NBER WORKING PAPERS SERIES
ENVIRONMENTAL IMPACTS OF A NORTH AMERICAN FREE TRADE AGREEMENT
Gene M. Grossman
Alan B. Krueger
Working Paper No. 3914
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
November 1991
This paper was prepared for the conference on the U.S.- Mexico
Free Trade Agreement. sponsored by SECOFI. We are grateful to
the Industrial Relations Section and International Finance
Section of Princeton University, and the National Science
Foundation for partial financial support. We thank Loren Baker.
Kainan Tang, and GuillerInO Frias for research assistance, and
Drusilla Brown, Gardener Evans, and Greg Schoepfle for sharing
their unpublished data. Joanne Gowa. Howard Gruenspecht. and
Jeff Mackie-Mason provided helpful comments and
discussion. This
paper is part of NBER's research program in International
Studies. Any opinions expressed are those of
the authors and not
those of the National Bureau of Economic Research.
NEER Working Paper #3914
November 1991
ENVIRONMENTAL IMPACTS OF A NORTH AMERICAN FREE TRADE AGREEMENT
ABSTRACT
A reduction in trade barriers generally will affect the
environment by expanding the scale of economic activity, by altering
the composition of economic activity, and by bringing about a change
in the techniques of production. We present empirical evidence to
assess the relative magnitudes of these three effects as they apply to
further trade liberalization in Mexico.
In Section 1. we use comparable measures of three air pollutants
in a cross-section of urban areas located in 42 countries to study the
relationship between air quality and economic growth. We find for two
pollutants (sulfur dioxide and smoke") that concentrations increase
with per capita GDP at low levels of national income, but decrease
with GD? growth at higher levels of income. Section 2 studies the
determinants of the industry pattern of U.S. imports from Mexico and
of value added by Mexico's maquiladora sector. We investigate whether
the size of pollution abatement costs in the U.S. industry influences
the pattern of international trade and investment. Finally, in
Section 3, we use the results from a computable general equilibrium
model to study the likely compositional effect of a NAFTA on pollution
in Mexico.
Gene H. Grossman Alan B. Krueger
Woodrow Wilson School Woodrow Wilson School
Princeton University Princeton University
Princeton, NJ 08544 princeton, NJ 08544
and NBER and NEER
Environmental advocacy groups In the United States have voiced their
concerns about a potential North American Free Trade Agreement (NAFTA). Some
went so far as to oppose the Congressional. granting of fast-track negotiating
authority to the President to enable American negotiators to enter Into talks
with their Mexican counterparts. The reservations of the lobbying groups
mirror a growing perception on the part of environmentalists worldwide that an
open world trading system may be inimical to the goal of preserving a clean,
healthy, and sustainable global commons.
The arguments linking trade liberalization with environmental
degradation have not been fully articulated.' With regard to a NAFFA, the
environmentalists have expressed a number of reasons for fearing that freer
trade and direct investment flows between the United States and Mexico may
aggravate pollution problems in Mexico and in the border region.2 At the
least discerning level, some have argued simply that any expansion of markets
and economic activity inevitably leads to more pollution and faster depletion
of scarce natural resources. A more pointed argument recognizes that
pollution already is a severe problem in Mexico and that the country's weak
regulatory infrastructure is strained to the breaking point. Under these
conditions, it is feared that any further industrialization that results from
the liberalization of trade and investment will exacerbate an already grave
situation.
Other environmentalists draw their conclusions by extrapolating the
experience of the maquiladora sector in Mexico. The maquiladoras are
See Low and Safedi (1991), who cite several examples of writings that view
open trade as detrimental to environmental protection.
2 See, for example, Gregory (1991), Kelly and Kemp (1991). National Wildlife
Foundation (1990), Leonard and Christensen (1991), and Ortman (1991).
2
predominantly foreign-owned firms that produce largely for export to the
United States under a Mexican policy that allows duty-free imports of foreign
components for further processing and re-export. Originally, maquiladoras
were required to locate within a 20-kilometer strip along the U.S. - Mexico
border in order to qualify for special customs treatment. The sector grew quite
rapidly and with little governmental oversight, and now is widely
regarded as being a major contributor to the perilous environmental and social
conditions in the border region. Environmental groups point to this sector as
a prime example of how unregulated expansion in response to trade opportunities
can create risks to worker safety and public health. They argue that
investments in this sector have been encouraged by the lax enforcement of
environment and labor protection laws in t4exico and fear that any further
expansion in trade and investment flows between the United States and Mexico
will be motivated by firms' desires to avoid the high costs of meeting U.S.
regulations.
