3.1 Global Wage Arbitrage: A Determinant of Industry Migrations

Technological developments during and after the Second World War made it possible to codify and digitalize tasks, bringing down the costs of transportation and drastically lowering that of communications. Trade liberalization created the right conditions for multinational companies to capitalize on international wage disparities.

According to the ^{United Nations Organization for Industrial Development (2013}, cited in the International Labor Organization, 2014), worldwide, apparel industry wages are on average 35% lower than average manufacturing industry wages. The vast majority of developing countries that were among the top 20 exporters for the apparel industry around the globe in 2014 (China, Bangladesh, Vietnam, India, Indonesia, Cambodia, Sri Lanka, and Malaysia) are those countries with the lowest minimum monthly wages in the industry (^{Organización Internacional del Trabajo, 2014}).

In order to keep labor costs competitive, companies may violate certain provisions of local laws, as demonstrated by ^{Vaughan-Whitehead (2014)} in a survey administered to 122 companies in 10 Asian countries. Some of the most severe labor irregularities included: a) 20% of the companies surveyed paid wages below the legal minimum; b) 43% of the companies did not offer any social security to their workers; c) employees worked on average 71 hours a week, when the legal maximum permitted is 60; and d) 61% of companies did not pay overtime pursuant to local laws.

In order to commit these infractions and avoid penalties by the authorities, companies resort to keeping two sets of books and informal employment. ^{Vaughan-Whitehead (2014)} found that 48% of companies surveyed use this double recording mechanism to hide overtime in the work day, and pay less in overtime wages than what is legally required. Moreover, the need for greater flexibility and pressures on labor costs have led to an increase in informal employment in the textile, apparel, leather, and footwear industries (^{Organización Internacional del Trabajo, 2014}).

^{Vaughan-Whitehead (2014)} asserts that for the majority of clothing suppliers in China, migrant workers represent between 70% and 100% of the total workforce, and that companies shield themselves behind the Hukou system to avoid providing them with social security. In case studies of three Chinese companies, Vaughan-Whitehead (2014) found that they paid wages slightly above the legal monthly minimum for total number of hours worked, which were far above 40 a week (up to 32 more).

Comparing wages in Mexico with those of the top Asian apparel industry exporters, it turns out that the Mexican minimum wage in 2001 was five times higher than that of China and more than eight times higher than that of Bangladesh. As of 2011, wages in the industry had fallen 29% in Mexico, to become competitive with some of its Asian peers. However, in spite of this decline, in the same year, Mexican wages were 1.6 times higher than those in China (^{Worker Rights Consortium, 2013}).

Finally, even though labor costs are the main determinant behind the relocation of the apparel industry, in practically all African countries, low wage levels have not automatically translated into a comparative salary advantage for apparel industry manufacturers, because these countries still face serious limitations in terms of the availability and cost of key services for the industry, like transportation, work skills, and a stable business climate (Fukinishi et al., 2013). The foregoing means that low labor costs are a necessary but insufficient condition for the relocation of the industry.

3.2 Trade Agreements: Understanding the Rise and Fall of Mexican Exports³

In order to protect local companies in developed economies from cheap imports from the developing world, between 1961 and 1994, international trade in the apparel industry was constrained by trade agreements. However, starting in 1995, the MFA was supplanted by the ATC, whose objective was to do away with international trade quotas in the textile and apparel industry over a ten-year period (^{Frederick and Gereffi, 2011}; ^{Fernández-Stark et al., 2011}; ^{World Trade Organization, 1996}; ^{Martin, 2007}).

Even with the quota system, imports from Asia continued to pose a serious challenge to local American companies, which, in response, essentially resorted to two strategies: offshoring and outsourcing. Brand manufacturers used offshoring to install subsidiaries for the most-labor intensive production processes in export processing zones, located primarily in Mexico and the Caribbean Basin, and took advantage of Rules 807/9802,⁴ and, subsequently, free trade agreements, to re-export production, free of tariffs, to the United States (^{Cepal, 1998}).

On the other hand, retailers and brand sellers opted for the outsourcing strategy, with full-package producers located principally in Asia (^{Appelbaum and Gereffi, 1994}, cited in ^{Bair and Dussel, 2006}). With the passing of time, brand manufacturers also began to use the outsourcing strategy, abandoning production altogether (^{Bair and Gereffi, 2003}; ^{Cepal, 1998}).

