that belong to an entity which owns
other maquiladora plants (multi-plant firms).
In Table 2 we present the estimation of our employment growth equation when using the Chinese
share of import penetration rate. In column 1, the coefficient of Chinese share of import penetration
in the US is found to be negative and significant at the 1 % level. The magnitude indicates that a
one standard deviation increase in the Chinese share of import penetration rate is associated with a
decrease in annual plant employment growth of 9.9 percentage points. The coefficient of the import
penetration rate calculated without Chinese imports, IM P is also found to be negative and significant
25
The sample averages of the ratio of skilled workers over unskilled workers for sectors classified under
HighCHT , namely Apparel, Footwear and Leather, Toys and Sporting Goods, Electronic devices and Electrical
equipment are 0.23, 0.28, 0.27, 0.36 and 0.36 respectively.
26
It is usual to find that younger and smaller firms and plants grow faster conditional on survival (Dunne
et al. (1989)). Jovanovic (1982) provides a theoretical foundation through learning.
16
at the 10 % level indicating that import competition in the US market in general is associated with
lower employment growth. In column 2 we also add Mexican industry hourly wages relative to the
US counterpart, as well as the US industry production index. The coefficient of the US production
index is positive and significant at the 5 % level, indicating a positive association between maquiladora
growth and the US production. The coefficient in front of the Chinese share of import penetration is
still significant at the 5 % level. Its magnitude indicates that a one standard deviation increase in the
Chinese share of import penetration rate is associated with a decrease in annual plant employment
growth of 8.7 percentage points.
In columns 3 and 4 of Table 2 we present instrumental variable regression results when we instrument
the Chinese imports variable, IM P CH, with the real exchange rate between China and the US
and the worldwide Chinese imports relative to the total world imports. Both are interacted with
the 1999 Chinese import shares. We also instrument, IM P , with the instruments described in the
previous section. The results confirm that Chinese imports in the US market are associated with lower
employment growth in Maquiladora industries.
In Table 3 in columns 1, 2 and 3 we present our results when we interact our Chinese competition
proxy with plant TFP, skill intensity as measured by the ratio of skilled workers to unskilled workers
and capital-labor ratio as measured by the rental expenditures of machinery, equipment and building
to total wages respectively. None of the interaction terms are significant, so there is no indication
that intensified Chinese competition as proxied with the Chinese share of the import penetration rate
in the US causes a disproportionate decrease in employment growth, especially in the group of lowproductivity plants, low-skill intensive plants or low capital-intensive plants within an industry. The
substitutability between the Chinese export bundle and the Maquiladora export bundle is quite high
and there is no apparent ranking between them. That is, we do not expect Chinese exports to the US
to exhibit higher substitutability with the lower end of the distribution of maquiladora products in
comparison to the upper end for a given industry. Although for example, as Bloom et al. (2009) and
Bugamelli et al. (2010) find, it is more plausible to think that imports from China to Europe compete
more with European firms’ products located at the low end of the distribution within industries.
17
5.2
Employment at the Extensive Margin
5.2.1
Entry of New Plants
In order to analyze the impact of Chinese competition on plant entry we aggregate the plant-level
data to industry-level and estimate the following equations:
EN T RYjt = γ0 + γ1 Zjt + γ2 IM P CHjt +
X
δtY Y eart +
t
X
δjI Industryj + jt
(8)
j
EN T RYjt is the total number of entrants in industry j at period t. We include industry dummies
to control for industry-specific factors that affect entry, such as different levels of sunk entry costs
associated with starting up a plant, say, in the apparel sector versus in auto parts. We also include
year dummies to control for aggregate shocks such as exchange rate that may affect the entry decision to
maquiladoras in the same way across sectors as it affects the relative production costs between Mexico
and the US. If intensified Chinese competition discourages entry of new export-assembly plants in
Mexico, we expect γ2 to be negative. Due to the count data nature of the dependent variable, we
estimated equation 8 using Poisson and negative binomial regressions. Our dependent variable exhibits
over-dispersion so we opted for the negative binomial model.
27
In column 1 of Table 4, we regress
EN T RYjt on the Chinese competition proxy, industry and year fixed effects. We find a negative and
significant effect of the Chinese share of import penetration on entry. Can this effect be generalized to
imports from everywhere else? Or is it especially true for Chinese competition? We add the import
penetration rate in column 2. We find a weakly significant effect of import penetration in the US
market on entry of offshore plants in Mexico. Another potential factor that may affect entry decisions
is cost of labor in the US relative to Mexico. We include industry hourly wages of unskilled workers
in the Mexican maquiladora sectors relative to the corresponding US industries in column 3. We find
the coefficient of the relative wage negative and significant at the 5 % level. As one may expect, cost
factors play an important role for entry of an offshore plant. We use a measure of the ’general level
of competitiveness’ of the US market in the last column: It is the industry-specific exchange rate
constructed using import partners’ shares in total imports in the particular US industry, lnM ER. An
increase in this measure refers to the appreciation of the US dollar. We find a negative and significant
27
In this specification our dependent variable conditional on our regressors is assumed to be distributed
with Negative Binomial distribution. It is a Poisson-like distribution but, unlike Poisson, equi-dispersion (that
0
is, mean equals variance V ar(yi |xi ) = exp(xi β) ) is not imposed. Variance is assumed to be V ar(yi |xi ) =
0
0
exp(xi β)+α∗(exp(xi )2 where α is an over-dispersion parameter, y is EN T RY , and x is our vector of regressors.
18
effect indicating that a decrease in the level competitiveness of the US industry is associated with
lower rate of entry to the Mexican maquiladoras. But the Chinese share of import penetration rate
keeps its sign and significance in column 4.
We focus on the Chinese share of import-penetration rate in the entry equation instead of the WTO
dummy approach due to the level of aggregation and the resulting decrease in the degrees of freedom.
Yet the results with the WTO dummy presented in the Table 5 confirm our finding that the intensified
competition from China as a result of China’s accession to WTO has a significantly negative effect on
entry of maquiladoras.
5.2.2
Plant Shut Downs
We analyze the impact of Chinese competition on maquiladora exit using a probit analysis. The results
using the WTO dummy are presented in Table 6. In these regressions, a full set of state by year fixed
effects are included (except in column 1) to control aggregate shocks such as exchange rate fluctuations
that may result in differing response across states; we also control for industry fixed factors that may
play a role in differing degrees of failure rates across industries. Plant specific factors such as size and
productivity that should affect shutdown decisions are also included. We find that the probability of
exit increases after 2001 as the coefficient of the WTO dummy is positive and significant at the 1 %
level in column 1. However, this does not necessarily indicate the impact of intensified competition.
There are other factors after 2001 that may well increase the probability of exit, such as the economic
slow-down of the US. When a full set of state by year dummies are included, the coefficient of the
interaction between the WTO dummy and plants that belong to the most affected sectors is found to be
positive and significant (in column 2) indicating that probability of exit increases disproportionately
among plants that are the most receptive to Chinese competition. In column 3 we include both
HighCHT and N oCHT to see the differential impact in comparison to moderately affected plants
(excluded group, M edCHT ). The coefficients have the right signs but are insignificant.28 In column
(4), we use a stricter classification defined by the third quarter, High3qCHT ,29 and the interaction
term is found to be positive and significant. We confirm that the probability of plant exits increases
after 2001 more so among plants that face the toughest competition from China.
28
The coefficient of interaction terms are found to be significant in column (3) when standard errors were
clustered for plants.
29
Instead of HighCHT group which is defined using median level of IM P CH in 2000.
19
Let us now discuss the coefficients of the plant-level variables. As one may expect we find a significant
and negative relationship between exit and size as well as between exit and productivity. We also
find evidence of the presence of non-linearities in the relationship between productivity and exit. The
impact of productivity on exit diminishes with productivity (negative and significant coefficient of
productivity square). A non-linear relationship is also found between plants’ age and probability of
exit, so as the plants age, the probability of exit decreases except for very young plants.30
We also look at the impact of Chinese competition on plant exit by directly using the Chinese share of
import-penetration rate in the corresponding US industry. The results of this exercise are presented
in Table 7. We find a positive and significant coefficient of the Chinese penetration rate. In column 2
aggregate demand and cost factors that may affect maquiladora exits are added in addition to plantlevel factors. As the general import penetration calculated without Chinese imports increases in the
US market, maquiladora plants’ likelihood of exit is not affected significantly in Mexico. However,
the coefficient of US production index is found to be negative and significant indicating close ties of
maquila plants with the US industries.31 The marginal effects of column 2 imply that a marginal
change in IM P CH from the average of 5.3 % is associated with a 13 % increase in probability of
plant exits. In column 3 the Chinese share of import-penetration is instrumented using the worldwide
share of Chinese imports as well as the real exchange rate between the US and China32 . The results
are robust.