A further concern of some environmental groups is that a NAFEA may
undercut regulatory standards in the United States. Spokespersons have made
the political-economic argument that, with freer trade, industry groups in the
United States will demand less stringent pollution controls in order to
preserve their international competitiveness, so that environmental standards
will tend toward a lowest common denominator. The environmentalists worry,
moreover, that existing environmental protection laws in the United States may
be seen as nontariff barriers to trade in the context of a regional trade
agreement.
While the environmental groups have raised a host of valid questions
they have so far been unable to provide convincing and well supported answers
3
to these questions. Many of their arguments fail to recognize all of the
implications of trade liberalization for resource allocation and natural
resource use in each of the trade partner countries. Moreover, die empirical
claims that have been made rely mostly on anecdotal evidence and on
extrapolation of the experience in one region or industry to the entirety of
economic activity in Mexico. Indeed, relatively little is known at any level
of generality about the relationship between a country's trade regime and its
rate of environmental degradatton. or even about the relationship between a
country's stage of economic development and its output of pollution.
Theoretical investigation of these topics has been limited, and empirical
studies are virtually non-existent.
It is useful to distinguish three separate mechanisms by which a change
in trade and foreign investment policy can affect the level of pollution and
the rate of depletion of scarce environmental resources) First, there is a
scale effect, capturing the simple intuition espoused by the environmental
advocates. That is, if trade and investment liberalization causes an
expansion of economic activity, and if the nature of that activity remains
unchanged, then the total amount of pollution generated must increase. The
environmental groups point, for example, to the deleterious environmental
consequences of the combustion of fossil fuels and to the air pollution that
is generated by the trucking industry. To the extent that economic growth
gives rise to an increased demand for energy, which then is generated by means
similar to the prevailing methods, there will be an increased output of
harmful pollutants that attends an increase in economic output. Similarly, to
A similar decomposition of the effects of economic growth on the output
of pollution has been proposed by the Task Force on the Environment and the
Internal Market (1990).
4
the extent that expanded trade gives rise to an increased demand for crossborder
transportation services without there being any change in trucking
practices, increased trade will contribute to a deterioration in air quality.
Second, there is a composition effect that results from any change in
trade policy. When trade is liberalized, countries specialize to a greater
extent in the sectors in which they enjoy competitive advantage. If
competitive advantage derives largely from differences in environmental
regulation, then the composition effect of trade liberalization will be
damaging to the environment. Each country then will tend to specialize more
completely in the activities that its government does not regulate strictly.
and will shift out of production in industries where the local costs of
pollution abatement are relatively great. On the other hand, if the sources
of international comparative advantage are the more traditional ones, namely
cross-country differences in factor abundance and technology, then the
implications of the composition effect for the state of the environment are
ambiguous. Trade liberalization will lead each country to shift resources
into the sectors that make intensive use of its abundant factors, The net
effect of this on the level of pollution in each location will depend upon
whether pollution-intensive activities expand or contract in the country that
on average has the more stringent pollution controls.
Finally, there is a technicue effect. That is, output need not be
produced by exactly the same methods subsequent to a liberalization of trade
and foreign investment as it has been prior to the change in regime. In
particular, the output of pollution per unit of economic product need not
remain the same. There are at least two reasons to believe that pollution per
unit of output might fall, especially in a less developed country. First,
S
foreign producers nay transfer modern technologies to the local economy when
restrictions on foreign investment are relaxed. More modern technologies
typically are cleaner than older technologies due to the growing global
awareness of the urgency of environmental concerns. Second, and perhaps more
importantly, if trade liberalization generates an increase in income levels,
then the body politic may demand a cleaner environment as an expression of
their increased national wealth. Thus, more stringent pollution standards and
stricter enfoccement of existing laws may be a natural political response to
economic growth.
In this paper we explore some of the empirical evidence that bears on
the likely environmental impacts of a NAflA. In Section 1, we shed some light
on the relative magnitudes of the scale and technique effects. We use a
cross-country sample of comparable measures of pollution in various urban
areas to explore the relationship between economic growth and air quality.
