Did NAFTA Increase Labor Market Integration
between the United States and Mexico?
Raymond Robertson
Department of Economics
Macalester College
1600 Grand Ave.
St. Paul, MN 55105
robertson@macalester.edu
Abstract: Has NAFTA led to wage convergence between the United States and Mexico?
Data from Mexican and United States household surveys finds little evidence of
increased labor market integration following the North American Free Trade Agreement.
Wages diverged when the peso crisis precipitated a drop in Mexican wages, but
alternative measures of labor market integration, which control for the peso crisis, find
little evidence of increased labor market integration.
The author thanks Bill Maloney and the NAFTA Brainstorming Workshop participants for valuable
comments and suggestions. Financial support from the World Bank is gratefully acknowledged. Jason
Heil provided excellent research assistance. Any errors are the responsibility of the author.
1
Did NAFTA increase labor market integration between Mexico and the United
States? Labor market integration has been one of the primary hopes and fears of the
North American Free Trade Agreement (NAFTA). This paper uses household-level
surveys and three different measures of integration to evaluate labor market integration
before and after NAFTA. While the question of integration is interesting on its own, this
paper also contributes to the debate about the mechanisms that integrate labor markets.
Trade, capital flows, and migration can all integrate labor markets. NAFTA liberalized
trade and investment, but did not liberalize migration. Therefore, the extent of detectable
labor market integration following NAFTA may help us understand the relative
importance of these factors.
The relative contributions of trade, migration, and capital flows in integrating
labor markets have created significant debate in economic literature.1 Williamson (1996)
uses real wages, GDP per worker-hour and GDP per capita as indicators of wage, and he
finds that the main cause of convergence is migration.2 Due to the elevated labor content
of migrants, GDP per capita is not affected in the same level as GDP per worker-hour or
real wages. Although mentioned in the study, changes in productivity in the form of
factor-saving improvements, commodity-price convergence, and schooling play a less
significant role. Williamson’s study observes three periods: late 19th century (18701913), 1914-1950 (Great War), and late 20th century. The first and third periods show
convergence while the second period is associated with divergence (and a movement
1
See, for example, O’Rourke and Williamson (1994), O’Rourke, Taylor, and Williamson (1996),
Saint-Paul, Gilles (1997, 1999).
2
The correlation is weaker when capital is allowed to “chase” labor in the computable general
equilibrium (CGE) models.
towards autarky).
Mokhtari and Rassekh (1989) employ another measure of convergence: the
coefficient of variation of wages (CVW). They use the mean and standard deviation of
the ratio of nominal manufacturing wages to the Consumer Price Index to construct the
CVW. As a result, commodity-price convergence (using openness as a proxy) shows a
strong impact on factor price equalization for the period (1961-1984) observed. While
the coefficient of variation of wages across countries is not as relevant in the two-country
case, the findings of a significant role for trade openness in explaining wage convergence
is very relevant for this study.
The contrast between these papers illustrates the value of studying the NAFTA
case. NAFTA was notable because it greatly facilitated capital and trade flows, but it
explicitly did not include liberalization on labor migration. If anything, migration
controls became stricter following NAFTA through policy changes in border
enforcement. Thus, finding increased labor market integration could suggest the
importance of trade and capital flows vis a vis migration, while a finding of little or no
increase in integration may suggest that migration plays a more dominant role in
integrating North American labor markets.
This paper presents some preliminary evidence on the degree of labor market
integration in North America before and after the NAFTA. Labor market integration can
be measured in a number of ways. This paper focuses on three criteria: absolute wage
convergence, the responsiveness of Mexican wages to United States wage “shocks,” and
the rate of convergence of United States and Mexican wages. Convergence of absolute
wage levels is a popular measure of labor market integration. The Mexican peso crisis of
December 1994, which was probably not related to NAFTA provisions, caused Mexican
2
wages, in dollar terms, to fall drastically. By 2001, wages had barely recovered to their
1994 levels. Given the peso-induced drop in the absolute value of wages, it is not
surprising that the wage gap between Mexico and the United States increased. The wage
growth rate, while lower between 1995-2001 period than during the 1987-1994 period,
increased in 1998. The post 1998 wage growth rate is the highest in the period, perhaps
suggesting that faster wage growth, and thus convergence, is still a possibility.
An alternative measure of labor market integration follows from an errorcorrections framework. If wages in two integrated regions exhibit a long-run
relationship, then wages should exhibit two other measurable behaviors. First, wages
shocks in one region would be transmitted to wages in the other region. Second, wages
should converge to their long-run equilibrium differential. In the case of the United
States and Mexico, shocks from the United States should be detectable in Mexico if the
two regions are integrated. In the short run, while the difference between the U.S. and
Mexican wages is constant, wages that are moved from the equilibrium differential
should move in order to return to that equilibrium differential. Robertson (2000)
observes these patterns between the United States and Mexico, suggesting that labor
markets in the two countries are integrated, although the difference between average
wage levels is very large.3
In the following sections, we apply these criteria to evaluate economic integration
between Mexico and the United States before and after NAFTA. In the first section, we
describe the economic and theoretical rationale for using absolute wage convergence as a
measure of integration and present some empirical results. In the second section we
3
In 2001, the United States GDP per capita was US$36,200 and the Mexican GDP per capita was
$9,100.
3
discuss the error corrections approach. In this section we also discuss robustness checks
and alternative wage comparisons. Specifically, we discuss the effects of exchange rate
movements and whether wage integration differs by industry. We conclude in the final
section.
II. Measuring Labor Market Integration: Absolute Wage Convergence
A. Theory and Motivation
Markets are generally considered to be integrated when prices converge. Thus,
the most intuitive measure of labor market integration is the equalization of wages. This
approach has been used widely in the literature. This metric is effective when discussing
the removal of barriers to migration (such as the fall of the Berlin Wall) or in areas in
which barriers to migration are relatively small (Boyer and Hatton 1994, Moazzami 1997,
Collins 1999). Reynolds (1992) summarizes the expectation for wage convergence in
the context of NAFTA. Trade theory suggests that wages, which are determined by
supply, demand, productivity, government policies, and other factors, could equalize as a
result of trade liberalization. The main reasons for this are well known to trade
economists and therefore will only be briefly described here. The first reason is that
differences in wages give rise to trade. In a two-factor, two-country, two-good model,
wages will differ across countries when relative factor supplies differ. Wages are high in
the labor-scarce country and low in labor-abundant countries. The increase in the
demand for the (cheaper) labor-intensive goods from the labor-abundant country should
increase the demand for labor and, therefore, increase wages. The reverse should happen
in the labor-scarce country, generating a trend towards convergence in absolute levels.
4
In the context of the United States and Mexico, it seems unlikely that trade with
Mexico would have a significant effect on U.S. wages because Mexico is just one of
many trading partners4 and many of these other trading partners have much lower wages
than Mexico (e.g. China). Mexico, on the other hand, could have much larger effects
from trade liberalization because 77.1% of Mexico’s total trade is with the United States5
and, therefore, tariff liberalization could affect a much wider range of production in
Mexico. To the extent that trade liberalization results in production shifts that increase
the demand for labor, we should expect to see production shifting into labor-intensive
activities.
Figure 1 shows the evolution of Mexico’s manufacturing exports between 1991
and 2002. Exports rise throughout the sample period, but increase at a higher rate
following NAFTA’s implementation (even before the peso crisis). This increase in
exports need not correspond to an increase in the share of workers working in
manufacturing, however, because exports could increase if goods formally sold
domestically were simply shifted abroad.