30
It is typical to find a higher exit rate among younger firms/plants, since they enter without full information
about their capabilities or opportunities; so as they age, their likelihood of exit decreases. However, offshore
plant dynamics may exhibit differences in comparison to manufacturing plants. When an offshore plant starts
operation, it starts with a business tie with a company with safe demand, with no or minimal role for a demand
accumulation process to affect plant dynamics, but as time goes by, the offshore plants’ probability of loosing
the business tie might increase, perhaps due to bankruptcy or other reasons. Our findings indicate a need for
a closer look at offshore plants dynamics.
31
Bergin et al. (2009) documents excess volatility of maquiladoras in comparison to the US counterpart
which may imply that the US firms respond to shocks more strongly in their offshore plants as opposed to
home plants.
32
As explained in the Section 4 these instruments are interacted with the cross-sectional (1999) share of
IM P CH.
20
5.3
Productivity
So far our analysis reveals significant and sizable impact of intensified Chinese competition on Mexican
maquiladoras. We now move to analysis of potential link between the competition and productivity.
Our difference in difference estimation results for plant TFP are presented in Table 8.
We find that the coefficient of the WTO dummy is positive and significant at the 1 % level in column
1, indicating a general shift in the productivity of export assembly plants in Mexico in the 2000s by
about 7.2 %. It is difficult to attribute this gain to intensified Chinese competition only, as there may
be other changes in the aggregate environment. But our difference in difference approach will be able
to extract the role of Chinese competition from other changes.
In column (2) of Table 8 we present the regression result when we only include the top group
HighCHT . The coefficient of the interaction between the WTO dummy and the group of sectors
that are under the most direct threat of Chinese competition, HighCHT , is found to be positive and
significant. It indicates that the productivity increase after China’s accession to WTO is higher for
plants that belong to sectors with stronger Chinese comparative advantage. More specifically, after
the WTO accession of China, the productivity of plants in group HighCHT becomes 5.6 % higher
than the productivity of the rest of the plants after controlling for aggregate shocks. The coefficient in
front of the entrant dummy is positive but insignificant.33 The exit dummy is negative and significant,
indicating that, on average, exiting plants are 3.5 percent less productive.
In column (3) of Table 8 we repeat the same analysis for the group of plants that belong to the
least affected sectors, N oCHT , (Chemicals, Transportation and Food). We find that the interaction
between the WTO dummy and N oCHT is negative and significant. More specifically, plants that
belong to the sectors where China does not have a comparative advantage, are on average 5.5 % less
productive in comparison to the rest of the plants after China’s accession to WTO. Note that, a negative
33
Entry is generally thought to be a negative contributor to the aggregate productivity as new entrants are
on average found to be less productive than the average. Foster et al. (2008) and Foster et al. (2009) find on
the other hand that entrants are not necessarily less productive than the incumbent plants after controlling
for demand side factors. If the demand accumulation process does not play a significant role among offshore
plants, then we do not expect younger plants to charge a lower price than the older plants, everything else is
being constant. Accordingly, our productivity estimates will not underestimate the productivity of entrants
due to the omitted price problem. So in one sense, we confirm the findings of Foster et al.(2008) that entrants
are not necessarily less productive than the incumbents after controlling for demand disadvantages of entrants.
21
sign in front of the N oCHT ∗ W T O does not mean that these plants (Chemicals, Transportation
and Food) actually experienced a decline in productivity. In column (4) we included both groups,
HighCHT , and N oCHT , so that the interaction terms will indicate the productivity differential
between that group and the excluded group, which is the group with moderate exposure to Chinese
imports. The coefficient of HighCHT ∗ W T O is positive and significant indicating the productivity
differential of 4.7 % between the plants that are under threat of Chinese competition and the plants that
are moderately exposed to Chinese competition. Our difference in difference estimates confirm that
heightened Chinese competition leads to within firm productivity increase in Mexican maquiladoras.
Does the competition have effect on productivity through entry and exit? Column (4) of Table 8 shows
that the coefficient of the interaction between exit and WTO dummy is positive indicating that the
positive impact of competition on productivity is not through plant exits.34
The interaction between entry and the WTO dummy is positive although not significant, it is interesting that most of the positive effect of entry is indeed after the competition with China has
intensified.35 Looking at the average skill intensity (the ratio of administrators and technicians over
unskilled workers) of the entrants (Table A-3) we find that the mean skill-intensity of entrants increased from 0.339 to 0.609 after 2001 which supports a hypothesis that intensified competition with
China trigger industrial upgrading. This increase is not due to a lower size of entrants after 2001 since
the mean size of entrants also increases slightly in comparison to the pre-2001 level. This is in line
with the Heckscher-Ohlin theory which suggests growth in skill-intensive jobs in Mexico as a result
of competition from low skill intensive China. Skill-intensity also increases among continuing plants
from 0.28 to 0.38 on average after 2001.
From Column (5) through column(8) we repeat the analysis with the inclusion of plant-fixed effects.
The basic picture does not change, that there is significant within productivity improvement after
2001 and that this effect is stronger for plants with more exposure to Chinese imports.
34
Sargent and Matthews (2009) surveys managers of about 100 maquila plants and they find no relationship
between the use of just-in-time inventory practices, technology-intensive production systems and total quality
management practices and the recently observed plant deaths. Our results with the extensive data-set confirm
their finding.
35
Our productivity measure is robust to changes in the skill composition of the workforce as it takes into
account skilled and unskilled workers as separate inputs. However, when productivity is calculated using total
hours, we see that the coefficient of the interaction between entry and WTO dummy is positive and statistically
significant.
22
What is the over-time impact of China’s accession to WTO on the productivity of Mexican maquiladoras? One may suspect that our results are driven by the US recession (assuming somehow that those
industries that are most vulnerable to Chinese competition are disproportionately affected by the recession) between 2001 and 2003. The productivity effect should be temporary if our results would be
driven by the US recession. The results presented in Table 9 tell us that the effect of competition on
productivity increases gradually over time. The productivity differential of the most effected group of
plants and the rest of the maquiladoras is 5.9 % on average in 2001, and this differential becomes 9.2 %
on average in 2006 (column 1). The coefficient of interaction terms not only increase their significance
but also increase their magnitudes during US recovery period (2004-2006). The estimates are also
robust to the inclusion of plant fixed effects (column (2)).
In Table 10 we present our results with the continuous proxy for Chinese competition, the Chinese
import penetration in the US market, IM P CH. When we control for import-penetration rate (imports
from everywhere else), the coefficient of IM P CH is still significant at the 1 % level. The magnitude
implies that one standard deviation increase in Chinese import-penetration increases productivity by
3.6 percentage points (column (2)). When we add interaction of our entry and exit dummies with the
Chinese competition proxy (columns 3 and 4) we confirm our previous findings in Table 8.
6
6.1
Discussion
Additional Robustness Checks
Most available plant-level data do not contain plant-level prices, and ours is not an exception. This
gives rise to the concern that demand shocks may be captured by our productivity measure. Considering productivity measure as a profitability measure, we cannot think of any positive demand shocks
after 2001 that only affect the industries where China’s comparative advantage is high. We also find
that plant growth is negatively affected in these industries. Additionally, price variation between
maquiladora plants must be limited compared to 3-4 digit manufacturing industries; maquiladora
plants mainly focus on downstream processes within 3-4 digit manufacturing sectors. Another concern
would be that our productivity estimates are based on a constant returns to scale assumption. If the
underlying scale elasticities are less than one, then we may be capturing a scale effect. Unfortunately
we cannot estimate the production function using Olley and Pakes methodology due to lack of investment data. We estimate it using Levinsohn and Petrin and use imported materials as a proxy variable;
23
we also estimate it using the fixed effect model. The fixed effect model is known to underestimate the
scale.36 The scale elasticities are found to be on average 0.9, which justifies our use of index methodology. The results on the impact of Chinese competition are robust to both Levinsohn-Petrin and
fixed effect estimates of the productivity measures. In Table 11 we present the productivity regression
when we use the alternative productivity measure derived from the fixed effect model.
We also exercise with different sub-samples. Particularly, in order to remove a possible differential
effect of the 1994 peso crisis and the implementation of NAFTA on maquiladora industries, we repeat
the analysis with the 1996-2006 sample.37 The results for productivity are presented in column (1) and
(2) of Table 12. They confirm our findings. One additional robustness check would be a regression
using data only from the before period to check for pre-existing differential trends. We checked if
there is any trend difference between the industries that are the most receptive to the competition
from China and others by performing ’pseudo difference in difference’ analysis for the years between
1990 and 1999 where we assign 1997, 1998, and 1999 as pseudo-after. Any significant effect is likely
to be a sign of a violation of the assumption of the difference-in-difference estimation. The results
are presented in column (3) and (4) of Table 12. They confirm that there is no other trend difference
between the treatment and the control group.
Finally in order to address the possible under-estimation of standard errors due to possible serial correlations in the dependent variables (Bertrand, et al. (2004)), we aggregate the data into two periods:
pre- and post-WTO. The results for productivity presented in Table 13 indicate significant increase
in productivity in the post-WTO period, more so among plants that are threatened by competition
from China.