After holding constant the identifiable geographic characteristics of
different cities, a common global time trend in the levels of pollution, and
the location and type of the pollution measurement device, we find that
ambient levels of both sulphur dioxide and dark matter suspended in the air
increase with per capita GD? at low levels of national income, but decrease
with per capita GD? at higher levels of income. The turning point comes
somewhere between $4,000 and $5,000, measured in 1985 U.S. dollars. For a
third measure of air quality, namely the mass of suspended particles found in
a given volune of air, the relationship between pollution and GD? is
monotonically decreasing.
Sections 2 and 3 address different aspects of the composition effect.
In section 2 we ask whether and to what extent the sectoral patterns of U.S.
6
foreign investment in Mexico and of Mexican exports to the United States are
affected by the laxity of environmental regulations in Mexico as compared to
the stricter enforcement of controls in the United States. We relate the
sectoral pattern of maquiladora activity, of U.S. imports from Mexico under
the offshore assembly provisions of the US. tariff codes, and of total US.
imports from Mexico to industry factor intensities, U.S. tariff rates, and the
size of pollution abatement costs in the U.S. industry. We find that the
traditional determinants of trade and investment patterns are significant
here, but that the alleged competitive advantages created by lax pollution
controls in Mexico play no substantial role in motivating trade and investment
flows.
Finally, in Section 3, we begin with the premise that resource
allocations in the United States, Mexico, and Canada have been guided by
competitive advantages generated by differences in factor endowments. We
borrow from Brown, Deardorff and Stern (1991) their estimates of the change in
resource allocation that might result from a NAFTA, and discuss the
implications of these predicted changes in the structure of production for
levels of pollution in each country.
1 Economic Growth and Urban Air Pollution
As we noted in the introduction, economic growth has offsetting
implications for the anthropogenic generation of air pollution. On the one
hand, some pollutants are & natural byproduct of economic activities such as
electricity generation and the operation of motor vehicles. As economic
activity expands, emissions of these pollutants tend to grow. On the other
hand, firms and households can control their pollution to some degree by their
7
choice of technology. Cleaner technologies produce less pollution per unit of
output. As a society becomes richer its members may intensify their demands
for a more healthy and sustainable environment, in which case the government
may be called upon to impose more stringent environmental controls,
Little is known about the empirical relationship between national income
and concentrations of various poLlutants. Investigation of this issue has
been hampered by the paucity of data on air pollution that is available on a
comparable basis for a representative sample of countries. However, since
1976 the World Health Organization (WHO) has collaborated with the United
Nations Environment Programme in operating the Global Environmental Monitoring
System (GEMS). The goal of this project has been to monitor closely the
concentrations of several pollutants in a cross-section of urban areas using
standardized methods of measurement. This data set, which to our knowledge
has not previously been analyzed by economists, provides us with an
opportunity to examine how air quality varies with economic growth.4
In the next subsection we describe the GEMS project, the types of
pollution that it monitors, and the data that it has generated. Section 1.2
gives the details of the statistical analysis that we have performed. Our
findings are presented in Section 1.3 and the implications for Mexico are
discussed in Section 1,4.
The GEMS data have been statistically analyzed by some enviroTuhlental
scientists (see World Health Organization (1984]), but they have neglected
to use
any economic variables in their exclusively bivariate analyses.
8
1.1 The GEMS Data3
The GEMS monitors air quality in urban areas throughout the world.
Daily (or, in some cases, weekly or less frequent) measurements are taken of
concentrations of sulphur dioxide (SO) and suspended particulate matter
Data on particulates, which are gases and liquids suspended in the air, are
collected by different methods (described further below) that alternatively
measure the mass of materials in a given volume of air and the concentration
of finer, darker matter, sometimes referred to as "smoke".
Sulfur dioxide is a corrosive gas that has been linked to respiratory
disease and other health problems.' It is emitted naturally by volcanoes,
decaying organic matter, and sea spray. The major anthropogenic sources of
SO are the burning of fossil fuels in electricity generation and domestic
heating, and the smelting of non-ferrous ores (World Resource Institute,
1988). other sources in some countries include automobile exhaust and the
chemicals industry (Kormondy, 1989). Sulfur dioxide emissions can be
controlled by the installation of flue gas desulfurization equipment
(scrubbers) on polluting facilities, and by switching electricity-generating
and home-heating capacity to lower sulfur grades of coal or away from coal
altogether.