In light of the threat of growing clothing imports and the shift away from production in brand manufacturers, fabric manufacturers began to capitalize on the growing demand for full-package services, by way of forward vertical integration. The idea was that this strategy would garner sales with customers who would otherwise subcontract their production in Asia, with producers who would very likely not use American textiles (^{Bair and Gereffi, 2002}, cited in ^{Bair and Dussel, 2006}).

The possibility for vertical integration with Mexican manufacturers to revert the long-term downward trend in production, and allow them to come out on top as major players, made American textile companies, such as Burlington Industries, Guilford Mills, Galey & Lord, Cone Mills, and Dan River, strong supporters of the North American (NAFTA) and Central American (^{UNCTAD, 2005}) free trade agreements.

The forward vertical integration of American textile companies came about through the establishment of subsidiaries, acquisitions, and joint venture investments with Mexican companies (^{Cutchin et al., 1999}). However, the strategy did not work out as expected, and in just a few years, practically all of the major textile companies that offered the full package abandoned this approach. Not only that, but also, between 2001 and 2005, many of the large companies that had invested in Mexico went bust (^{Bair and Dussel, 2006}).

Rules 807/9802, the export processing zones, and the North American Free Trade Agreement (NAFTA) all endowed Mexico with access to a market protected by the quota system, the United States, which allowed the country to rapidly increase clothing exports. Between 1990 and 2000, Mexico's share in global exports skyrocketed (see Figure 1), going from 0.5% in 1990 to 4.4% in 2000 (see Appendix). As exports rose, so too did the number of jobs and factories for assembling clothing items in the maquiladora export industry. The number of factories went from 277 in 1990 to 1,088 by the year 2000, with the number of employees going from 42,464 to 282,755, respectively (^{INEGI, 1990-2006}).

Source: Created by the authors based on data from the ^{World Trade Organization (2015}, ²⁰¹⁶), ^{Martin (2007)}.

Figure 1. Share of Mexican Exports in the Apparel Industry and its Relationship to Trade Agreements 

However, the year 2000 marked a change in trend in the share of Mexican exports out of global exports, mainly due to the elimination of the quota system, China's joining the World Trade Organization (WTO), and Article 303 of NAFTA. Due to the advancement of the ATC, by 2002, quotas had already been removed for 51% of the international clothing and textiles trade, gradually reducing the protections afforded by the quota system and the advantages that Rules 807/9802 and NAFTA had given Mexico in the 1990s. China's accession to the WTO (December 11, 2001) intensified competition between this country and the American market. When Article 303 of NAFTA took effect in 2000, taxes on maquiladora factories were raised, considering them to be national enterprises, and chipping away at their competitiveness (^{Carrillo, 2007}).

The voiding of the quota system, as well as the other preceding factors mentioned, caused Mexico to sink from the fourth-highest spot on the list of top industry exporters in 2000 to being ranked twenty-first in 2014 (see Appendix). Mexico's falling share of global exports led to the closure of factories and destroyed apparel industry jobs. In 2000, Mexico was home to 1,088 clothing assembly factories in the maquiladora export industry.⁵ By 2006, only 484 remained. Jobs also fell by 40% in the same time period, from 282,755 employees in 2000 to 169,490 in 2006 (^{INEGI, 1990-2006}).

Losses in the Mexican piece of the global apparel industry export pie, and bankruptcies for many American textile companies with plants in Mexico, in the wake of the implementation of the ATC, made evident that the competitiveness of the Mexican apparel industry was based on neither wages nor productivity, but rather on trade agreements.

3.3 Fashion, Quality, and Production Volumes

Several factors can help explain why certain production niches are still held by high-wage countries: production volumes are very low, production is high quality, requiring direct collaboration between designers, producers, and retailers, and the goods are very sensitive to production times (^{Doeringer and Crean, 2006}).

The first group of these niches is composed of high-quality products, fashion for the high-end market, made-to-order clothing items, small fashion product orders for regional chains and independent stores, and products that use exotic fabrics that require special care and skilled labor. The second set of niches comprises products directed at specialized mid- and low-range markets, principally, surgical wear, big and tall sizes, uniforms, emergency orders of basic fashion products when production abroad has fallen behind, repairs of products damaged during shipping, modification of products that are no longer selling, etc. (^{Doeringer and Crean, 2006}). The third consists of products known as fast fashion.