36
One drawback of the L-P approach in our context is that not all of the plants in the data-set report
imported materials, which causes an efficiency lost. Another is related to the assumption of the evolution
of capital. Since we only have information about the rental capital, it may be arguable to treat the capital
variable as a state variable. So we think the fixed effect model is more reliable.
37
See for example Verhoogen (2008) for an excellent work that shows differential impact of the 1994 peso
crisis on Mexican manufacturing plants’ quality upgrading motives. However, his work concerns differential
impact across exporters versus non-exporters. All plants in our data-set by definition are exporters.
24
6.2
Potential Sources of Productivity Gain
Our analysis shows that reallocation within industries does not play a significant role in the productivity gain that Maquiladora industries underwent in response to heightened competition. However,
there could still be significant reallocation between sectors as more productive sectors grow and less
productive ones shrink. We decompose aggregate productivity growth between 1999 and 2006 into
components of within-plant, between plants within industry, turnover and between sectors reallocation.38 Table A-4 presents the results. Between 1999 to 2006, the aggregate productivity increase in
Maquiladoras is calculated as 7.8 %. We find a significant role of between sector reallocation, in that
50 % of the observed growth can be attributed to reallocation between Maquiladora sectors. Confirming our findings, only 3.8 % of the total growth is due to reallocation between continuing plants
within industries while more than 30 % of the observed growth is due to within-plant productivity
improvement. When we repeat the exercise only for sectors that are the most threatened by Chinese
competition, HighCHT , we find that aggregate productivity growth is 10 % and that 36 % of it can
be attributed to within-plant productivity improvement.
The plant-level survey (EIME) does not allow us to do a deeper analysis for the sources of within-plant
productivity gain. However, INEGI also conducts technology surveys, called ”Encuesta Nacional De
Empleo, Salarios, Tecnologı̀a Y Capacitaciòn” (National Survey of Employment, Wages, Technology
and Training).39 This survey was also conducted among a sample of maquiladoras in 1999, 2001 and
2005. In Tables A-5 to A-8 we present information regarding operational and technological capabilities
from ENESTyC surveys which help us gain insight into potential sources of productivity gain.40 ,
41
Summary statistics indicate that surveyed maquila plants are mainly large foreign (US) owned
38
The details of the decomposition analysis is presented in the appendix.
The survey covers all plants with 100 or more employees and a sample of smaller plants.
40
We thank Dr. Alberto Ortega y Venzor for granting us access to the Encuesta Nacional De Empleo,
39
Salarios, Tecnologı̀a Y Capacitaciòn (ENESTyC) survey. We thank Mauro Antonio Gascón Varela and Gabriel
Arturo Romero Velasco from INEGI for hosting us at INEGI offices in Aguascalientes and patiently answering
our questions about the data-sets.
41
ENESTyC surveys are designed as separate cross-sections. Because of this, assigned plant identification
numbers are not unique across time. Close to 700 continuously surveyed plants are identified by INEGI,
constructing a balanced panel, which does not, however, contain any maquila plants. It is still possible to
match plants but only through company names and addresses. Until now, we are not able to access this info
due to confidentiality reasons. So that for maquila plants, plant identifications were not possible across time
nor between EIME and ENESTyC data-sets. Industry affiliation information for maquila plants in 2005 are
25
plants. 589, 675 and 791 maquila plants are surveyed in 1999, 2001 and 2005 respectively. When we
compare 1999 with 2005 results, we see that the number of plants that report performing research and
development increases from 39 to 46 %. Plants that report performing product development increases
from 21 to 32 %. Table A-7 presents percentage capacity utilization among maquila plants as well
as among non-maquila manufacturing plants. The average capacity utilization increases substantially
from 81 to 86 % among maquilas. Interestingly, there is no comparable increase in capacity utilization
among non-maquila manufacturing plants. In Table A-8, we present information on management
techniques that is derived from ENESTyC 2005.42 Approximately 20 % of 642 plants implements Just
in Time inventory methods before 2000 and this increases to 41 % in the beginning of 2005. Similarly
the percentage of firms implementing other productivity enhancing management techniques such as
Total Quality Management, Job Rotation, Process Re-engineering, Re-arrangement of Equipment43
also doubles or more than doubles between 2000 and 2005.
7
Conclusion
We analyze the impact of competition with Chinese products in the United States market on the
offshore plants of the US companies specializing in similar processes as Chinese plants. Analyzing the
impact of Chinese competition on Mexican export assembly plants, we find that employment growth
and entry are negatively affected by Chinese competition. We also provide evidence that heightened
competition with China triggered maquiladora exit. We do not find evidence that Chinese competition affected plants’ growth or exit disproportionately within industries so as to lead to productivity
improvement through reallocation. Instead we find that competition especially affected the most
unskilled labor-intensive sectors among the ones that are threatened most by Chinese competition,
leading to sectoral reallocation.
Using a difference in difference approach we quantify a positive, and both economically and statistically
not yet available to us, so we are unable to provide sector-specific information at this time.
42
This question (Section 12, question 6 in ENESTyC 2005) asks whether a specified organization technique
has been implemented, and the starting year of the implementation.
43
Re-arrangement of Equipment is defined in the survey as ”management and organization of machinery,
equipment and facilities to carry out more efficient production and decrease the possibility of occupational
hazards”. The other productivity enhancing management concepts are well-known and their definitions follows
standard management literature.
26
significant productivity improvement in Mexican maquiladoras, which can be attributed to intensified
competition from China. We find that most of the productivity gain experienced within industries
is due to within plant improvement. The results provide evidence in support of models that imply
a positive relationship between international competition and within firm productivity. The results
also support field studies and anecdotal evidence of industrial upgrading in Mexican Maquiladoras in
response to competition with low-wage locations such as China.
Overall we identify an interesting link between competition of two popular offshoring destinations
(Mexico and China) for, mainly US based, multinationals and within plant productivity improvement;
a link that may relieve some of the worries that Mexican policy makers express over growing trade
from China.
27
References
[1] Aw, Bee, Xiaomin Chen and Mark J. Roberts (1997), ”Firm-level Evidence on Productivity Differentials, Turnover, and Exports in Taiwanese Manufacturing”, Journal of Development Economics,
Vol.66, No:1, pp. 51-86.
[2] Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith and Peter Howitt (2005), ”Competition and Innovation: An Inverted-U Relationship”, The Quarterly Journal of Economics
[3] Arrow, Kenneth (1962), ”Economic Welfare and Allocation of Resources for Invention”, appeared
in The Rate and the Direction of the Inventive Activity, Princeton University Press, pp. 609-626.
[4] Bair, Jennifer and Gereffi (2003), ”Upgrading, Uneven Development, and Jobs in the North
American Apparel Industry”, Global Networks, No: 3-2, pp.143-169.
[5] Bergin, Paul, Robert Feenstra, and Gordon Hanson, ”Outsourcing and Volatility”, forthcoming
in American Economic Review.
[6] Bernard, Andrew B, Bradford Jensen, and Peter K. Schott (2006), ”Survival of the Best Fit: Exposure to Low-Wage Countries and the (uneven) Growth of U.S. Manufacturing Plants”, Journal
of International Economics, 68, pp. 219-237.
[7] Bertrand, Marianne, Esther Duflo and Sendhil Mullainathan (2004), ”How Much Should We
Trust Differences-in-Differences Estimates?”, The Quarterly Journal of Economics, No:119(1),
pp. 249-275.
[8] Bloom, Nicholas, Mirko Draca, John Van Reenen (2009), ”Trade induced technical change? The
impact of Chinese imports on innovation and information technology”, Working Paper, Stanford
University.
[9] Bugamelli, Matteo, Silvia Fabiani and Enrico Sette (2010), ”The pro-competitive effect of imports
from China: an analysis of firm-level price data”, Banca D’italia Working Papers No. 737.
[10] Canas, Jesus, and Roberto Coronado (2002), ”Maquiladora Industry: Past, Present and Future”,
El Paso Business Frontier, Federal Reserve Bank of Dallas, El Paso Branch.
[11] Carillo, Jorge and Arturo Lara (2005), ”Mexican Maquiladoras: New Capabilities of Coordinatiion and Emergence of New Generation of Companies”, Innovation: Management, Policy and
Practice Vol. 7-2, pp. 256-273
28
[12] Caves, W. Douglas, R. Laurits Christensen, and W. Erwin Diewert (1982) ”The Economic Theory
of Index Numbers and Measurement of Input, Output and Productivity”, Econometrica, Vol 50,
No:6, pp-1393-1414.
[13] Devlin, Robert, Antoni Estevadeordal and Andrés Rodrı́guez-Clare (2006), The Emergence of
China: Opportunities and Challenges for Latin America and the Caribbean, Washington D.C.:
Inter-American Development Bank.
[14] Dixit, Avinash, and Joseph Stiglitz (1977) ”Monopolistic Competition and Optimum Product
Diversity”, American Economic Review, Vol 67, pp-297-308.