Particulates arise from dust, sea spray, forest fires, and volcanoes.
Most of these naturally produced particles are relatively large. Finer
The GEMS data for 1977.1984 are published by the World Health Organization
in the series Mr quality in Selected Urban Areas. Unpublished data for 1985-
1988 have been kindly provided to us by Gardener Evans of the U.S. EPA.
6 Lave and Seskin (1970) find for example, that variation in SO2 and
population density together explain two-thirds of the variation
in death from
bronchitis in a sample of U.S. cities-
9
particles are emitted by industry and from domestic fuel combustion (World
Resources Institute, 1988). Larger particles reduce visibility but have a
relatively minor health impact, whereas the finer particles can cause eye and
lung damage and can aggravate existing respiratory conditions (U.S. EPA,
1982). Particulate emissions from anthropogenic processes can be reduced via
the installation of control equipment and by switching to fuels that, when
burned, emit fewer particles.
The GEMS sample of cities has been changing over time. Sulfur dioxide
was monitored in 47 cities spread over 28 different countries in 1977, 52
cities in 32 countries in 1982, and 27 cities in 14 countries in 1988.
Measurements of suspended particles were taken in 21 cities in 11 countries in
1977, 36 cities in 17 countries in 1982, and 26 cities in 13 countries in
1998, while data for darker matter (smoke) are available for 18 cities in 13
countries for 1977. 13 cities in nine countries for 1982, and seven cities in
four countries for 1988. In all, there are 42 countries represented in our
sample for 502, 19 countries in our sample for dark matter, and 29 countries
in our sample for suspended particles. The participating cities are located
in a variety of developing and developed countries and have been chosen to be
fairly representative of the geographic conditions that exist in different
regions of the world (Betmett et at., 1985). In most of the cities included
in the project, air quality measurements are taken at two or three different
sites, which are classified either as center city or suburban, and as
commercial, industrial, or residential. Multiple sites in the same city are
monitored in recognition of the fact that pollutant concentrations can vary
dramatically with local conditions that depend in part upon land use.
Observations at most sites are made on a daily basis and the data set includes
10
measures of the mean, median, 80th, 95th, and 98th percentile of daily
observations in a given site for a given year.
Sulfur dioxide concentrations have been determined by a number of well
accepted methods (see WHO, 1984). The reliability of these methods has been
checked in independent studies, and an intercomparison exercise was performed
using one particular method as a reference point (Bennett et al., 1985). It
was concluded that the measurements by alternative methods are roughly
comparable, although particular meteorological conditions can affect the
various methods differently. With these results in mind, we have chosen to
pool our sample of observations of 502 concentration, but to allow for a dummy
variable to reflect the method of measurement at each site.
Suspended particles are measured by two main methods. High volume
gravimetric sampling determines the mass of particulates in a given volume of
air while the smoke-shade method assesses the reflectance of the stain left on
a filter paper that ambient air has been drawn through. The former method
measures the total weight of suspended particles while the latter is
predominantly an indication of dark material in the air. Since the two
methods yield incomparable measures that capture different aspects of
particulate air pollution, we treat the data generated by gravimetric and
smoke-shade methods separately in our analysis.7
Table I provides the mean, median and standard deviation for the 50th
and 95th percentiles of daily observations in our sample of cities for each of
A few sites used nephelometric methods to measure suspended particles;
i.e., they measured the light loss due to scattering when a light beam is passed
through a sample of particle-laden air. This method gauges the mass of suspended
particles, much as does the high volume gravimetric method. Since the estimates
are comparable in many cases, we pooled the observations from these two types of
instruments, but included a dummy variable to allow for device-specific
measurement differences.
11
the three types of pollution. Figure 1 displays the
corresponding histograms.