To briefly consider the shift in production before and after NAFTA, we draw
upon data from the quarterly Encuesta Nacional de Empleo Urbano (National Urban
Employment Survey, or ENEU ). These data are available from the first quarter of 1987
to the last quarter of 2001. The surveys are conducted in municipalities throughout
Mexico and are used to construct unemployment statistics. From these data, I
constructed the share of urban employment in manufacturing in each quarter. To begin, I
restrict the sample to workers between 16 and 70 years who are not self-employed,
4
Mexico’s share of total U.S. goods trade in March 2002 was 12.7%. In that same month, the top
ten trading partners together made up about 70% of total U.S. goods trade.
5
working in the public sector, or have missing earnings. To be consistent with later work,
I also focus on five urban areas: Central Mexico (Mexico City and Mexico State) and
four border cities (Tijuana, Ciudad Juarez, Nuevo Laredo, and Matamoros). Both males
and females are included.
The share of these workers employed in manufacturing over time is found in
Figure 2. The fraction of workers employed in manufacturing falls steadily between
1989 and 1994, when NAFTA went into effect. In contrast, between January 1, 1994 (the
data NAFTA became effective) and the peso crisis (December 1994), the share of
manufacturing employment rose. It falls somewhat during 1995 (the worst period of the
crisis) but continues to climb until 1999, when it again begins to fall. Overall, however,
the share of employment in manufacturing rose in Mexico following NAFTA.
Shifting into manufacturing only increases the demand for labor if manufacturing
is labor intensive. Of course, manufactured goods are produced with a wide range of
labor-intensive techniques. Textiles is one of the most labor-intensive industries within
Mexico. Figure 3 shows the evolution of the share of urban employment in the textile
industry. As with manufacturing, the share employed in textiles falls until NAFTA. The
peso crisis reverses the increase until the beginning of the recovery in 1998. The share of
employment in textiles increases following NAFTA when the entire period (1994-2001)
is considered, although the share seems to fall near the end of the sample.
Figures 1-3 provide some circumstantial evidence that NAFTA did indeed lead to
the kind of changes that theory suggests may lead to absolute wage convergence.
Industrial employment shifted towards manufacturing in general and textiles in particular
following NAFTA. Do Mexican wages show evidence of convergence towards United
5
Based on March 2002 statistics. See http://dgcnesyp.inegi.gob.mx/bdine/bancos.htm .
6
States wages following NAFTA?
B. Empirical Evidence
To formally address this question, I turn again to the ENEU surveys. The survey
design is much like the U.S. Current Population Surveys (CPS). One difference is that
the ENEU follows households for five quarters before dropping them from the sample.
While matching individuals across periods may seem desirable, in practice it is very
difficult because households, and not individuals, are revisited. To approximate the panel
effect, I instead employ the pseudo panel approach described by Deaton (1985). This
approach generates wage averages for different groups in the population and tracks the
wages of these groups over time. This approach has the advantage of allowing us to
follow “individuals” over the entire sample period.
To implement the pseudo-panel technique, I classify workers into 48 groups that
are identified with six education levels and eight age groupings. The education levels are
based on the continuous years of education variable, which is then sorted into the
following categories: fewer than 5 years, 5 to 6 years, 7 to 9 years, 10 years, 11 years,
and 12 or more years. The age groups are 5 years apart starting at 15, except for the last
group. The last group includes workers between 50 and 65 in the first year. As is
customary in the pseudo-panel technique, the age group boundaries advance through time
so that we follow cohorts. Using the United States CPS, I calculate the average wages for
the same age-education groups in the United States and then pair these cohorts with their
demographic counterparts in Mexico.
Figure 4 contains the evolution of the average ratio of log dollar wages of these
groups (weighted by sample size) for Tijuana, Matamoros, and Mexico City relative to
7
their United States counterparts. These cities are chosen because they contrast in ways
that help us understand integration. Tijuana and Matamoros are border cities that are
more closely integrated with the United States labor markets than Mexican interior cities
are (Robertson 2000). Tijuana has a larger number of maquiladora plants than
Matamoros, and also is closer to a major United States city. Not surprisingly, the United
States border patrol apprehends many more people trying to enter the United States
illegally in Tijuana (Hanson, Robertson, and Spilimbergo 2002) than in other parts of the
border.
At first glance, figure 4 shows two important features. First, prior to the NAFTA
and the peso crisis, the ratio of Mexican wages to United States wages was increasing,
suggesting absolute wage convergence. It is very important to mention at this point that
the Mexican real exchange rate was also appreciating at this time: the nominal exchange
rate was not adjusting to offset Mexican inflation. Thus, wages increased in dollar terms
even if they did not increase relative to Mexican inflation. This changed dramatically
when the peso crashed in December 1994. The adjustment in the nominal value of the
peso resulted in a sharp drop in the dollar value of Mexican wages. Following NAFTA, a
very similar phenomenon has occurred with the Mexican exchange rate: the nominal
value of the exchange rate has not adjusted to offset the differences in inflation between
Mexico and the United States so that the dollar value of Mexican wages has been
increasing relative to United States wages.
The second main feature evident in figure 4 is the difference in both the absolute
and relative wage changes between regions within Mexico. The Mexican border region
has higher wages than the Mexican interior, and the wages in Tijuana are higher, on
average, than the wages in Matamoros. Furthermore, during the peso crisis, the wages in
8
the border region in general, and most evident in Tijuana in particular, did not fall as
much as wages in the interior. Wages in Tijuana also recover more quickly from the
crisis than wages in the interior.
The lack of adjustment in the nominal exchange rate makes wage comparisons
using the nominal exchange rate difficult. An alternative approach is demonstrated in
figures 5a and 5b. Rather than changing Mexican wages to dollars using the nominal
exchange rate, I calculate real wages in the United States and Mexico using the consumer
price index of each country (using 1992 as the base year). I then divide each series by the
average of the 1992 values. This creates a wage index for each country that is in the same
units and therefore can be compared directly. Figure 5a shows the evolution of the ratio
of the Mexican (normalized) wages (for Tijuana and Mexico City) to United States
(normalized) wages.
Prior to NAFTA, the wages in the interior (Mexico City) rose more quickly than
in the border region. As shown before, the drop in wages is larger due to the peso crisis
in the interior. The recovery, or wage growth following NAFTA, is more rapid in
Tijuana. Using trend breaks in a linear time trend confirms these differences. Prior to
NAFTA, the trend coefficient on Mexico City’s wage values is 0.016, while for Tijuana
this coefficient is 0.011. Following NAFTA, Mexico City’s trend coefficient is 0.029
and Tijuana’s is 0.031 (a statistically significant difference). Figure 5b shows the
estimated trend lines for each series. The difference before NAFTA is apparent, and,
following NAFTA the trends look very similar.
9
III. An Error Corrections Approach
A. Theory and Motivation
One problem with using absolute wage convergence as a metric of integration
between the United States and Mexico is that the United States government dedicates
considerable resources to preventing migration from Mexico.6 One reason often given
for patrolling the border is that such border enforcement is necessary to maintain the
wage differential. If the observed differential is constant, and the previous section
suggests that, over the entire period, there was little change in the wage gap, then an error
corrections approach generates two alternative measures of integration.