Fast fashion is a concept developed in Europe to sell to teenagers and young women. The business model combines the following elements: a) rapid response; b) frequent changes in the assortment; c) fashion designs at accessible prices (^{Caro and Martínez-de-Albéniz, 2015}). Fast fashion is characterized by its reliance on local production or production in nearby countries, which has been a lifesaver for the industry in developed countries (^{Abernathy et al., 2006}; ^{Doeringer and Crean, 2006}, cited in Caro and Martínez-de-Albéniz, 2015). The model permits a reduction in excessive inventory of finished product, by postponing purchasing decisions until there is enough evidence as to the behavior of demand, even if it means that production costs are higher (Caro and Martínez-de-Albéniz, 2015).

The empirical evidence shows that the prices of exports in the apparel industry in developed countries are on average higher than the prices of their imports. For example, in 2014, the average price of Germany's imports from the rest of the world was 28 dollars per kilogram, while the price of German exports was 42 dollars. This pattern holds true that year for all of the countries selected (see Table 1). This would seem to suggest that the apparel industry in developed countries has become concentrated on serving market niches where the competition does not depend on labor costs, but rather on the aforementioned characteristics, which allows them to produce and export goods that are on average more expensive than those they import.

3.4 Economies of Scale, Infrastructure, and Governmental Support

Besides competitive labor costs and preferential trade agreements, the location of the apparel industry in developing countries is also affected by the economies of scale, infrastructure, and tax incentives offered by host countries to investors.

The Chinese export market and domestic market (the latter absorbing more than half of the industry's local production) (^{Gereffi and Frederick, 2010}, cited in Fukinishi et al., 2013) has enabled the Chinese apparel industry to capitalize in economies of scale in two ways (^{Appelbaum, 2008}; ^{Tewari, 2006}; and ^{Gereffi, 2009}, cited in Frederick and Gereffi, 2011): a) by creating clusters revolving around a large vertically integrated company; and b) by creating product clusters. The Chinese government has fostered the creation of both types of clusters through fiscal incentives, and by building ports and roads.

Transportation and communications infrastructure plays an essential role in the competitiveness of the apparel industry. China has boosted exports and reduced time-to-market by building and improving roads and ports and developing clusters in coastal cities (^{Frederick and Gereffi, 2011}; ^{Tewari, 2006}, cited in Frederick and Gereffi, 2011).

Compared to China, Mexico has faced serious disadvantages in the apparel industry. First, in spite of sharing a low-wage policy, salaries in the Mexican apparel industry in 2011 continued to be higher than those of the Asian nation (^{Worker Rights Consortium, 2013}), Second, contrary to what China has done, Mexico has not made significant investments in transportation infrastructure or logistics to improve its competitiveness (^{Frederick and Gereffi, 2011}; ^{Robinson, 2010}, cited in Frederick and Gereffi, 2011).

Another factor to consider is the exchange rate. In 1994-1995, the Mexican peso suffered a sharp devaluation, which incentivized the growth of exports in the industry, and led more maquiladora factories to sprout up in the country (^{Cepal, 1998}; ^{Gereffi and Bair, 1998}). However, by 2000, the costs of maquiladora production in Mexico had risen due to, among other factors, the overvaluation of the Mexican peso against the dollar (^{Carrillo, 2007}).

Finally, the following aspects must be considered in determining the location of apparel industry production: the business climate, quality and on-time deliveries, reliability, availability of qualified labor and productivity, access to quality inputs at competitive prices, capacity of manufacturers to offer full-package services, etc. (^{UNCTAD, 2005}; ^{Freund and Wallace, 2004}).

3.5 Technology: Could it be a Central Factor in the Location of Production?

According to ^{Book et al. (2010)}, the need for automation in the apparel industry has been widely acknowledged by many since the early 1980s. For that reason, during that decade, the United States, Japan, and Europe spent millions of dollars on research programs to automate the apparel manufacturing industry. Although none of these programs managed to fully automate production, some advances were made (^{Byrne, 1995}).

Even though automation has made progress in the design, patterning, and cutting phases, not enough has been done to make it competitive with cheap labor in developing countries. The foregoing becomes evident with a look at the growing share of developing nations, with their limited use of technology, in global apparel industry exports (see Appendix). According to ^{Bhavani and Tendulkar (2001}, cited in ^{Yunus and Yamagata, 2012}), production in the developing world is done through processes that combine various general and specialized machines with manual, mechanical, and electronic systems operated by skilled and unskilled workers.