The median of daily observations on SO2 range from a minimum of zero to a
maximum of 291 micrograms per cubic meter (pg n13) of air, whereas the 95th
percentile of daily measures range from zero to 1022 pg m3.8 These numbers
can be compared with the World Health Organization recommendation that annual
average SO2 concentrations ought not to exceed 40-60 pg m and that 98th
percentile concentrations ought not to exceed 100-150 pg m3. The median of
daily observations for suspended particles varied from zero to 715 pg nC3
while that for the 95th percentile observation ranged from [5 to 1580 pg
The WHO guidelines for suspended particles list 60-90 pg nC3 as the safe limit
for the annual mean and 150-230 pg as the safe limit for the 98th
percentile. Finally, the median of daily observations of dark matter (or
smoke) in the sample of sites varied from zero to 312 pg m3, while the 95th
percentile observation varied from two to 582 pg m3. The WHO recommends that
dark matter not exceed 50-60 pg & in annual average and 100-150 pg nC3 in the
98th percentile of daily observations.
1.2 Estimation
Concentrations of pollutants in the air depend upon the amounts that are
emitted by natural and anthropogenic sources and on the ability of the
atmosphere to absorb and disburse the gases or particles. Thus, our analysis
of the relationship between growth and air quality must allow for an influence
of city and site characteristics on the observed concentrations of the various
pollutants in addition to the dependence on national product.
a Actlly, SO concentrations are never literaLly zero, but the machines
are unable to detect very low levels of the gas.
12
We have sought to explain the median and 95th percentile of daily
observations for SO2 suspended particles (gravimecric and nephelomecric
methods) and dark matter (smoke-shade method). As explanatory variables, we
have included functions of per capita GOP in the country where the site is
located, characteristics of the site and city, and a time trend. We used the
Summers and 1-leston (1991) data for per capita CD?, which attempt to measure
output in relation to a common set of international prices. Initially, we
allowed the coefficient on per capita CD? to vary across income ranges by
including a dummy variable in our regressions for each $2,000 interval of per
capita GDP. These relatively unrestricted regressions suggested that a cubic
function of per capita CD? would fit the data fairly wellS The cubic
equations are the main focus of our subsequent analysis.9
In the equation for concentrations of SO2, we included dummy variables
for the location within the city (central city or suburban) and for the land
use of the area near the testing site (industrial, commercial, or
residential). We also included a dummy variable for the method of measurement
<gas bubbler or otherwise). Another dummy indicated whether the city was
Located along a coastline or not (reflecting the disbursement properties of
the local atmosphere). We included a variable for the population density of
the city and a dummy variable for whether the city was located in a country
ruled by a Communist government.'6 Finally, a linear time trend was included
Co allow for the possibility that pollution has been abating (or worsening)
We also estimated equations in which we entered per capita GDP in
quadratic form. In general, the quadratic equations do not fit quite as well as
the cubic equations, though in many cases the shape of the estimated relationship
between income and pollution is found to be roughly the same.
' Population densities were collected from several different sources.
These sources and other details of our data set are available upon request.
13
worldwide, in response to increased global awareness or for other reasons)'
The regressions for suspended particles and dark matter included a
similar sat of right-hand-side variables, except that we did not include a
dummy variable for the method of monitoring dark matter, because all
measurements were taken in the same way. Since dust is an important natural
source of particulate matter, we included as an additional explanatory
variable a dummy that indicates whether the measurement site is located within
100 miles of a desert.'2
Some commentators have argued that a country's level of pollution might
be directly related to its openness to international trade, perhaps becAuse
environmental regulations tend to a least common denominator. To test this
hypothesis, we estimated one set of regressions in which we included the trade
intensity of the country in which the site is located (ratio of the sum of
exports and imports to CDP) as a separate determinant of the concentration of
the air pollutants)3
For each pollutant, we estimated a random effects model, allowing for a
component of the error term that is common to a given year's observations at
different sites located in the same city." We find that the variance of the
We began our analysis with separate dummy variables foreach year in our
sample, but the estimates of this model strongly suggested a simple, linear time
trend.
12
However, we coded the desert dummy variable as zero for Cordoba.
Argentina. in view of the fact that a mountain range lies between this city and
the nearby desert. The regressions fit somewhat better
with Cordoba treated this
way, although none of our conclusions about the relationship of particulate
pollution to CDP depends upon this designation.
The data on trade intensities were taken from the World Bank database.