The basic estimation equation in this approach is
∆wmex
= β 0 + β1∆wusjt + β 2 ( wmex − wus ) jt −1 + et .
jt
(1)
This equation is derived from a supply and demand framework more completely
described by Robertson (2000). The equation describes the relationship between the
change in the Mexican log wage for group j (defined here as an age-education group) at
time t , the change in the log wage in the United States, and the difference between the
U.S. and Mexican log wage levels (the error correction term).
In this approach, the β1 term captures the effect of the U.S. wage shock on
Mexican wages, and β2 measures the rate of convergence back to the equilibrium
differential. Larger differences between the two suggest a faster rate of return to the
equilibrium differential. These two coefficients are measures of integration. More
6
According to the 2000 INS Statistical Yearbook, the INS apprehended 1,814,729 aliens in fiscal
year 2000. Of these, approximately 96 percent were Mexicans.
10
integrated labor markets should have larger coefficients (in absolute value). Robertson
(2000), for example, shows that wages in Tijuana are more responsive to shocks and have
a faster rate of convergence to the equilibrium differential than wages in the interior of
Mexico, suggesting that the Tijuana (and the rest of the Mexican border region in
general) is more integrated with the United States labor market than the Mexican interior
is. This evidence is consistent with migration being a major factor integrating these labor
markets.
The framework described in (1) can be modified to formally test the differences in
the estimated coefficient values across regions and industries. The significance and sign
of coefficients estimated from interacting region and industry dummy variables with the
relevant variables reveals whether there are statistically significant differences across
regions and industries as well as whether certain regions or industries are more integrated
with United States labor markets.
B. Empirical Results: Regional Differences before and after NAFTA
The data used to estimate (1) and its variants are the same log wage means for
matched U.S.-Mexican age-education cells. The data are quarterly and run from 1987 to
2001. The maximum number of observations possible is 14,400 (48 age-education
groups, 15 years, 4 quarters per year, and 5 Mexican regions). Not all cells can be
matched due to demographic differences between the countries and following age groups
over time, so that the actual number of observations used in estimation is 13,145.
Table 1 provides the results from the basic model in which we include regional
effects for the entire sample period. This table simply extends Robertson (2000) by
11
adding four more post-NAFTA years of data. The four border cities (Tijuana, Ciudad
Juarez, Matamoros, and Nuevo Laredo) are included as interaction effects, leaving the
main effects to reflect Mexico City. The results are consistent with the basic findings of
the earlier paper. The shock term is positive and significant for Mexico City, and the
estimated effects of U.S. wage shocks are larger (although only significantly so for
Ciudad Juarez) in the border cities. The convergence term is negative and significant, as
expected, and the rate of convergence is larger for the border cities. The cities with the
highest rates of migration and largest number of maquiladoras have the fastest rates of
convergence back to the equilibrium differential. Table 1 also shows that these results
are robust across education groups, especially for less educated workers.
The effects of NAFTA are estimated by modifying the basic equation used in
table 1 by including a dummy variable equal to 1 for the post NAFTA years (1994 and
on). In addition to the main effect, this dummy variable is interacted with the shock and
convergence terms found in Table 1. These results are found in table 2. The main
effects are qualitatively very similar to those found in table 1. The interactions with the
NAFTA effects, however, are generally not significant and generally indicate less
integration following NAFTA.7
The peso crisis in 1995 and the devaluation in 1987 may affect the results because
the nominal value of the peso is used to calculate the dollar value of Mexican wages. In
addition, the extreme movements of the peso may affect wages in ways that could be
mistaken for the effects of trade liberalization because the peso crisis occurs in the first
year of NAFTA. As a first pass of addressing these problems, I exclude 1987 and 1995
7
The sign of the interaction coefficients on the shock terms are often negative and are generally
positive on the convergence coefficients.
12
from estimation. Table 3 contains the results from the base regression (comparable to
table 1). The results without 1987 and 1995 seem to be stronger in the sense that now
the regional shock terms are positive and significant in column 1. The border effects are
also stronger in all three columns. Thus, it seems as if removing the periods with large
movements in the peso improves the estimation results.
The results of including NAFTA interaction terms when 1987 and 1995 are
excluded are found in table 4. The main shock and convergence terms are still significant
and have the expected sign. Overall, the results seem similar to those found in table 1.
Like table 2, however, table 4 shows little evidence of a significant increase in integration
by either measure. The main difference between tables 2 and 4 is that in table 4 the rate
of convergence now seems to be slightly higher following 1994, but the difference is not
statistically significant.
C. Empirical Results: Industry Differences before and after NAFTA
Rather than regions, one may think that wages in industries are closely linked
across countries. Since trade and investment are generally focused in manufacturing, it
may be the case that wages in manufacturing industries are more closely linked than
wages in other industries. There are, however, some reservations about this approach. If
workers are mobile across industries, then it is not clear that there should be systematic
wage differences across industries that would be transmitted across countries.8 Robertson
8
The phenomenon of inter-industry wage differentials has spawned a very large literature, and
significant differences in industry wages persist in Latin America. These differentials are also
highly correlated across developed countries. There is relatively little research on these
differentials in developing countries.
13
and Dutkowsky (2002) find that labor market adjustment costs at the 2-digit level of
Mexican manufacturing are small – about an order of magnitude smaller than in
developed countries. Another problem is that even narrowly defined industries may be
very heterogeneous. The most relevant example of this case would be the kind of vertical
specialization implied by Feenstra and Hanson (1997). They describe a kind of
production fragmentation in which the final production of a good is broken into stages,
and each stage may have a different production factor intensity. They suggest that capital
flows into Mexico may be within a very narrowly defined industry, but the kinds of tasks
performed (e.g. design versus assembly) may be very different. These different stages
may therefore be subject to different shocks that may or may not be transmitted clearly
across borders.
Even with these problems, however, focusing on industries is intuitive and helpful
if industries are linked through investment or product prices. A shock to the demand for
textiles in the United States, for example, may affect the demand for textile labor in
Mexico if foreign investment and trade link labor markets. Services, such as construction
and commerce, may be less integrated under these mechanisms. If there are no
systematic differences across industries, then it may be that shocks are not transmitted
though industry-specific channels (increasing support for the migration hypothesis), or
that shocks that are transmitted by industry-specific channels are quickly diffused
throughout the economy.
To estimate the effects of industry-specific shocks, I change the definition of the
age-education categories to accommodate industries. To preserve the integrity of the cell
sizes, I kept the number of cells at 48 but changed the composition. Now the 48 cells are
defined by 2 education groups (less than 12 years of education, 12 or more years), 4 age
14
groups, and 6 industry groups. The industry groups include construction (the base
industry), textiles, machinery and transport equipment, other manufacturing, commerce
(wholesale and retail trade), and other services. The manufacturing industries were
divided into these groups because the majority of the maquiladora investment is in the
first two industries, regardless of where in the country the maquiladoras are located.
Table 5 contains the estimation results for the base equation and the industry
dummy variables and interaction terms. The first main result from the table is that none
of the shock terms are significant. The second main result is that the convergence terms
for construction are significant and have the expected sign. The textile industry has a
statistically significant higher rate of convergence to the U.S.-Mexican wage differential
than construction, and the other services sector has a statistically significant lower rate of
convergence than construction. None of the other industries have significantly significant
marginal effects. Regional controls and time effects were not included.