[15] Foster, Lucia, John Haltowanger, J. C. Krizan (1999), ”Aggregate Productivity Growth: Lessons
from Microeconomic Evidence”, NBER Working Paper No. 6803.
[16] Foster, Lucia, John Haltiwanger, and Chad Syverson (2008) ”Reallocation, Firm Turnover and
Efficiency: Selection on Productivity or Profitability?, American Economic Review
[17] Foster, Lucia, John Haltiwanger, and Chad Syverson (2009) ”The Slow Growth of New Plants:
Learning about Demand?”, working paper.
[18] Gallagher, Kevin P. and Roberto Porzecanski (2007), ”What a Difference a Few Years Makes:
China and the Competitiveness of Mexican Exports”, Oxford Development Studies, Vol. 35, No.
2
[19] Gallagher, Kevin P., Juan Carlos Moreno-Brid, and Roberto Porzecanski (2008), ”The Dynamism
of Mexican Exports: Lost in (Chinese) Translation?, World Development, 36(8), 1365-1380.
[20] Goldberg, Linda (2004), ”Industry-Specific Exchange Rate for the United States”, FRBNY Economic Policy Review
[21] Good, David, Ishaq Nadiri and Robin Sickles (1997), ”Index Number and Factor Demand Approaches to the Estimation of Productivity”, H. Pesaran and P. Schmidt, eds. Handbook of
Applied Econometrics Vol. II-Microeconometrics, Malden, MA: Blackwell Publishers, 14-80.
[22] Griliches, Zvi and Haim Regev (1995), ”Productivity and Firm Turnover in Israeli Industry:
1979-1988”, Journal of Econometrics, Vol. 65, pp. 175-203.
[23] Gruben, William C. (2001), ”Was NAFTA Behind Mexico’s High Maquiladora Growth”, Economic and Financial Review, Third Quarter, Dallas FED.
[24] Hanson, Gordon (2002), ”The Role of Maquiladoras in Mexico’s Export Boom”, mimeo, University of California at San Diego, La Jolla.
29
[25] Hanson, Gordon and Raymond Robertson (2008), ”China and the Manufacturing Exports of
Other Developing Countries”, NBER Working papers No 14497.
[26] Hart, Oliver (1983), ”The Market Mechanism as an Incentive Scheme”, The Bell Journal of
Economics, Vol. 14, No. 2, pp. 366-382
[27] Hermalin, E. Benjamin (1992), ”The Effects of Competition on Executive Behavior” The RAND
Journal of Economics, Vol. 23, No. 3, pp. 350-365
[28] Holmes, Thomas J. and James A. Schmitz, (2009), ”Competition and Productivity: A Review of
Evidence”, Federal Reserve Bank of Minneapolis, Research Development Staff Report 439.
[29] Iacovone Leonardo, Wolfgang Keller, and Ferdinand Rauch (2010), ”Innovation When The Market
Is Shrinking: Firm-level Responses to Competition From China”, mimeo.
[30] International Labour Office (ILO) (2003), ”Employment and Social Policy in Respect of Export Processing Zones (EPZs)”, Governing Body, Committee on Employment and Social Policy,
Geneva, March 2003, pp. 1-22
[31] Jovanovic, Boyan (1982),”Selection and the Evolution of Industry”, Econometrica, Vol. 50, No.
3, pp. 649-670
[32] Lall, Sanjaya and John Weiss (2004), ”China’s Competitive Threat to Latin America: An Analysis
for 1990-2002”, Oxford Development Studies, Vol.33, No.2, pp.163-194.
[33] Levinsohn, James and Amil Petrin (2003), ”Estimating Production Functions Using Inputs to
Control for Unobservables,” The Review of Economic Studies, 70 317-342.
[34] Lindquist, Diane (2004), ”The Maquiladora Roars Back”, The San Diego Union Tribune, June
29, 2004.
[35] Lopez-Cordova, J. Ernesto, Alejandro Micco and Danielken Molina (2008), ”How Sensitive are
Latin American Exports to Chinese Competition in the U.S. Market?”, World Bank Policy Research Working Paper Series, No 4497.
[36] Mendoza, Jorge Eduardo (2010) ”The Effect of the Chinese Economy on Mexican Maquiladora
Employment”, The International Trade Journal, Vol.24 No.1, pp. 52-83.
[37] Moulton, Brent R. (1990) ”An Illustration of a Pitfall in Estimating the Effects of Aggregate
Variables in Micro Units”, The Review of Economics and Statistics, 72(2), pp.334-338.
30
[38] Olley, Steven and Ariel Pakes (1996), ”The Dynamics of Industry in the Telecommunications
Equipment Industry,” Econometrica, 64 (6) 1263-1297.
[39] Peters, Enrique Dussel (2005), ”The Implications of China’s Entry into the WTO for Mexico”,
Global Issue Papers, No:24, the Heinrich Böll Foundation
[40] Salop, Steven (1977), ”The Noisy Monopolist: Imperfect Information, Price Dispersion and Price
Discrimination”, The Review of Economic Studies, Vol. 44, No. 3, pp. 393-406
[41] Sargent, John and Linda Matthews (2006), ”The Drivers of Evolution/Upgrading in Mexico’s
Maquiladoras: How Important is Subsidiary Initiative?”, Journal of World Business, Vol. 41,
pp.233-246.
[42] Sargent, John and Linda Matthews (2009), ”China versus Mexico in the Global EPZ Industry:
Maquiladoras, FDI Quality, and Plant Mortality”, World Development, Vol.37, pp. 1069-1082.
[43] Scharfstein, David (1988) ”Product-market competition and managerial slack” Rand Journal of
Economics 19, pp. 147155.
[44] Schmidt, Klaus (1997) ”Managerial Incentives and Product Market Competition”, The Review
of Economic Studies, Vol 64, No. 2, pp.191-213.
[45] Schmitz, James A. (2005), ”What determines productivity? Lessons from the dramatic recovery
of the U.S. and Canadian iron ore industries following their early 1980s crisis.”, Journal of Political
Economy, 113(3), pp.582-625.
[46] Truett, Lila, and Dale Truett (2007), ”NAFTA and The Maquiladoras: Boon or Bane?” Contemporary Economic Policy, Vol. 25, No. 3, pp. 374-386
[47] Verhoogen, A. Eric (2008), ”Trade, Quality Upgrading and Wage Inequality in the Mexican
Manufacturing Sector”, Quarterly Journal Of Economics, Vol. 123, No. 2, pp.489-530.
31
Tables and Figures
22
20
Logarithm of US Dollar
8
18
16
US
Canada
Japan
Switzerland
South Korea
Germany
Holland
14
12
10
1994
1996
1998
2000
2002
2004
2006
Source: Banco de Mexico
Figure 1: Capital Equipment Investment in Maquiladora Industry By Country
32
7
2.5
x 10
Nation−wide Total (INEGI)
Aggregated from the Plant−level Data Set
thousand USD
2
1.5
1
0.5
0
1990
1992
1994
1996
1998
Year
2000
2002
2004
2006
(a) In Level
0.5
Nation−wide Total (INEGI)
Aggregated from the Plant−level Data Set
0.4
growth rate
0.3
0.2
0.1
0
−0.1
−0.2
1990
1992
1994
1996
1998
Year
2000
2002
2004
2006
(b) Growth Rates
Figure 2: Maquila Export Value Added (Source: INEGI )
33
Table 1: The Impact of Chinese Competition on Employment Growth I
Dependent Variable
WTO
HighCHT*WTO
NoCHT*WTO
(1)
∆lnE
(2)
∆lnE
−0.277∗∗∗
(0.034)
−0.062
(0.045)
0.058
(0.041)
−0.080∗
(0.035)
0.104∗
(0.041)
0.051∗
(0.022)
0.124∗∗∗
(0.027)
−0.087∗∗∗
(0.019)
−0.512∗∗∗
(0.029)
−0.788∗∗∗
(0.037)
−0.967∗∗∗
(0.049)
−1.057∗∗∗
(0.061)
No
X
X
3068
18222
0.206
0.115∗∗
(0.041)
0.067∗∗
(0.022)
−0.060
(0.045)
−0.229∗∗∗
(0.028)
−0.493∗∗∗
(0.029)
−0.755∗∗∗
(0.036)
−0.918∗∗∗
(0.048)
−1.008∗∗∗
(0.061)
X
–
X
3068
18222
0.230
(3)
∆lnE
(4)
∆lnE
0.084∗
(0.035)
−0.063
(0.038)
0.048
(0.038)
0.113∗∗
(0.041)
0.067∗∗
(0.022)
−0.063
(0.045)
−0.230∗∗∗
(0.028)
−0.491∗∗∗
(0.028)
−0.752∗∗∗
(0.036)
−0.916∗∗∗
(0.048)
−1.006∗∗∗
(0.061)
X
–
X
3068
18222
0.230
0.116∗∗
(0.041)
0.067∗∗
(0.022)
−0.060
(0.045)
−0.229∗∗∗
(0.028)
−0.493∗∗∗
(0.029)
−0.755∗∗∗
(0.036)
−0.918∗∗∗
(0.048)
−1.008∗∗∗
(0.061)
X
–
X
3068
18222
0.230
High3qCHT*WTO
lnT F Pijst−1
Multi-plant Dummy
Age Dummy 1
Age Dummy 2
Size Dummy 2
Size Dummy 3
Size Dummy 4
Size Dummy 5
Year by State Fixed Effects
State Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
(5)
∆lnE
0.055
(0.037)
−0.205∗∗∗
(0.056)
0.112∗∗
(0.041)
0.064∗∗
(0.022)
−0.071
(0.045)
−0.230∗∗∗
(0.027)
−0.492∗∗∗
(0.028)
−0.751∗∗∗
(0.036)
−0.919∗∗∗
(0.048)
−1.013∗∗∗
(0.061)
X
–
X
3068
18222
0.232
Note: Dependent variable is the change in the logarithm of employment between t and t-1. Robust standard errors are
reported in parentheses. They are clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10
%, 5% and 1% levels respectively. The constant is included but not reported.