That is, if is the total residual in the equation for some pollutant
at site i in city j at time t, we assume that pj — ait + where ait is the
common.to.the-city component, is the idiosyncratic component, and
14
common-to-the-city component of the estimated residuals is relatively large in
comparison to the idiosyncratic (site-by-tine) component, so that ordinary
least squares would give an inconsistent estimate of the standard errors of
the regression. We also calculated one set of estimates that allowed for
fixed site effects, In other words, we included a separate dummy variable for
each of the different sites in our various samples - The fixed effects were
intended to capture the unobservable topographical and meteorological
conditions at a site that might contribute to its ability to absorb or
disburse pollution. Of course, when we included the fixed site effects we
dropped the dummy variables reflecting the location of the site within the
city and the location of the city on a coastline or near a desert, since these
influences were no longer separately identified. The model with fixed site
effects provides an especially stringent test of the relationship between
national income and pollution, inasmuch as it ignores all information
contained in the cross-country variation in pollution levels and relies
instead only on the variation in air quality that resulted at the various
sites from changes in CDP during the twelve years of GEMS observations (and
then only that part of the variation that cannot be explained by a common
linear time trend).
1.3 Findings
The results of our various estimates of the random effects models for
the three pollutants are given in Tables 2-4. These regressions do not
include the model with fixed site effects, which we discuss separately below.
— 0. Our estimation takes into account the unbalanced nature of this
panel data set.
15
Figure 2 displays, for the median of daily observations on each
Pollutant, the estimated coefficients on the dunimy
variables indicating
whether a country1s per capita COP falls in the range from $2,000 to $3999
from $4,000 to $5,999, and so on.
The coefficient estimates have been plotted
above the midpoint in the range; e.g.. above $3,000 for CDP in the range frort
$2,000 to $3999. These coefficients should be
interpreted as indicating the
amount of extra pollution a.country with a per capita GD?
in a given range is
likely to have, holding constant the values of other
explanatory variables
(site location, city population density, etc.), relative to a country with a
per capita COP in the range from zero to $2,000. The figure also shows the
estimated amount of extra pollution that is associated
with a given level of
per capita COP (relative to a country with a per capita GDP of $1,000) that
comes from regressing the pollution concentrations on per capita GD?, per
capita COP-squared, per capita COP-cubed, and the remaining
explanatory
variables. The figure shows that, in each case, the cubic functional form
approximates well the shape of the relationship between pollution and GOP that
is indicated by the less restrictive regressions,
Figure 3
depicts the estimated relationship between per capita GD? and
S0, derived form the cubic equations, for both the 50th and 95th percentile
of daily observations, For both measures, the concentration of 502 rises with
per capita CDP at low levels of national income, falls with per capita CDP in
the broad range between $5,000 and $14,000 (1985 U.S. dollars), and then
levels off or perhaps begins to rise again.'1 The turning point in the
There are only two countries in our sample (the United States and Canada)
with per capita incomes in excess of $16,000, so the fact that the estimated
curves turn upward in this range probably should not be viewed as strong evidence
for a renewed positive relationship between national product and SO2 pollution
at high income levels,
16
predicted relationship for the median of daily observations comes at $4,119,
while that for the 95th percentile observation occurs at $4,630. We estimate
that a country with per capita CD? of $5,000 will have a 20 pg nC3 greater
concentration of 502 for the 95th percentile of its daily observations, as
compared to a country with a per capita CD? of $1,000, all else equal.
Table 2 indicates that the hypothesis that SO2 pollution is unrelated to the
level of CD? can be rejected at the 1/100th of one percent significance level
in the regression for the median observation and at the seven percent level in
the regression for the 95th percentile observation (see columns 2 and 5).
Table 2 reveals that several other variables contribute significantly to
the cross-city variation in concentrations of SO2. For example, cities
located on a coastline are estimated to have lesser concentrations of S0:
6.68 pg m3 lower for the median of daily observations and 46.79 pg m3 lower
for the 95th percentile observation. Concentrations of 502 are higher in the
center city than in the suburbs, lower in residential areas than in commercial
areas, and higher in industrial areas than in commercial areas (although this
effect is not statistically significant at conventional significance levels).
More densely populated cities suffer greater concentrations of SO2, all else
equal. We also find that SO2 pollution has been significantly greater in
cities Located in Communist-ruled countries. Finally, we note that 802 levels
have been trending downward in our sample of cities even after controlling for
the effects of income and other variables. The downward trend may reflect an
increasing global awareness of the health probLems associated with SO2, and
the expanding efforts that are being made worldwide to limit suLfur emissions.