One of the most striking characteristics of Mexican industries is the relatively
high rate of female labor force participation. Furthermore, as in developed countries,
females tend to specialize in certain economic sectors. These two phenomena suggest
that the rate of convergence and responsiveness to international wage shocks may differ
by gender. I therefore recalculated the cell average wages by creating averages that also
differed by gender.
The results of estimating the industry equations are found in tables 6 (males) and
7 (females). As in each of the earlier tables, the results are also presented when the
sample is divided by education group. The only significant result for males is that the
convergence term for construction is significant and carries the expected sign. The shock
terms for construction are positive, as expected, but the estimated coefficients are not
15
statistically significant. None of the interaction terms are significant. The coefficient
estimates of both the responsiveness to shocks and the rate of convergence are about the
same size, but the lack of statistical significance comes from the relatively large standard
errors. Since regional controls are not included, these results suggest that the integration
that was suggested by tables 1-4 is more a function of regional differences than industrial
differences, again lending support to the migration hypothesis.
The results for females, found in table 7, merit comment. Here the shock terms
are generally insignificant, with the exception of other manufacturing. This result
primarily emerges for less educated female workers. It is interesting, however, that the
convergence term for other manufacturing is not significant and the estimate of the
marginal effect is positive, suggesting slower convergence. This is the kind of result that
would emerge if workers in the same industry were subject to common shocks, but there
was no strong link between the industries. The other convergence estimates are more
interesting. Construction workers have large estimated rates of convergence. Only more
educated women in textiles have significantly faster convergence rates. It is interesting
that more educated women in textiles also have relatively strong, albeit not statistically
significant, responses to United States shocks. This, again, may illustrate the influence of
regional effects.
The two other interesting effects are found in the estimated convergence rates for
electric machinery and transport equipment. These industries have significantly slower
rates of convergence than construction even though they have high rates of trade and
foreign investment. The main result that does, however, support the trade and
investment hypothesis is slower rates of convergence in other services because other
services are generally insulated from the international economy except, possibly, from
16
migration. This may also reflect the very different mix of female employment in the two
countries, because Mexican women in other services may be more highly concentrated in
domestic help than women in the United States.
To facilitate comparisons with tables 1-4, we also explore the marginal effects of
NAFTA and again differentiate between males (table 8) and females (table 9). To avoid
exchange rate effects, we now define the NAFTA period as 1997-2001. The results in
table 8 suggest some significant differences during the NAFTA period. First, the main
shock terms are generally insignificant. During the NAFTA period, however, wages in
textiles exhibit a strong and positive response to U.S. shocks – especially for the more
educated workers. This contrasts with a slightly slower adjustment speed of wages in the
industry. Overall, the adjustment speed in textiles is not statistically different from that
of construction. In general, the estimates of the adjustment terms are negative and
statistically significant for construction, but the industry-specific marginal effects
generally are not significant. Effects of the NAFTA period differ across education
groups. At least for construction workers, the more educated workers seem to show
faster adjustment in the NAFTA period, while less educated workers have a slower
adjustment in the same period.
Again, the results for females are somewhat more varied. In general, the shock
terms are only significant for females in other manufacturing. The convergence terms are
significant and negative for construction, but most of the industry marginal effects are
insignificant. As in table 7, the main exceptions are electric machinery and transportation
equipment. Wages in these industries exhibit slower adjustment for the whole sample of
females and for the less-educated workers.
The NAFTA effects for females are again interesting. . The shock estimates
17
during the NAFTA period for textiles and other manufacturing, for example, are very
large and suggest almost a negative responsiveness. This could be consistent with the
hypothesis of capital moving from the United States into Mexico that moved jobs. That
is, as wages fell in the United States, wages increased in Mexico for women in these
industries. In this case, there would be a change in the equilibrium differential in wages
for these industries and the convergence term would not necessarily apply. Even so, the
rate of convergence in textile wages is the fastest for the most educated workers, and is
slightly smaller during the NAFTA period (there is no difference in the convergence term
for other manufacturing in the NAFTA period). These results suggest that female
employment in the textile industry may be a place to evaluate, in later research, the
hypothesis that NAFTA facilitates the movement of jobs from the United States to
Mexico in later research.
The reader may have noticed that regional controls were not included and,
without them, the estimates of the main and marginal effects were less precise. This may
be consistent with the migration hypothesis if integration is has a stronger relationship
with regions than to trade. Within regions, however, we should be able to get more
precise estimates of industry effects. Tables 10 and 11 contain estimates of the base
industry regression by city (table 10) and when NAFTA effects are included (table 11).
As expected, the results are more consistent with the basic model when the
estimation is done separately for each city. The main shock effects (in construction) are
positive, although only significant in Nuevo Laredo. It is interesting that the effects of
shocks in the United States are strongest in the textile sector in Tijuana. The rest of the
marginal effects are generally not significant. The convergence terms, on the other hand,
are all negative and significant for construction. All of the marginal effects are
18
significant in Mexico City, with the largest being in textiles and electric machinery and
transportation equipment. A similar result emerges in Tijuana, in which all marginal
effects are negative and significant with the exception of electric machinery and transport
equipment. Textiles again has a very large (in absolute value) coefficient, suggesting that
the textile industry in Tijuana, rather than the textile industry nation wide, is closely
integrated with the United States labor markets. This weighs against the trade
hypothesis.
Table 11 contains the effects of the NAFTA on these estimates. When the
NAFTA (1997-2001) dummy is included, the shock terms are still generally insignificant.
The effect of NAFTA is relatively large and positive in textiles in Tijuana and Ciudad
Juarez, and large and negative in Matamoros. The main convergence terms remain
negative and significant across all cities, but the rate of convergence is only higher for
construction in Tijuana and Ciudad Juarez. The rate of convergence does not change in
Mexico City in a statistically significant way. The estimated rate of convergence slows
in the electric and transportation equipment industry in Tijuana, but no other significant
effects emerge there. The only other significant effects that do emerge are for textiles in
Matamoros (slightly slower convergence between 1997-2001) and Nuevo Laredo
(slightly faster convergence between 1997-2001). Overall, however, there is very little
evidence of a significant increase in integration across industries in the 1997-2001 period.
IV. Conclusions
While there has been a significant increase in trade and foreign direct investment
in Mexico following NAFTA, this paper finds surprisingly little evidence of increased
19
integration of labor markets. There is only slight evidence that wages have converged in
absolute levels following NAFTA, but this is largely due to the large negative shock that
occurred with the December 1994 peso crisis. Some evidence does emerge that suggests
that the long-run trend in absolute wage convergence has increased following NAFTA,
but this could be due to the recovery from the crisis.
Using an alternative approach that is based upon a constant differential in the two
nations’ wages, we also estimate the responsiveness to shocks and the short-run rate of
convergence back to the equilibrium differential. This approach reveals very limited
evidence of increased labor market integration following NAFTA. The Mexican border
region remains more integrated with the United States than the Mexican interior, but
there is little change in this pattern following NAFTA. Furthermore, neither the border
nor the interior seems more integrated. We also examined integration using industries as
our unit of analysis. There is some evidence that the overall rate of convergence in
Tijuana and Ciudad Juarez increased, but this effect seems more closely tied to a nontradable industry (construction).
These results may contribute to the debate about what factors integrate labor
markets in North America. Theory tells us that trade, capital flows, and migration can all
integrate labor markets. NAFTA liberalized trade and capital flows, but did not relax
restrictions on migration. To the extent that there is very little evidence that integration
increased - especially in trade and FDI-intensive industries – lends at least circumstantial
support towards the migration hypothesis.