34
Table 2: The Impact of Chinese Competition on Employment Growth II
Specification
Dependent Variable
IM P CHjt−1
IM Pjt−1
(1)
OLS
∆lnE
(2)
OLS
∆lnE
(3)
IV
∆lnE
(4)
IV
∆lnE
-1.535
(0.437)∗∗∗
-0.961
(0.440)∗
-1.347
(0.407)∗∗
-0.563
(0.431)
0.116
(0.035)∗∗
0.135
(0.240)
0.112
(0.040)∗∗
-0.064
(0.046)
-0.227
(0.028)∗∗∗
0.067
(0.022)∗∗
-0.493
(0.028)∗∗∗
-0.754
(0.036)∗∗∗
-0.924
(0.48)∗∗∗
-1.022
(0.061)∗∗∗
X
X
3070
18222
0.232
-2.415
(0.378)∗∗∗
-2.809
(0.409)∗∗∗
-1.275
(0.240)∗∗∗
0.119
(0.030)∗∗∗
-0.065
(0.033)∗
-0.232
(0.022)∗∗∗
0.063
(0.023)∗∗
-0.494
(0.019)∗∗∗
-0.758
(0.021)∗∗∗
-0.924
(0.030)∗∗∗
-1.019
(0.036)∗∗∗
X
X
2643
17746
0.229
0.4857
0.5247
0.133
(0.031)∗∗∗
-0.076
(0.033)∗
-0.232
(0.023)∗∗∗
0.066
(0.024)∗∗
-0.488
(0.020)∗∗∗
-0.746
(0.022)∗∗∗
-0.908
(0.031)∗∗∗
-1.000
(0.038)∗∗∗
X
X
2551
16713
0.231
0.5301/0.8312
0.1147
U SP roductionIndexjt−1
RelW agejt−1
lnT F Pijst−1
Age Dummy 1
Age Dummy 2
Multi-plant Dummy
Size Dummy 2
Size Dummy 3
Size Dummy 4
Size Dummy 5
Plant Fixed Effects
Year by State Fixed Effects
Number of Plants
Number of Observations
R2
SheaP artialR2
Sargan Test (P-value)
0.115
(0.040)∗∗
-0.062
(0.045)
-0.227
(0.028)∗∗∗
0.066
(0.022)∗∗
-0.493
(0.028)∗∗∗
-0.755
(0.036)∗∗∗
-0.921
(0.048)∗∗∗
-1.014
(0.061)∗∗∗
X
X
3070
18222
0.231
Note: Dependent variable is the change in the logarithm of employment between t and t-1. Robust standard
errors are reported in parentheses. For the OLS estimates standard errors are clustered for each industry in
each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1% levels respectively. For IV regressions
instruments we use are explained in the text. The constant is included but not reported.
35
Table 3: The Impact of Chinese Competition on Employment III
Dependent Variable
IM P CHjt−1
IM Pjt−1
lnT F Pijt−1
(1)
∆lnE
(2)
∆lnE
(3)
∆lnE
-1.117
(0.420)∗∗
-0.457
(0.439)
0.175
(0.047)∗∗∗
-1.540
(0.437)∗∗∗
-0.974
(0.441)∗
0.121
(0.041)∗∗
0.009
(0.013)
-1.600
(0.432)∗∗∗
-1.009
(0.436)∗
0.122
(0.041)∗∗
Skill Intensity (NP/P)ijt−1
Capital-Labor Ratio (K/L)ijt−1
0.018
(0.009)∗
IM P CHjt−1 ∗ lnT F Pijt−1
-1.152
(0.623)
IM P CHjt−1 ∗ Skill Intensity (NP/P)ijt−1
-0.029
(0.088)
IM P CHjt−1 ∗ Capital Intensity (K/Y)ijt−1
Plant-Level Controls
Industry-Level Controls
Year by State Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
Yes
Yes
X
X
3068
18222
0.226
Yes
Yes
X
X
3062
18206
0.225
0.040
(0.175)
Yes
Yes
X
X
3050
18159
0.228
Note: Dependent variable is the change in the logarithm of employment between t and t-1. Robust standard
errors are reported in parentheses. They are clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate
significance at the 10 %, 5% and 1% levels respectively. Plant-level controls include multi-plant dummy and
size dummies and age dummies. Industry-level controls include Mexican industry hourly wages relative to
the corresponding US industry, and the production index of the corresponding US industry. The constant is
included but not reported.
36
37
M XW agejt
U SW agejt )
176
975
-490
-2.853
(0.233)∗∗∗
X
X
-4.721
(1.076)∗∗∗
(1)
Negative Binomial
EN T RY
176
1056
-489
-2.922
(0.249)∗∗∗
X
X
-5.326
(1.063)∗∗∗
-1.558
(0.722)∗
(2)
Negative Binomial
EN T RY
176
1118
-488
-2.879
(0.236)∗∗∗
X
X
-2.930
(1.077)∗∗
-4.744
(1.038)∗∗∗
(3)
Negative Binomial
EN T RY
176
1208
-480
-3.065
(1.139)∗∗
-4.035
(0.953)∗∗∗
-3.140
(0.271)∗∗∗
-4.246
(0.989)∗∗∗
(4)
Negative Binomial
EN T RY
10 %, 5% and 1% levels respectively. The constant is included but not reported.
Dependent variable is the total number of entrants at period t and industry j. Robust standard errors are reported in parentheses. ∗ , ∗∗ and ∗∗∗ indicate significance at the
N
χ2
Log pseudolikelihood
Industry Fixed Effects
Year Fixed Effects
ln(α) (over-dispersion parameter)
Industry Specific Exchange Rate (lnM ERjt )
Relative Wage (
IMP
IMPCH
Specification
Variables
Table 4: The Impact of Chinese Competition on Entry to Mexican Offshoring Industry I
Table 5: The Impact of Chinese Competition on Entry to Mexican Offshoring Industry II
Specification
Variables
WTO*HighCHT
(1)
Negative Binomial
EN T RY
(2)
Negative Binomial
EN T RY
(3)
Negative Binomial
EN T RY
-2.813
(0.248)
X
X
0.410
(0.164)∗
-2.815
(0.246)
X
X
-0.336
(0.138)∗
0.233
(0.185)
-2.891
(0.265)
X
X
176
906
-494
176
906
-496
176
931
-493
-0.434
(0.124)∗∗∗
WTO*NoCHT
ln(α) (over-dispersion parameter)
Industry Fixed Effects
Year Fixed Effects
N
χ2
Log pseudolikelihood
Dependent variable is the total number of entrants at period t and industry j. Robust standard errors are reported in parentheses.
∗
, ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1% levels respectively. The constant is included but not reported.
38
Table 6: The Impact of Chinese Competition on Maquiladora Exits (WTO
Dummy)
Specification
Variables
WTO
WTO*HighCHT
WTO*NoCHT
(1)
Probit
χ
(2)
Probit
χ
(3)
Probit
χ
0.525
(0.102)∗∗∗
0.117
(0.135)
-0.138
(0.170)
0.164
(0.067)∗
0.137
(0.074)
-0.091
(0.097)
-0.335
(0.018)∗∗∗
-0.431
(0.108)∗∗∗
0.391
(0.081)∗∗∗
0.121
(0.040)∗∗
-0.241
(0.057)∗∗∗
0.074
(0.053)
No
X
X
-0.353
(0.019)∗∗∗
-0.406
(0.107)∗∗∗
0.390
(0.082)∗∗∗
0.095
(0.042)∗
-0.295
(0.058)∗∗∗
0.040
(0.051)∗∗∗
X
–
X
-0.353
(0.019)∗∗∗
-0.408
(0.106)∗∗∗
0.392
(0.082)∗∗∗
0.095
(0.042)∗
-0.295
(0.058)∗∗∗
0.039
(0.051)∗∗∗
X
–
X
-0.111
(0.090)
0.237
(0.088)∗∗
-0.353
(0.019)∗∗∗
-0.410
(0.106)∗∗∗
0.396
(0.083)∗∗∗
0.095
(0.042)∗
-0.301
(0.058)∗∗∗
0.027
(0.052)∗∗∗
X
–
X
19372
0.181
18504
0.227
18504
0.227
18504
0.228
WTO*High3qCHT
Size
Productivity
Productivity Square
Multi-plant Dummy
Age Dummy 1
Age Dummy 2
Year by State Fixed Effects
State Fixed Effects
Industry Fixed Effects
N
Pseudo R2
(4)
Probit
χ
Dependent variable is the indicator variable that takes 1 if the plant does not participate the next period
(t+1). Robust standard errors are reported in parentheses. They are clustered for each industry in each
year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1% levels respectively. Size variable is
measured by logarithm of labor. The constant is included but not reported.