Columns 3 and 6 of Table 2 present estimates of a model for 802
determination that includes trade intensity as an additional explanatory
17
variable, Contrary to the fears of some environmentalists, we find that 502
Levels are significantly lower in cities located in countries that conduct a
great deaL of trade (relative to their GOP). We have no good economic
explanation for this finding.
Figure 4 depicts the estimated (cubic) relationship between dark matter
and per capita CDP for the median and 95th percentile of daily observations.
Apparently, the nature of the relationship is much the same as for 502. The
concentration of smoke in the air rises with per capita GOP at low levels of
income, peaks at around $5,000 (1985 U.S. dollars). and faLls with GOP at
higher income levels until it eventually levels off. We see in columns 2 and
5 of Table 3 that the three CD? variables are jointly significant in the
determination of dark matter pollution at the 1/100th of one percent
significance level. Moreover, the size of the estimated effects are quite
large. We estimate that a country with a per capita CD? of $5,000 will have a
higher concentration of smoke by about 90 pg in its median of daily
observations and 220 pg m3 in its 95th percentile observation, compared to
one with a per capita COP of $1,000. Recall that the WHO recommends that
concentrations of smoke not exceed 50-60 pg m for the mean observation and
100-150 pg of3 for the 98th percentile observation.
Not surprisingly, the dummy variable indicating proximity to a desert
has a positive and significant coefficient in the regressions explaining
concentrations of dark matter. A location on a coast reduces a citys
concentration of this type of pollution, and again the effect is statistically
significant. We find that smoke pollution levels are greater in center cities
than in suburbs, and smaller in residential areas than in commercial areas.
Also, dark matter, like 502, rises with population density, although this
18
effect is significant only in the regression for the median of daily
observations. Finally, there appears to be neither a global trend in this
type of pollution (once the upward movement in world incomes has been
accounted for) nor a significant association with trade intensity.
Unlike the other two pollutants, the mass of suspended particles in the
air appears to fall in response to increases in per capita GD? at low levels
of economic deveLopment (see Figure 5). This relationship continues until per
capita GOP reaches about $9,000, whereupon economic growth has no further
effect on the concentration of suspended particles. Again, the estimated
effects are large in comparison to the WHO guidelines, and again the three CD?
variables are jointly significant in the determination of this measure of air
quality.
As with dark matter, cities situated near to a desert are likely to
experience higher concentrations of suspended particles than cities located
elsewhere. This effect is both quantitatively large and highly significant.
The coefficients on the Communist dummy, the center-city dummy and the
industrial area dummy all are positive and statistically significant in the
regressions for both the median and 95th percentile of daily observations.
The global trend in suspended particle pollution apparently has been downward,
although the coefficient on the year variable is statistically significant
only in the regressions for the median of daily observations. Finally, a
country's trade intensity has a small and statistically insignificant effect
on this form of pollution.
Table 5 reports the estimated coefficients from a random effects model
for each type of pollutant that also allows for fixed site effects. The
numbers of site dummy variables that were included in the various regressions
19
are shown in the table. These estimates of the relationship between per
capita GD? and the various measures of air quality rely only on the
covariation between GD? and concentrations of the pollutants over time within
the individual sites, and not on the cross-country variation in pollution and
GD? at a given moment in time. The estimated relationships for the median of
daily observations of s2 and dark matter hold up remarkably well. (see Figure
6). In each case, the data continue to indicate an inverted-U shaped
relationship between pollution and national income with peak levels of
pollution occurring for per capita incomes in the range from $4,000 to $5,000.
Only the estimated coefficients from the fixed site effects modeL for
suspended particles suggest a different relationship between per capita GD?
and concentrations of the pollutant than is found in the regressions without
site effects. In this case, the estimation indicates a monotonically
increasing relationship between particulate pollution and national output in
the sample range of output levels.
1.4 Implications for Mexico
Unfortunately, Mexico has not participated in the GEXS project and
reliable measures of its air pollution are not available (US GAO, 199lb).