20
References
Boyer, G. and Hatton, T. (1994) "Regional Labour Market Integration in England and
Wales, 1850-1913" in Grantham, G. and MacKinnon, M. (eds.) Labor-market Evolution
Routledge, London and New York.
Collins, William J. “Labor Mobility, Market Integration, and Wage Convergence in Late
19th Century India” Explorations in Economic History v36, n3 (July 1999): 246-77.
Deaton, Angus “Panel Data from Time Series of Cross-Sections” Journal of
Econometrics v30, n1-2 (Oct.-Nov. 1985): 109-26.
Feenstra, Robert C. ; Hanson, Gordon H. “Foreign Direct Investment and Relative
Wages: Evidence from Mexico's Maquiladoras” Journal of International Economics v42,
n3-4 (May 1997): 371-93.
Hanson, Gordon H. ; Robertson, Raymond ; Spilimbergo, Antonio “Does Border
Enforcement Protect U.S. Workers from Illegal Immigration?” Review of Economics and
Statistics v84, n1 (February 2002): 73-92.
Moazzami, Bakhtiar “Regional Wage Convergence in Canada: An Error-Correction
Approach” Canadian Journal of Regional Science v20, n3 (Autumn 1997): 341-50.
Mokhtari, Manouchehr; Rassekh, Farhad. "The Tendency Towards Factor Price
Equalization Among OECD Countries." The Review of Economics and Statistics v71,
issue 4, pp. 636-642. November, 1989.
O’Rourke, Kevin; Williamson, Jeffrey G. "Late Nineteenth-Century Anglo-American
Factor-Price Convergence: Were Heckscher and Ohlin Right?" Journal of Economic
History v54, issue 4, pp. 892-916. December, 1994.
O’Rourke, Kevin; Taylor, Alan M.; Williamson, Jeffrey G. "Factor Price Convergence in
the Late Nineteenth Century." International Economic Review v37, issue 3, pp. 499-530.
August, 1996.
Reynolds, Clark W. “Will a Free Trade Agreement Lead to Wage Convergence?
Implications for Mexico and the United States” in Bustamante, Jorge A.; Reynolds, Clark
W.; Hinojosa Ojeda, Raul A., eds. U.S.-Mexico relations: Labor market interdependence
Stanford: Stanford University Press (1992): 477-86.
Robertson, Raymond “Wage Shocks and North American Labor-Market Integration”
American Economic Review v90, n4 (September 2000): 742-64.
Robertson, Raymond ; Dutkowsky, Donald H. “Labor Adjustment Costs in a Destination
Country: The Case of Mexico” Journal of Development Economics v67, n1 (February
2002): 29-54.
21
Saint-Paul, Gilles “Economic Integration, Factor Mobility, and Wage Convergence”
International Tax and Public Finance v4, n3 (July 1997): 291-306.
Saint-Paul, Gilles “Economic Integration, Factor Mobility, and Wage Convergence” in
Razin, Assaf; Sadka, Efraim, eds. The economics of globalization: Policy perspectives
from public economics Cambridge; New York and Melbourne: Cambridge University
Press (1999): 313-32.
Williamson, Jeffrey G. Globalization, Convergence, and History. Journal of Economic
History v56, issue 2, pp. 277-306. June, 1996.
22
Figure 1: Mexican Manufacturing Exports
15000
Millions US$
10000
5000
0
1991m1
1994m1
1997m1
date
2000m1
2002m3
Figure 2: Share of Manufacturing in Urban Employment in Mexico
(mean) mfg
.378425
.306502
1987q1
1989q1
1991q1
1993q1
1995q1
date
1997q1
1999q1
2001q1
23
Figure 3: Textile Share of Total Urban Employment in Mexico
(mean) tex
.038402
.022146
1987q1
1989q1
1991q1
1993q1
1995q1
date
1997q1
1999q1
2001q1
24
Figure 4: Mexican/United States Dollar Wage Ratios
Tijuana
Matamoros
Mexico City
.811867
.643433
87
89
91
93
95
Year 1900+
97
99
101
Notes: These series are the ratio of Mexican to United States log monthly earnings, in which the log
monthly earnings were first converted to dollars using the nominal exchange rate. These (log) dollar values
were then arithmetically averaged across all age-education groups (using cell sizes as weights) and used to
calculate the ratio of log wages.
25
Figure 5a: Normalized Wage Ratios
Tijuana
Mexico City
1.27728
.976901
1987q1
1989q1
1991q1
1993q1
1995q1
date
1997q1
1999q1
2001q1
Notes: This graph shows the ratio of the Mexican inflation-adjusted wage relative to the United States
inflation-adjusted wage, when the wages in both countries are normalized to be equal to 1 for 1992. The
real values are compared without using the nominal exchange rate.
Figure 5b: Trend Analysis
Tijuana
Fitted values
Mexico City
Fitted values
1.27728
.965621
1987q1
1989q1
1991q1
1993q1
1995q1
date
1997q1
1999q1
2001q1
26
Table 1: Basic Results
Eneu 1987-2001
(1)
All Ed
Tijuana
-0.080
(0.039)*
C. Juarez
-0.122
(0.032)**
Matamoros
-0.092
(0.030)**
N. Laredo
-0.064
(0.047)
Shock (Mexico City)
0.154
(0.026)**
Tijuana
0.062
(0.039)
C. Juarez
0.067
(0.030)*
Matamoros
0.080
(0.043)
N. Laredo
0.007
(0.036)
Convergence (Mexico City)
-0.252
(0.031)**
Tijuana
-0.159
(0.024)**
C. Juarez
-0.111
(0.020)**
Matamoros
-0.095
(0.019)**
N. Laredo
-0.058
(0.026)*
Constant
-0.563
(0.067)**
Observations
13145
R-squared
0.34
(2)
More Ed
-0.035
(0.043)
-0.040
(0.034)
-0.080
(0.042)
-0.002
(0.060)
0.364
(0.048)**
0.024
(0.058)
0.071
(0.099)
-0.116
(0.103)
-0.127
(0.065)
-0.338
(0.022)**
-0.145
(0.030)**
-0.069
(0.026)*
-0.100
(0.026)**
-0.025
(0.035)
-0.743
(0.046)**
6463
0.37
(3)
Less Ed
0.022
(0.021)
-0.148
(0.017)**
-0.021
(0.017)
-0.031
(0.023)
0.159
(0.022)**
0.062
(0.041)
0.090