39
Table 7: The Impact of Chinese Competition on Maquiladora
Exits (IMPCH)
Specification
Variables
IMPCH
(1)
Probit
χ
(2)
Probit
χ
(3)
IV
χ
2.022
(0.769)∗∗
2.537
(0.919)∗∗
-0.353
(0.019)∗∗∗
-0.406
(0.106)∗∗∗
0.393
(0.082)∗∗∗
-0.296
(0.057)∗∗∗
0.039
(0.051)∗∗∗
0.096
(0.042)∗
X
X
2.043
(0.785)∗∗
1.008
(0.638)
-0.156
(0.075)∗
0.458
(0.636)
-0.354
(0.019)∗∗∗
-0.405
(0.106)∗∗∗
0.391
(0.082)∗∗∗
-0.300
(0.057)∗∗∗
0.028
(0.051)∗∗∗
0.094
(0.042)∗
X
X
18504
0.228
18504
0.229
IMP
US Production Index
Relative Wage (
M XW agejt
U SW agejt )
Size
Productivity
Productivity Square
Age Dummy 1
Age Dummy 2
Multi-plant Dummy
Year by State Fixed Effects
Industry Fixed Effects
N
Pseudo R2
-0.352
(0.013)∗∗∗
-0.410
(0.088)∗∗∗
0.394
(0.069)∗∗∗
-0.294
(0.055)∗∗∗
0.040
(0.053)∗∗∗
0.097
(0.043)∗
X
X
18504
Dependent variable is the indicator variable that takes 1 if the plant does not participate
the next period (t+1). Robust standard errors are reported in parentheses. They are
clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10
%, 5% and 1% levels respectively. Size variable is measured by logarithm of labor. The
constant is included but not reported. In the IV regression, the Wald test does not reject
exogeneity of IM P CH.
40
41
X
No
X
No
3062
18572
0.213
-0.038
(0.012)∗∗
0.072
(0.012)∗∗∗
0.043
(0.015)∗∗
-0.028
(0.020)
0.015
(0.004)∗∗∗
-0.018
(0.006)∗∗
-0.018
(0.005)∗∗∗
0.013
(0.007)
(1)
lnTFP
No
X
X
No
3062
18572
0.225
-0.035
(0.012)∗∗
0.018
(0.004)∗∗∗
-0.008
(0.006)
-0.010
(0.005)
0.009
(0.007)
0.056
(0.009)∗∗∗
(2)
lnTFP
No
X
X
No
3062
18572
0.223
-0.034
(0.012)∗∗
-0.055
(0.015)∗∗∗
0.017
(0.004)∗∗∗
-0.009
(0.007)∗
-0.011
(0.005)∗
0.010
(0.007)
(3)
lnTFP
0.047
(0.010)∗∗∗
-0.029
(0.017)
0.017
(0.004)∗∗∗
-0.010
(0.007)∗
-0.011
(0.006)∗
0.002
(0.009)
0.023
(0.014)
-0.037
(0.019)∗
0.004
(0.025)
No
X
X
No
3062
18572
0.225
(4)
lnTFP
No
X
No
X
3062
18572
0.062
-0.010
(0.009)
0.011
(0.006)
-0.018
(0.009)∗
-0.012
(0.006)∗
0.013
(0.006)∗
0.040
(0.008)∗∗∗
(5)
lnTFP
No
X
No
X
3062
18572
0.060
-0.010
(0.009)
-0.031
(0.012)∗∗
0.011
(0.006)
-0.007
(0.009)
-0.012
(0.006)
0.014
(0.006)∗
(6)
lnTFP
0.036
(0.009)∗∗∗
-0.011
(0.014)
0.011
(0.006)
-0.018
(0.009)
-0.013
(0.006)∗
0.013
(0.006)
0.010
(0.013)
-0.026
(0.011)∗
0.026
(0.016)
No
X
No
X
3062
18572
0.063
(7)
lnTFP
year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1% levels respectively. The constant is included but not reported.
Note: Dependent variable is the logarithm of productivity. Robust standard errors are reported in parentheses. They are clustered for each industry in each
State Fixed Effects
Year by State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
Exit*WTO
Exit Dummy
Entrant*WTO
Entrant Dummy
Age Dummy 2
Age Dummy 1
Multi-Plant Dummy
NoCHT*WTO
HighCHT*WTO
WTO
Dependent Variable
Table 8: The Impact of Chinese Competition on Productivity I
Table 9: The Impact of Chinese Competition on
Productivity II
Dependent Variable
HighCHT*1992
HighCHT*1993
HighCHT*1994
HighCHT*1995
HighCHT*1996
HighCHT*1997
HighCHT*1998
HighCHT*1999
HighCHT*2000
HighCHT*2001
HighCHT*2002
HighCHT*2003
HighCHT*2004
HighCHT*2005
HighCHT*2006
Multi-Plant Dummy
Age Dummy 1
Age Dummy 2
Year by State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
(1)
lnTFP
(2)
lnTFP
0.014
(0.046)
−0.004
(0.039)
0.004
(0.033)
0.010
(0.030)
0.011
(0.030)
0.024
(0.030)
0.026
(0.032)
0.025
(0.031)
0.032
(0.032)
0.059
(0.029)∗
0.067
(0.029)∗
0.070
(0.029)∗
0.082
(0.031)∗∗
0.082
(0.031)∗∗
0.092
(0.034)∗∗
0.017
(0.004)∗∗∗
−0.008
(0.006)
−0.011
(0.005)∗
X
X
No
3257
20742
0.234
0.016
(0.028)
0.002
(0.026)
0.015
(0.023)
0.030
(0.023)
0.037
(0.022)
0.039
(0.021)
0.036
(0.024)
0.040
(0.024)
0.044
(0.024)
0.066
(0.022)∗∗
0.070
(0.023)∗∗
0.069
(0.022)∗∗
0.079
(0.025)∗∗
0.087
(0.024)∗∗∗
0.096
(0.025)∗∗∗
0.013
(0.006)∗
−0.018
(0.009)∗
−0.015
(0.006)∗∗
X
No
X
3257
20742
0.068
Note: Dependent variable is the logarithm of productivity. Robust standard errors are reported in parentheses. They are clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5%
and 1% levels respectively. The constant is included but not reported.
42
Table 10: The Impact of Chinese Competition on Productivity III
Dependent Variable
IMPCH
(1)
lnTFP
(2)
lnTFP
(3)
lnTFP
(4)
lnTFP
0.488
(0.142)∗∗∗
0.018
(0.004)∗∗∗
-0.009
(0.007)
-0.011
(0.005)
0.009
(0.007)∗∗
0.538
(0.135)∗∗∗
0.165
(0.079)∗
0.018
(0.004)∗∗∗
-0.009
(0.007)
-0.011
(0.006)∗
0.010
(0.007)∗
-0.034
(0.012)∗∗
-0.035
(0.012)∗∗
X
X
No
18572
0.223
X
X
No
18572
0.224
0.510
(0.136)∗∗∗
0.165
(0.079)∗
0.017
(0.004)∗∗∗
-0.010
(0.007)
-0.012
(0.006)∗
0.005
(0.009)
0.082
(0.088)
-0.045
(0.015)∗∗
0.143
(0.109)
X
X
No
18572
0.224
0.406
(0.121)∗∗∗
-0.012
(0.083)
0.012
(0.006)
-0.016
(0.009)
-0.012
(0.006)
0.012
(0.007)
0.025
(0.082)
-0.023
(0.011)∗
0.189
(0.096)
X
No
X
18572
0.062
IMP
Multi-Plant Dummy
Age Dummy 1
Age Dummy 2
Entrant Dummy
Entrant*IMPCH
Exit Dummy
Exit*IMPCH
Year by State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Observations
R2
Sargan Test (P-value)
(5)
lnTFP
IV
0.459
(0.077)∗∗∗
0.012
(0.006)
-0.017
(0.010)
-0.012
(0.007)
0.013
(0.005)∗∗
-0.009
(0.006)
X
No
X
18175
0.0526
Note: Dependent variable is the logarithm of productivity. Robust standard errors are reported in parentheses. They are
clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1 % levels respectively.