Greater evidence as to the limited use of technology in developing countries comes from the industry's top exporter (China). Pursuant to ^{Zhang et al. (2016)}, the rate of adoption for mechatronic equipment (industrial robots) and digital equipment in the Chinese apparel industry has not risen significantly; storage equipment is inadequate given the high level of production, and design is still done primarily on paper samples.

Mexico is another clear example of a country with limited use of technology in this industry (^{González, 2006}). According to the National Employment, Salaries, Technology, and Training in the Manufacturing Sector Survey (ENESTYC, 2005; ^{INEGI STPS, 2005}), in 2004, only 31.39% of apparel industry companies made use of automatic or numerically controlled equipment (computerized or not), and 0.04% used robots; the rest used only manual equipment (39.56%) and tool machines (29.02%). Compared to 2000, the use of advanced technology has fallen. In that year, 32.38% of companies were using automatic or numerically controlled equipment, and just 0.13% robots (^{INEGI. STPS, 2001}). Because a more updated version of the ENESTYC has not been released, and neither have similar surveys, the evolution of technology usage in the Mexican apparel industry is approximated via the change in the capital/labor ratio in this sector between 2003 and 2013.

According to data from the 2004 and 2014 economic censuses, the apparel industry has shrunk in size, measured by added value (-44%) and number of people employed (-32%). The total stock of industrial machinery and equipment has also been reduced, although not to the same degree as the other variables mentioned, so the capital/labor ratio rose (35%) between 2003 and 2013. However, this greater intensity in capital use was not reflected in a rise in productivity. On the contrary, productivity fell by 17% (see Table 2).

There are at least three reasons why in spite of the increase in the capital/labor ratio, productivity could have declined in the apparel industry: 1) greater idle capacity installed; 2) lower added value goods were produced; and 3) workers were not sufficiently trained to use new machines.

Higher idle capacity can be ruled out. According to the Monthly Industrial Survey for 2003 (^{INEGI, 2003}), the apparel industry reported utilized capacity of 77.1%, and the Monthly Manufacturing Industry Survey (^{INEGI, 2013a}) indicated that the capacity utilized in the industry in 2013 was 81.6%. To verify that the industry is producing goods with lower added value, its share in total gross production for 2003 and 2013 was estimated, and it was found that the figure went from 50% in 2003 to 34% in 2013 (see Table 2).

Finally, if the increase in the capital/labor ratio were due to the incorporation of more advanced technology, a reduction in the share of workers in total employment would be expected. However, the share of workers in total employed persons rose marginally during the time period, going from 92.8% of workers employed by a company in 2003 to 93.2% in 2013 (information broken down for personnel not employed by a company is not available, although their share in total industry employment continues to be low, 10% in 2013) (^{INEGI, 2013b}). The empirical evidence suggests that falling productivity is principally due to the fact that the industry is producing goods with lower added value.

Considering the use of advanced/automated technology in the apparel industry in 2003 (15.2% cutting, 6.5% sewing, 13% ironing, and 1.4% air transportation systems) (^{CEC-ITAM, 2008}), the rise of 35% in the capital/labor ratio between 2003 and 2013, and the 17% decline in labor productivity, we believe it is very likely that an increase in the degree of automation in the apparel industry in 2003-2013 would have been insignificant, making Mexico a clear example of the low automation of this industry in developing countries.

The scant penetration of new technologies in the apparel industry is not exclusive to developing countries, though, in particular when it comes to industrial robots, one of the technologies that has made the greatest leaps forward in the manufacturing industry. In 2013, globally, the manufacturing industry had 1.1 million industrial robots in operation, 26% more than in 2008, while in the same time period, the textile, leather, and clothing industries together went from 1,127 to 1,228 robots (^{IFR, 2015}). Compared to industries like the automotive, electrical-electronics, metal, and plastics and chemical products sectors, the textile, leather, and clothing industries are lagging extremely behind in their usage of industrial robots, and furthermore, there is no foreseeable change in this trend.

Even though according to ^{Byrne (1995)}, new technologies have managed to permeate the majority of stages/tasks in the apparel industry production process, the cost and flexibility of labor have gotten in the way of widespread adoption. The highest barrier to the penetration of new technologies resides in the sewing phase, because the heterogeneity of cuts and the flexibility of fabrics still present challenges for some automated equipment (^{Book et al., 2010}; ^{Paraskevi, 2012}). ^{Michelini and Razzoli (2013)} assert that the sewing process can be automated for mass production, but that the use of this technology is not compatible with the economies of scope that characterize the current markets.