(0.027)**
0.104
(0.027)**
0.029
(0.037)
-0.328
(0.024)**
-0.131
(0.012)**
-0.145
(0.012)**
-0.068
(0.014)**
-0.052
(0.015)**
-0.744
(0.047)**
6682
0.38
Notes: All equations include year effects. Robust standard errors in parentheses. * significant at 5%; **
significant at 1%
27
Table 2: Effects of NAFTA on Shock and Convergence Estimates
Nafta Main Effect
Tijuana
C. Juarez
Matamoros
N. Laredo
Shock (Mexico City)
Tijuana
C. Juarez
Matamoros
N. Laredo
Convergence (Mexico City)
Tijuana
C. Juarez
Matamoros
N. Laredo
Constant
Observations
R-squared
(1)
(2)
(3)
All Ed
(Nafta)
0.018
(0.054)
-0.129
(0.029)**
-0.122
(0.031)**
-0.120
(0.036)**
-0.129
(0.038)**
0.093
-0.106
(0.035)** (0.074)
0.050
-0.007
(0.050)
(0.068)
0.040
0.003
(0.041)
(0.058)
0.080
-0.026
(0.054)
(0.096)
-0.009
0.075
(0.048)
(0.084)
-0.083
0.003
(0.007)** (0.027)
-0.125
-0.015
(0.021)** (0.013)
-0.087
-0.005
(0.020)** (0.004)
-0.088
0.002
(0.023)** (0.004)
-0.079
-0.018
(0.022)** (0.006)**
-0.130
(0.015)**
13145
More Ed
(Nafta)
0.008
(0.115)
-0.121
(0.038)**
-0.116
(0.054)*
-0.110
(0.057)
-0.120
(0.058)*
0.352
-0.491
(0.046)** (0.137)**
-0.027
0.005
(0.082)
(0.118)
0.002
0.011
(0.133)
(0.126)
-0.113
-0.181
(0.124)
(0.174)
-0.235
0.351
(0.091)*
(0.157)*
-0.094
-0.006
(0.006)**
(0.063)
-0.119
-0.022
(0.028)**
(0.027)
-0.085
-0.009
(0.037)*
(0.010)
-0.085
0.001
(0.036)*
(0.009)
-0.077
-0.019
(0.035)*
(0.010)
-0.146
(0.013)**
6463
Less Ed
(Nafta)
-0.043
(0.041)
-0.106
(0.023)**
-0.120
(0.030)**
-0.120
(0.040)**
-0.125
(0.034)**
0.048
-0.020
(0.029)
(0.046)
0.062
-0.010
(0.055)
(0.078)
0.059
-0.012
(0.037)
(0.066)
0.114
-0.002
(0.045)* (0.108)
0.030
0.024
(0.051)
(0.091)
-0.082
-0.027
(0.005)** (0.021)
-0.113
-0.027
(0.020)** (0.013)
-0.087
-0.008
(0.019)** (0.005)
-0.086
-0.004
(0.026)** (0.004)
-0.076
-0.022
(0.020)** (0.006)**
-0.136
(0.008)**
6682
0.09
0.10
0.09
Notes: Robust standard errors in parentheses. * significant at 5%; ** significant at 1%
28
Table 3: Without 1987 and 1995
Tijuana
C. Juarez
Matamoros
N. Laredo
Shock (Mexico City)
Tijuana
C. Juarez
Matamoros
N. Laredo
Convergence (DF)
Tijuana
C. Juarez
Matamoros
N. Laredo
Constant
Observations
R-squared
(1)
All Ed
-0.163
(0.029)**
-0.145
(0.026)**
-0.118
(0.025)**
-0.101
(0.046)*
0.107
(0.027)**
0.085
(0.039)*
0.068
(0.033)*
0.102
(0.043)*
0.042
(0.037)
-0.173
(0.028)**
-0.198
(0.025)**
-0.115
(0.019)**
-0.101
(0.019)**
-0.074
(0.028)*
-0.299
(0.051)**
11586
0.20
(2)
More Ed
-0.127
(0.030)**
-0.085
(0.029)**
-0.112
(0.034)**
-0.041
(0.057)
0.277
(0.134)*
0.101
(0.069)
0.080
(0.099)
-0.072
(0.088)
-0.111
(0.054)*
-0.244
(0.043)**
-0.189
(0.033)**
-0.087
(0.026)**
-0.109
(0.026)**
-0.042
(0.036)
-0.420
(0.071)**
5685
0.22
(3)
Less Ed
-0.083
(0.020)**
-0.166
(0.018)**
-0.058
(0.021)*
-0.082
(0.028)**
0.118
(0.024)**
0.079
(0.042)
0.085
(0.032)*
0.121
(0.035)**
0.065
(0.038)
-0.239
(0.032)**
-0.178
(0.016)**
-0.142
(0.015)**
-0.076
(0.019)**
-0.073
(0.020)**
-0.514
(0.067)**
5901
0.25
Notes: All equations include year effects. Robust standard errors in parentheses. * significant at 5%; **
significant at 1%.
29
Table 4: NAFTA Effects without 1987 and 1995
(1)
All Ed
(Nafta)
N
-0.015
(0.043)
Tijuana
-0.201
(0.025)**
C. Juarez
-0.157
(0.026)**
Matamoros
-0.146
(0.031)**
N. Laredo
-0.140
(0.040)**
Shock (Mexico City)
0.082
-0.101
(0.031)*
(0.063)
Tijuana
0.088
-0.014
(0.047)
(0.066)
C. Juarez
0.054
0.002
(0.042)
(0.059)
Matamoros
0.130
-0.067
(0.047)** (0.088)
N. Laredo
0.040
0.030
(0.051)
(0.086)
Convergence (DF)
-0.064
-0.005
(0.006)** (0.022)
Tijuana
-0.180
-0.013
(0.022)** (0.010)
C. Juarez
-0.105
-0.006
(0.018)** (0.003)
Matamoros
-0.103
0.003
(0.021)** (0.002)
N. Laredo
-0.085
-0.017
(0.024)** (0.005)**
-0.076
Constant
(0.010)**
11586
Observations
R-squared
0.11
(2)
More Ed
(Nafta)
-0.041
(0.101)
-0.200
(0.030)**
-0.156
(0.046)**
-0.149
(0.045)**
-0.128
(0.059)*
0.316
-0.474
(0.114)* (0.082)**
0.124
-0.069
(0.089)
(0.117)
0.063
-0.018
(0.123)
(0.135)
0.022
-0.286
(0.122)
(0.186)
-0.196
0.316
(0.082)* (0.148)*
-0.062
-0.023
(0.010)** (0.057)
-0.175
-0.025
(0.027)** (0.022)
-0.107
-0.012
(0.034)** (0.008)
-0.107
-0.001
(0.031)** (0.006)
-0.080
-0.020
(0.037)* (0.009)*
-0.070
(0.015)**
5685
0.11
(3)
Less Ed (Nafta)
-0.044
(0.037)
-0.166
(0.026)**
-0.153
(0.028)**
-0.125
(0.041)**
-0.136
(0.040)**
0.050
-0.026
(0.028) (0.043)
0.081
-0.011
(0.053) (0.075)
0.067
-0.011
(0.043) (0.067)
0.144
-0.032
(0.044)** (0.098)
0.078
-0.022
(0.052) (0.091)
-0.081
-0.017
(0.007)** (0.018)
-0.166
-0.015
(0.023)** (0.012)
-0.106
-0.007
(0.019)** (0.005)
-0.092
0.000
(0.028)** (0.003)
-0.084
-0.019
(0.025)** (0.006)**
-0.114
(0.011)**
5901
0.13
30
Table 5: Industry Error Correction Models
Males and Females
(1)
(2)
All Ed
More Ed
Shock (Construction)
0.068
-0.074
(0.053)
(0.291)
Textiles
-0.015
0.502
(0.081)
(0.316)
Electric & Transport
-0.045
0.131
(0.059)
(0.342)
Other Mfg.
0.075
0.074
(0.087)
(0.355)
Commerce
0.040
0.554
(0.104)
(0.355)
Other Services
-0.067
0.329
(0.079)
(0.323)
Convergence (Constr.)
-0.138
-0.357
(0.016)**
(0.022)**
Textiles
-0.100
-0.133
(0.035)**
(0.047)*
Electric & Transport
-0.010
0.039
(0.022)
(0.032)
Other Mfg.
