Causes of Growth and Decline in Mexico’s Maquiladora Apparel Sector

Abstract

Under the North American Free Trade Agreement (NAFTA) beginning in 1994, the maquiladora sector was the dynamic manufacturing sector in Mexico, and its apparel subsector was especially so, more than quadrupling in employment from December 1993 to July 2000. Yet NAFTA’s influence on apparel employment is hard to find in a careful time series econometric analysis. Instead, much of employment growth is explained for 1980 to the present by changes in US demand as measured by real US gross domestic product or by real US apparel spending, by US/Mexico relative labor cost as proxied by the real peso‐dollar exchange rate, and by the relaxation of quotas on US apparel imports from Mexico in 1988–1990. In equations including these variables, tests for a structural break at the time of NAFTA find an effect which is either insignificant or else quite small and in some models negative. Possible explanations for this surprising result are discussed, along with the implications for cost–benefit analysis of free trade agreements.

Key Words:

• Employment
• exchange rate
• apparel
• liberalization
• NAFTA
• quotas

Jel Classifications:

• F15
• F23
• O24

Acknowledgement

I gratefully acknowledge helpful conversations about this paper with Bernie Morzuch, as well as useful comments from an anonymous referee. Any errors that remain are solely my responsibility.

Notes

1. TSUS stands for Tariff System of the United States. A predecessor program, narrower in scope and with somewhat different provisions, had existed since the 1930s, however.

2. However, because apparel quotas are by category, not an overall limit on physical (square meter equivalents) or dollar apparel imports, quotas could affect the type of clothing produced, more than the total quantity.

3. Hansen (2001) explains that in searching for an unknown break date we must compare the F‐test statistic to different and higher set of critical values as published in Andrews (1993). The 1990:III peak falls slightly short of meeting that test. However, for each date the null is that there is no break at that date, and the alternative hypothesis is that there is a break at that precise date. Our purpose here is not to find the precisely exact break date, but to find the approximately correct break date, producing one or more good models. Then, if a break at the time of NAFTA is not already included in such models, we test for the presence of a NAFTA break. To clarify: NAFTA was not left to the end, but was tested for at early, middle and late stages of the analysis. The presentation in the tables merely illustrates the insignificance of NAFTA dummies in several especially clear‐cut cases.

4. Adjustments in break dates were based on exploring possible breaks at months near 1990:9 and 2000:9, especially in order to reduce serial correlation in the error terms.