The constant is included but not reported.
43
44
No
X
X
No
3062
18572
0.050
-0.091
(0.017)∗∗∗
0.081
(0.018)∗∗∗
0.096
(0.025)∗∗∗
-0.017
0.027
-0.025
(0.006)∗∗∗
-0.014
(0.012)
-0.012
(0.008)
0.005
(0.011)
(1)
lnT F PF X
X
–
X
No
3062
18572
0.063
-0.081
(0.018)∗∗∗
-0.017
(0.006)∗∗
0.010
(0.012)
-0.007
(0.009)
-0.009
(0.011)
0.106
(0.013)∗∗∗
(2)
lnT F PF X
X
–
X
No
3062
18572
0.060
-0.079
(0.018)∗∗∗
-0.080
(0.019)∗∗∗
-0.018
(0.006)∗∗
0.008
(0.012)
-0.010
(0.008)
-0.009
(0.011)
(3)
lnT F PF X
0.098
(0.015)∗∗∗
-0.026
(0.022)
-0.018
(0.006)∗∗
0.007
(0.012)
-0.010
(0.009)
-0.023
(0.013)
0.043
(0.023)∗∗
-0.091
(0.028)∗∗
0.018
(0.036)
X
–
X
No
3062
18572
0.063
(4)
lnT F PF X
X
–
No
X
3062
18572
0.053
-0.050
(0.012)∗∗∗
-0.014
(0.009)
-0.008
(0.014)
-0.012
(0.009)
-0.006
(0.007)
0.066
(0.011)∗∗∗
(5)
lnT F PF X
X
–
No
X
3062
18572
0.050
-0.047
(0.012)∗∗∗
-0.034
(0.015)∗
-0.014
(0.009)
-0.005
(0.015)
-0.011
(0.010)
-0.005
(0.008)
(6)
lnT F PF X
0.067
(0.013)∗∗∗
0.005
(0.017)
-0.014
(0.009)
-0.008
(0.015)
-0.012
(0.010)
-0.011
(0.008)
0.017
(0.020)
-0.092
(0.016)∗∗∗
0.066
(0.023)∗∗
X
–
No
X
3062
18572
0.054
(7)
lnT F PF X
levels respectively. The constant is included but not reported.
standard errors are reported in parentheses. They are clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10 %, 5% and 1%
Note: Dependent variable is the logarithm of productivity where productivity is derived from the estimated production function using fixed effect model. Robust
Year by State Fixed Effects
State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
Exit*WTO
Exit Dummy
Entrant*WTO
Entrant Dummy
Age Dummy 2
Age Dummy 1
Multi-Plant Dummy
NoCHT*WTO
HighCHT*WTO
WTO
Dependent Variable
Table 11: Robustness Check: The Impact of Chinese Competition on Productivity (FIXED)
Table 12: Additional Robustness Checks: Alternative Samples and Pseudo
Checks
Sample
Dependent Variable
HighCHT*WTO
NoCHT*WTO
Multi-Plant Dummy
Age Dummy 1
Age Dummy 2
Entrant Dummy
Exit Dummy
Year by State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
(1)
1996-2006
lnTFP
(2)
1996-2006
lnTFP
(3)
1990-1999
lnTFP
(4)
1990-1999
lnTFP
0.043
(0.007)∗∗∗
-0.014
(0.017)
0.018
(0.005)∗∗∗
-0.005
(0.007)
-0.013
(0.006)∗
0.019
(0.008)∗
-0.046
(0.013)∗∗∗
X
X
No
2659
13880
0.240
0.031
(0.009)∗∗∗
-0.006
(0.012)
0.020
(0.009)∗
-0.014
(0.009)
-0.009
(0.006)
0.014
(0.007)∗
0.000
(0.011)
X
No
X
2659
13880
0.064
0.007
(0.019)
-0.035
(0.022)
0.020
(0.005)∗∗∗
-0.015
(0.016)
0.002
(0.015)
0.005
(0.010)
-0.036
(0.020)
X
X
No
2179
9985
0.194
0.015
(0.016)
-0.016
(0.018)
0.005
(0.008)
-0.033
(0.014)∗
-0.010
(0.009)
0.004
(0.007)
-0.024
(0.012)∗
X
No
X
2179
9985
0.036
Note: Dependent variable is the logarithm of productivity. Robust standard errors are reported in parentheses. They are clustered for each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at the 10
%, 5% and 1% levels respectively. The constant is included but not reported.
45
Table 13: Robustness Check: Pre and Post
WTO
Dependent Variable
WTO
HighCHT*WTO
NoCHT*WTO
Multi-Plant Dummy
Entrant Dummy
Exit Dummy
State Fixed Effects
Industry Fixed Effects
Plant Fixed Effects
Number of Plants
Number of Observations
R2
(1)
lnT F P
(2)
lnT F P
0.071
(0.014)∗∗∗
0.048
(0.017)∗∗
-0.018
(0.024)
0.010
(0.009)
0.004
(0.009)
-0.056
(0.010)∗∗∗
X
X
No
3151
4667
0.225
0.072
(0.011)∗∗∗
0.043
(0.013)∗∗∗
-0.015
(0.019)
0.007
(0.010)
0.014
(0.008)
-0.046
(0.009)∗∗∗
X
3151
4667
0.084
Note: Dependent variable is the logarithm of productivity. Robust
standard errors are reported in parentheses. They are clustered for
each industry in each year. ∗ , ∗∗ and ∗∗∗ indicate significance at
the 10 %, 5% and 1% levels respectively. The constant is included
but not reported. The time-series dimension of the data is collapsed
for pre and post WTO periods by taking average of productivity
for each plant. The indicator variables such as entry and exit are
defined as one if the plant satisfies the condition in any of the year
throughout 1990-2000 and 2001-2006 periods.
46
APPENDIX
A
Descriptive Statistics Tables
Table A-1: Descriptive Statistics of the Plant Level Data Set
1990
Unskilled Workers (head count)
Skilled Workers (head count)
Materials
Capital
Value Added
Gross Output
1995
Unskilled Workers (head count)
Skilled Workers (head count)
Materials
Capital
Value Added
Gross Output
2000
Unskilled Workers (head count)
Skilled Workers (head count)
Materials
Capital
Value Added
Gross Output
2005
Unskilled Workers (head count)
Skilled Workers (head count)
Materials
Capital
Value Added
Gross Output
Total
Unskilled Workers (head count)
Skilled Workers (head count)
Materials
Capital
Value Added
Gross Output
Mean
Standard Deviation
Median
Observation
226
58
89722
1949
28988
121769
475
142
339306
4409
61407
391755
71
13
12500
495
8131
22618
1194
1194
1194
1194
1194
1194
258
61
164188
3499
35417
205730
552
154
635639
10317
76766
717517
80
13
16599
992
10198
28121
1425
1425
1425
1425
1425
1425
327
83
193324
4095
54800
252991
770
202
693715
9347
128617
806776
88
18
15319
1096
12807
31073
1995
1995
1995
1995
1995
1995
345
94
289591
5546
76162
370876
747
213
982330
10827
167276
1092718
104
24
29733
1731
21402
56792
1678
1678
1678
1678
1678
1678
288
75
187436
3867
50703
242821
658
186
724009
9668
123993
818302
82
17
16754
998
12581
31962
27548
27548
27548
27548
27548
27548
Note: Values are expressed in thousand 2002 Mexican peso.
Authors’ calculation.
47
Source: Plant-level Survey of Maquiladoras (INEGI).
Table A-2: Descriptive Statistics: Panel Information
Year
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Number of Plants
1194
1285
1372
1443
1430
1425
1548
1632
1741
1862
1995
2083
1848
1688
1662
1678
1662
Years in the Panel
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Number of Plants
362
59
57
86
103
171
127
158
179
186
222
272
392
406
404
357
228
Note: The left hand side of the table shows the number of plants in a given year. The
right hand side of the table shows the distribution of plants over their length of the stay
in the panel.
Table A-3: Descriptive Statistics: Average SkillIntensity
Number of Employees
1990-2000
2001-2006
Overall
333.549
409.955
Entrants
83.438
87.827
1990-2000
2001-2006
Overall
0.289
0.401
Entrants
0.339
0.609
SkilledW orkers
U nskilledW orkers
Source: Plant-level Survey of Maquiladoras (INEGI). Authors’ calculation.
Table A-4: Decomposition of Aggregate Productivity Growth over 1999-2006
All
HighCHT
Total Growth
Within Plant
0.078
0.100
0.024
0.036
Between Plants
(Within Industry)
0.003
0.002
The decomposition exercise follows the analysis presented in section B.2.
48
Net Entry
Between Sectors
0.012
0.021
0.039
0.040
Table A-5: Evolution of Technology among Maquilas I: Descriptive Statistics
Year
Number of Plants
Average Employment
(Head Count)
Foreign Owned
Capital (%)
1999
2001
2005
589
675
791
933.093
1245.003
800.510
79.822
78.514
82.729
Source: Encuesta Nacional De Empleo, Salarios, Tecnologı̀a Y Capacitaci`n (ENESTYC).