-0.024
0.027
(0.020)
(0.032)
Commerce
0.005
0.062
(0.020)
(0.038)
Other Services
0.038
0.142
(0.016)*
(0.026)**
Constant
-0.288
-0.633
(0.028)**
(0.049)**
Observations
12686
5790
R-squared
0.20
0.26
(3)
Less Ed
0.061
(0.058)
-0.074
(0.077)
-0.057
(0.064)
0.062
(0.097)
-0.027
(0.094)
-0.098
(0.081)
-0.096
(0.009)**
-0.067
(0.025)*
-0.007
(0.012)
-0.019
(0.014)
0.001
(0.008)
0.029
(0.009)**
-0.199
(0.019)**
6896
0.23
Notes: Robust standard errors in parentheses. The main effects and year effects are included in the
estimation but are not reported here.
31
Table 6: Sector Error Correction Models
Males
Shock (Construction)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Convergence (Constr.)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Constant
Observations
R-squared
(1)
All Ed
0.062
(0.058)
-0.047
(0.080)
-0.004
(0.061)
-0.015
(0.080)
0.039
(0.097)
-0.030
(0.066)
-0.148
(0.017)**
-0.071
(0.040)
-0.024
(0.033)
-0.032
(0.023)
-0.017
(0.025)
0.039
(0.018)*
-0.315
(0.034)**
12213
0.19
(2)
More Ed
0.000
(0.343)
0.269
(0.386)
0.132
(0.354)
-0.111
(0.421)
0.612
(0.436)
0.035
(0.345)
-0.398
(0.022)**
-0.067
(0.043)
-0.004
(0.043)
0.022
(0.033)
0.046
(0.028)
0.139
(0.031)**
-0.721
(0.061)**
5588
0.26
(3)
Less Ed
0.056
(0.061)
-0.086
(0.073)
-0.020
(0.063)
-0.023
(0.084)
-0.035
(0.084)
-0.040
(0.070)
-0.099
(0.009)**
-0.034
(0.020)
0.003
(0.008)
-0.022
(0.011)
-0.003
(0.007)
0.033
(0.008)**
-0.206
(0.022)**
6625
0.20
Notes: Robust standard errors in parentheses. The main effects and year effects are included in the
estimation but are not reported here.
32
Table 7: Sector Error Correction Models
Females
Shock (Construction)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Convergence (Constr.)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Constant
Observations
R-squared
(1)
All Ed
-0.023
(0.037)
0.060
(0.059)
0.037
(0.045)
0.227
(0.057)**
0.053
(0.087)
-0.012
(0.045)
-0.271
(0.024)**
-0.027
(0.051)
0.136
(0.039)**
0.048
(0.035)
0.036
(0.031)
0.129
(0.026)**
-0.501
(0.044)**
9361
0.18
(2)
More Ed
0.124
(0.428)
0.294
(0.515)
-0.059
(0.443)
-0.044
(0.530)
0.026
(0.436)
0.168
(0.459)
-0.346
(0.027)**
-0.252
(0.092)*
-0.052
(0.040)
-0.041
(0.032)
-0.107
(0.079)
0.080
(0.033)*
-0.670
(0.061)**
3932
0.25
(3)
Less Ed
-0.043
(0.019)*
0.044
(0.046)
0.031
(0.030)
0.233
(0.047)**
0.044
(0.083)
-0.031
(0.026)
-0.250
(0.036)**
0.038
(0.048)
0.169
(0.036)**
0.071
(0.054)
0.055
(0.038)
0.142
(0.036)**
-0.493
(0.073)**
5429
0.18
Notes: Robust standard errors in parentheses. The main effects and year effects are included in the
estimation but are not reported here.
33
Table 8: NAFTA Effects
Males
(1)
(2)
(3)
All Ed
(Nafta)
More Ed
(Nafta)
Less Ed (Nafta)
Shock (Construction)
0.071
-0.081
0.063
-0.284
0.055
-0.055
(0.079)
(0.055)
(0.204)
(0.505)
(0.085) (0.057)
Textiles
-0.125
0.208
0.071
0.493
-0.125
0.124
(0.088) (0.077)** (0.263)
(0.556)
(0.089) (0.073)
Electric & Transport
-0.043
0.132
0.169
-0.164
-0.046
0.114
(0.083)
(0.074)
(0.227)
(0.522)
(0.086) (0.068)
Other Mfg.
0.049
-0.121
-0.218
0.698
0.094
-0.216
(0.131)
(0.142)
(0.253)
(0.870)
(0.142) (0.151)
Commerce
0.048
0.014
0.502
0.407
-0.006
-0.015
(0.107)
(0.072)
(0.369)
(0.587)
(0.098) (0.071)
Other Services
-0.071
0.040
-0.079
0.342
-0.062
0.011
(0.095)
(0.104)
(0.232)
(0.564)
(0.102) (0.106)
Convergence (Constr.) -0.119
0.012
-0.304
-0.083
-0.087
0.028
(0.019)** (0.008) (0.019)** (0.031)* (0.005)** (0.006)**
Textiles
-0.073
0.003
-0.068
0.034
-0.049
0.004
(0.044)
(0.006)
(0.051)
(0.015)* (0.021)* (0.004)
Electric & Transport
-0.024
-0.002
0.051
-0.007
-0.005
0.002
(0.035)
(0.003)
(0.060)
(0.009)
(0.009) (0.002)
Other Mfg.
-0.027
-0.000
0.064
0.006
-0.028
0.002
(0.028)
(0.003)
(0.046)
(0.011) (0.009)** (0.002)
Commerce
-0.009
0.005
0.089
0.009
-0.006
0.009
(0.030)
(0.003)
(0.046)
(0.009)
(0.006) (0.001)**
Other Services
0.040
0.004
0.203
0.009
0.021
0.007
(0.022)
(0.003) (0.026)** (0.009) (0.005)** (0.002)**
Constant
-0.205
-0.426
-0.153
(0.031)**
(0.022)**
(0.009)**
Observations
12213
5588
6625
R-squared
0.07
0.14
0.05
Notes: Robust standard errors in parentheses. The main effects are included in the estimation but are not
reported here.
34
Table 9: NAFTA Effects
Females
(1)
Shock (Construction)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Convergence (Constr.)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Constant
Observations
R-squared
All Ed
(Nafta)
(2)
More Ed (Nafta)
-0.046
(0.045)
0.077
(0.099)
0.012
(0.058)
0.307
(0.083)**
0.091
(0.086)
-0.028
(0.063)
-0.255
(0.026)**
-0.002
(0.059)
0.162
(0.040)**
0.071
(0.039)
0.064
(0.035)
0.159
(0.026)**
0.091
(0.201)
-0.136
(0.224)
-0.015
(0.202)
-0.226
(0.251)
-0.140
(0.230)
-0.033
(0.205)
-0.028
(0.028)
0.042
(0.022)
0.032
(0.022)
0.036
(0.022)
0.046
(0.022)*
0.034
(0.021)
-0.029
(0.358)
0.812
(0.438)
-0.166
(0.390)
0.734
(0.427)
0.264
(0.370)
0.404
(0.408)
-0.254
(0.034)**
-0.269
(0.111)*
-0.007
(0.058)
-0.017
(0.046)
-0.067
(0.092)
0.159
(0.037)**
-0.407
(0.041)**
9361
0.08
0.437
(0.124)**
-1.155
(0.322)**
0.092
(0.141)
-3.499
(0.382)**
-0.490
(0.611)
-0.769
(0.274)*
-0.155
(0.054)**
0.068
(0.045)
0.034
(0.023)
-0.009
(0.025)
0.041
(0.023)
0.021
(0.023)
-0.383
(0.050)**
5429
0.16
(3)
Less Ed (Nafta)
-0.058
(0.037)
0.052
(0.101)
0.011
(0.050)
0.255
(0.060)**
0.077
(0.089)
-0.069
(0.042)
-0.254
(0.048)**
0.064
(0.062)
0.190
(0.048)**
0.090
(0.063)
0.083
(0.050)
0.157
(0.048)**
0.042
(0.202)
-0.075
(0.231)
0.021
(0.203)
-0.067
(0.257)
-0.088
(0.235)
0.054
(0.206)
-0.026
(0.047)
0.073
(0.046)
0.068
(0.045)
0.068
(0.046)
0.078
(0.046)
0.073
(0.046)
-0.476
(0.073)**
5429
0.07
Notes: Robust standard Errors in parentheses. The main effects are included in estimation but are not
reported.