Authors’ calculation.
Table A-6: Evolution of Technology among
Maquilas II: Research and Development
Year
Performed R&D
(%)
Performed Product
Development (%)
1997
1999
2004
38.879
39.704
45.891
20.713
20.593
32.111
Source: Encuesta Nacional De Empleo, Salarios, Tecnologı̀a Y
Capacitaciòn (ENESTyC). Authors’ calculation.
In ENESTyC
1999, R&D questions refers to the year 1997. In ENESTyC 2001,
R&D questions refers to the year 1999 and in ENESTyC 2005,
R&D questions refers to the year 2004.
Table A-7: Evolution of Technology among Maquilas III: Capacity Utilization
Year
1997
Mean
Median
Standard Deviation
Observation
1999
Mean
Median
Standard Deviation
Observation
2004
Mean
Median
Standard Deviation
Observation
Maquila Plants
Percentage Capacity Utilization
Manufacturing Plants (Panel) All Manufacturing Plants
81.3
85
20.0
589
76.0
79
15.2
686
72.3
78
23.7
6806
82.6
85
16.4
675
76.2
80
17.4
690
75.7
80
20.1
8178
85.7
90
16.4
786
76.9
80
17.9
689
76.9
80
18.3
6364
Source: Encuesta Nacional De Empleo, Salarios, Tecnologı̀a Y Capacitaci`n (ENESTyC 1999, 2001, and 2005). Authors’ calculation.
Second column is among non-maquila manufacturing plants that are surveyed continuously by ENESTyC. The third column is among
all manufacturing (non-maquila) plants that are surveyed.
49
Table A-8: Evolution of Technology among
Maquilas IV: Management and Organization Techniques
Year
2000
(%)
2005
(%)
Just in Time
Statistical Process Control
Total Quality Management
Job Rotation
Rearrangement of Equipment
Process Re-engineering
Number of (Maquila) Plants
20.1
41.6
31.3
22.7
26.0
19.6
642
41.1
61.6
60.2
46.8
55.0
45.5
791
Source: Encuesta Nacional De Empleo, Salarios, Tecnologı̀a
Y Capacitaciòn (ENESTYC) 2005.
Authors’ calculation.
The information is derived from ENESTYC 2005, Section
12, question 6 which asks whether a specified technique has
been implemented, and, if so, the starting year of the implementation. The survey also presents brief descriptions of
each technique to prevent recording errors.
B
B.1
Sources and Construction of the Data
Calculation of Plant TFPs
We use a KLEM approach and calculate multi-factor productivity using gross-output measures. Good,
Nadiri and Sickles (1997) discusses the extension of the total factor productivity index that incorporates
both the chaining approach and the hypothetical firm approach of Caves, Christensen and Diewert
(1982) which is suitable for panel data-setting. We construct a hypothetical firm whose subcomponent
expenditure shares are the arithmetic mean expenditure shares and whose subcomponent quantities
are the geometric means of the subcomponent quantities for each cross section. We then chain the
hypothetical firms together over time.
lnT F Pit = (qit − qt ) +
t
X
(qs − q s−1 ) − [
s=2
X
j
0.5 ∗ (αit
+ αtj )(xjit − xjit )+
j=k,sl,ul,e,m
t
X
X
s=2 j=sl,ul,k,e,m
50
0.5 ∗
(B-1)
(αsj
+
j
αs−1
)(xjs
−
xjs−1 )]
where qit is the logarithm of deflated sales of plant i, and xjit is the logarithm of the input j used
by plant i at period t where type of input is indicated with superscript j = sl, ul, k, e, m. sl denotes
skilled labor measured by the total number of administrative and technical personnel, ul denotes
unskilled labor as measured by the total number of workers, k denotes capital measured by the deflated
rental expenditures on buildings, machinery and equipment, e denotes energy measured by deflated
expenditures on fuel and electricity and m denotes materials measured by deflated expenditures on
domestic and imported materials. The bar indicates an average over the relevant variable (e.g. qt
indicates the natural logarithm of the geometric average for output across all plants at period t).
Scale elasticity α’s are calculated using costs shares.
B.2
Productivity Decomposition
Following Griliches and Regev (1995) and Foster et al. (1999), we decompose the aggregate productivity growth into within-plant, reallocation within sector, turnover and reallocation between sectors
components as follows:
∆Pt =
X
j
sjt [
X
sjt ∆Pitj ] +
X
j
i∈Cj
X
j
sjt [
X
sjt [
X
(pji − P j )∆sjit ] +
X
j
i∈Cj
(sjit (pjit − P j )) −
i∈Nj
Ptj ∆sjt +
X
(B-2)
(sjit−1 (pjit−1 − P j ))]
i∈Xj
Here a bar over a variable indicates the average of the variable over the base (t-1) and end year (t), i
denotes plant, j denotes industry. Market shares are measured using export value-added. Pt denotes
P
the aggregate industry productivity, Pt = j sjt Ptj where sjt is the industry j’s share of total export
value-added. Cj refers to a set of continuing plants in sector j, Nj refers to a set of entering plants
P
P
in sector j, and Xj refers to a set of exiting plants in sector j. The first term, j sjt [ i∈Cj sjt ∆Pitj ]
denotes the within-plant component, the second term is within-industry reallocation term, the third
term is reallocation between sector followed by the net entry component.
B.3
Matching NAICS with Maquiladoras
INEGI has conducted an annual survey of the universe of plants registered under the maquiladora
program and constructed one service and eleven manufacturing sectors, called ’economic group’. In this
paper, the 11 maquiladora industries were matched with the US NAICS. To do so, we use survey results
conducted by INEGI, Dirección De Estadı́sticas De Comercio Exterior, Registros Administrativos Y
51
Precios. Maquiladora sectors are tied to the US industries vertically within generally 4-digit industries
(plants import from and export to the same 4-digit industries). In Table B-1, we provide the names of
the eleven maquiladora sectors and the corresponding US NAICS. We further confirm this matching
by converting export trade data in 2-digit HS as reported by Banco de México to 3-digit NAICS.
B.4
Import Penetration Rates and other Aggregate Variables
To calculate import penetration in the U.S., data from the Center for International Data at U.C. Davis
on exports and imports by industry and country were used. The information is provided in 6-digit
NAICS classification. Output information is provided by the Bureau of Economic Analysis (BEA).
The matching is based on the combination of 4-digits NAICS as described in Table B-1.
The source for the MFN tariff average rates for the US industries is the Trade, Production and Protection database of the World Bank. The source for the world-wide Chinese imports as well as world-wide
total imports is the World Bank. U.S. hourly wages data is from the Bureau of Labor Statistics(BLS).
To estimate maquiladora hourly wages, total worker wages reported by the plant-level data set were
used. Since they are in Mexican pesos they were divided by the nominal exchange rate and aggregated
to industry-level. The source for the nominal exchange rate between the US dollar and Mexican peso is
Banco de México. Relative wages is then constructed by dividing Mexican hourly wage by the their U.S.
industry counterpart. Both US and maquiladora wages do not include benefits to workers. US wages
are based on 3-digit NAICS matching. The source for U.S. industrial production index is the Federal
Reserve Board of Governors. Except the first industry, ECOGROUP 1, which only corresponds to a
single 4-digit NAICS, matching of the aggregate variables except the import-penetration rate is based
on 3-digit NAICS. We use industry-specific exchange rate measures for the US manufacturing industries
constructed by Linda Goldberg. The data can be downloaded from http://www.newyorkfed.org/
research/global_economy/industry_specific_exrates.html These measures are constructed by
using the time histories of the weights of U.S. trading partners in the imports of each U.S. industry.
52
53
Description
Selection, preparation, packing and canning of food
Apparel and textile knitting and sewing
Footwear manufacturing and leather and hide tanning
Furniture and other wood and metal products assembly
Chemical products
Transportation equipment (and accessories) assembly
Assembly and repair of tools, equipment and parts, except electrical
Electronic (and electrical) devices, machinery, accessories
Electrical machinery, equipment and accessories
Sporting goods and toy assembly
Other manufacturing industries
NAICS Code
3114
3151, 3152, 3159, 3169
3161, 3162
3323, 3371,3379
3251, 3252, 3253, 3254, 3255, 3256, 3259
3362, 3363, 3369
3331, 3332, 3334, 3339
3341, 3342, 3343, 3344, 3345, 3346, 3352
3351, 3353, 3359
339920, 339931, 339932
334510, 3391, 3399, 339992 (excluding 339920-339931-339932)
Source: DEPARTMENT OF EXTERIOR COMMERCIAL STATISTICS, ADMINISTRATIVE AND PRICING REGISTRY, INEGI
Ecogroup No
1
2
3
4
5
6
7
8
9
10
11
Table B-1: Industry Descriptions