35
Table 10: Industry Estimation by City
Shock (Construction)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Convergence (Constr.)
Textiles
Electric & Transport
Other Mfg.
Commerce
Other Services
Constant
Observations
R-squared
(1)
Mexico City
0.031
(0.029)
-0.043
(0.079)
0.108
(0.057)
0.151
(0.089)
0.115
(0.102)
0.001
(0.070)
-0.130
(0.019)**
-0.097
(0.020)**
-0.103
(0.024)**
-0.043
(0.020)*
-0.047
(0.020)*
-0.071
(0.027)*
-0.313
(0.050)**
2774
0.31
(2)
Tijuana
0.097
(0.080)
0.352
(0.116)**
0.010
(0.108)
0.001
(0.087)
0.045
(0.114)
-0.036
(0.109)
-0.276
(0.033)**
-0.377
(0.065)**
-0.063
(0.047)
-0.181
(0.050)**
-0.107
(0.036)**
-0.080
(0.033)*
-0.586
(0.056)**
2533
0.30
(3)
C. Juarez
0.199
(0.124)
-0.102
(0.164)
-0.177
(0.125)
-0.092
(0.283)
-0.114
(0.262)
-0.061
(0.156)
-0.284
(0.039)**
-0.103
(0.039)*
0.094
(0.050)
-0.074
(0.048)
-0.029
(0.044)
-0.016
(0.045)
-0.607
(0.077)**
2516
0.22
(4)
Matamoros
0.001
(0.184)
0.371
(0.231)
0.054
(0.186)
0.369
(0.236)
0.176
(0.222)
-0.092
(0.203)
-0.249
(0.028)**
-0.115
(0.044)*
-0.019
(0.037)
-0.119
(0.039)**
-0.061
(0.050)
-0.079
(0.035)*
-0.548
(0.059)**
2482
0.26
(5)
N. Laredo
0.208
(0.069)**
0.149
(0.146)
-0.151
(0.072)*
-0.055
(0.207)
0.231
(0.090)*
-0.150
(0.120)
-0.332
(0.031)**
-0.182
(0.045)**
0.078
(0.029)**
-0.065
(0.042)
-0.064
(0.046)
-0.013
(0.032)
-0.782
(0.064)**
2381
0.28
Notes: Robust standard errors in parentheses. The main effects are included in the estimation but are not
reported here.
36
0.07
R-squared
0.14
2533
97-01
(0.010)
-0.006
(0.012)
-0.003
(0.013)
0.005
(0.007)
0.008
(0.014)
-0.002
(0.028)*
-0.060
(0.199)
0.073
(0.267)
0.011
(0.394)
0.410
(0.144)
0.121
(0.230)
0.106
(0.140)
-0.038
0.10
2516
(0.073)**
-0.338
(0.048)
0.050
(0.055)
0.013
(0.067)
-0.043
(0.045)*
0.106
(0.061)
-0.096
(0.042)**
-0.192
(0.182)
-0.065
(0.330)
-0.113
(0.391)
-0.203
(0.166)
-0.204
(0.214)
-0.122
(0.166)
0.163
Main
C. Juarez
(3)
(4)
(0.005)
-0.001
(0.007)
0.009
(0.007)
-0.001
(0.004)
0.001
(0.011)**
0.033
(0.022)
-0.025
(0.314)
0.296
(0.399)
0.168
(0.372)
-0.387
(0.230)
0.326
(0.261)
-0.495
(0.224)
-0.273
-0.021
(0.024)
-0.023
(0.295)
0.045
(0.298)
0.095
(0.418)
-0.214
(0.140)
0.002
(0.278)
-0.160
(0.111)
0.037
97-01
0.002
(0.006)*
0.014
(0.004)*
0.012
(0.009)
-0.014
(0.005)
2381
0.11
0.11
(0.056)**
-0.328
(0.033)
0.055
(0.054)
-0.039
(0.064)
-0.098
(0.040)
0.039
(0.051)** (0.007)**
-0.205
(0.030)**
-0.171
(0.166)
-0.171
(0.116)
0.189
(0.280)
0.051
(0.086)
-0.140
(0.235)
0.201
(0.080)
0.138
Main
N. Laredo
(5)
2482
(0.052)**
-0.264
(0.032)
0.007
(0.060)
-0.051
(0.054)
-0.086
(0.034)
0.020
(0.061)**
-0.169
(0.030)**
-0.151
(0.206)
-0.299
(0.231)
0.074
(0.166)**
0.491
(0.133)
-0.130
(0.214)**
0.680
(0.129)
0.105
97-01
Matamoros
Main
Notes: Robust standard errors in parentheses. The main effects are included in the estimation but are not reported here.
2774
(0.036)**
(0.011)
(0.026)**
(0.030)
-0.008
(0.008)
0.002
(0.006)
0.006
(0.005)**
0.015
(0.016)
0.011
(0.022)*
-0.058
(0.334)
0.054
(0.358)
0.446
(0.344)
-0.197
(0.317)
0.046
(0.443)
0.426
(0.307)
-0.008
97-01
-0.221
(0.005)
(0.015)
-0.004
(0.035)
-0.055
(0.063)*
-0.140
(0.056)
-0.035
(0.079)**
-0.405
(0.025)**
-0.147
(0.204)
-0.054
(0.215)
-0.103
(0.214)
0.082
(0.217)
-0.019
(0.262)
0.147
(0.189)
0.046
Main
-0.124
-0.000
(0.004)
(0.016)
-0.019
0.006
(0.004)
(0.016)
-0.013
-0.002
(0.006)
(0.027)**
-0.016
-0.010
(0.005)
(0.031)*
-0.099
0.002
(0.013)
(0.015)**
-0.067
0.025
(0.216)
(0.129)
-0.077
0.002
(0.175)
(0.126)
-0.010
-0.048
(0.248)
(0.102)*
0.141
-0.223
(0.220)
(0.106)
0.248
0.181
(0.194)
(0.095)
0.055
0.105
(0.159)
(0.080)
-0.100
-0.018
-0.002
Observations
Constant
Other Services
Commerce
Other Mfg.
Electric & Transport
Textiles
Convergence (Constr.)
Other Services
Commerce
Other Mfg.
Electric & Transport
Textiles
Shock (Construction)
97-01
Tijuana
Mexico City
Main
(2)
(1)
Table 11: Industry-City Estimation with NAFTA Effects
37