Thus, predictions for Mexico must be inferred
from relationships that hoLd in
other countries at similar stages of development. Surely
the available
evidence suggests that air quality has deteriorated with economic growth in
Mexico (US GAO, l99lb). Our estimates indicate
that this experience is common
in poor countries, but that the positive association between two pollutants
and economic output ceases when the typical country reaches a per capita
income level of about $5,000 (1985 U.S.
dollars). We note that Summers and
20
Heston (1991) put Mexico's per capita CDI' in 1988 at $4,996. Thus, we might
expect that further growth in Mexico, as may result from a free trade
agreement with the United States and Canada, wild lead the country to
intensify its efforts to alleviate its environmental problems.
Recent measures taken by the government of Mexico suggest that the
country already may have reached the turning point in tens of air pollution.
In the last year, the Salinas government has reduced the lead content of
petrol, ordered several power stations to burn natural gas instead of sulfurgenerating
fuel oil, and shut down oil refineries and private firms that were
found to be major sources of air pollution (The Economist. May 18, 1991).
Also, new cars are being fitted with catalytic convertors, a new fleet of
cleaner buses has been introduced, and drivers have been banned from using
their cars in Mexico City one day each week. To beef up enforcement, the
budget of the environmental protection ministry has been increased sevenfold
(New York Times, Sept. 22, 1991). Further growth may enable the government to
implement fully its planned $2.5 billion, four-year program to clean up Mexico
City.
2 Pollution Abatement Costs and the Pattern of U.S. - Mexico Investment
and Trade
A main source of concern about a NAFTA is that it will enable firms to
circumvent U.S. environmental protection laws. If the costs of meeting
pollution controls are high in the United States and low or negligible in
Mexico, then the asymmetry in standards or enforcement efforts can create a
competitive advantage for Mexican producers and can motivate U.S. fins to
relocate their production facilities south of the border. In these
21
circumstances liberalization of trade and investment flows can strengthen the
incentives for "environmental dumping
A number of authors1 including for example Pethig (1976), Siebert
(1977). 'lobe (1979) and KcGuire (1982), have studied the theoretical
relationship between environmental regulations and the pattern of trade. They
find that strict environmental standards or costly controls weaken a country's
competitive position in pollution-intensive
industries and diminish its
exports (or increase its imports) of the product of such sectors. Countries
that fail to regulate industrial pollutioo increase their specialization
in
activities that damage the environment. Kccuire has extended these results to
include direct foreign investment: controls cause firms active in the
pollution-intensive industry to relocate their activities to the less
regulated countries.
While these theoretical predictions are plausible and intuitive, they
have found little support in previous empirical
studies of trade and
investment patterns. For example, Tobey (1990) has tested the hypothesis that
environmental regulations have altered the pattern of trade in goods produced
by "dirty" industries.
lie finds that a qualitative variable describing the
stringency
of environmental controls in 23 countries fails to contribute to
the determination of their net exports of the five most pollutionintensi'le
commodities. Similarly. Walter (1982) and Leonard (1988) conclude that there
is little evidence that pollution abatement costs
have influenced the location
decisions of multinational firms. Apparently,
the cross-country variation in
the costs of meeting environmental controls is not so
large as to be a major
factor in the determination of nations' comparative advantages.
Our purpose in this section is to address
this issue in connection with
22
the pattern of U.S. imports from Mexico and the pattern of U.S. foreign
investments in Mexico- There is some evidence from a GAO survey suggesting
that a few American furniture manufacturers may have moved their operations to
Mexico in response to the State of California's tightening of air pollution
control standards for paint coatings and solvents (US CAO, l991c). But the
question remains open as to whether the overall sectoral pattern of U.S.
economic relations with Mexico has been meaningfully affected by the higher
costs of pollution abatement in the United States. If the pattern of
specialization has been so influenced, then the composition effect of a
further liberaLization of trade and investment may be damaging to the
environment.
Using data from the Bureau of Census' 1988 survey of pollution abatement
costs in American industries (U.S. DOC, 1988). we have conducted three sets of
tests. We have studied the 1987 pattern of U.S. imports from Mexico in 3digit
SIC manufacturing industries, the pattern of 1987 U.S. imports from
Mexico that have entered the country under the offshore assembly provisions
(again at the 3-digit SIC level), and the sectoral pattern of value added by
maquiladora plants in (approximateLy) 2-digit industry categories. In each
case we have investigated whether pollution abatement costs in the United
States help to explain the pattern of Mexican specialization and trade.
Our estimates of the determinants of the pattern of manufactured imports
from Mexico are recorded in the first two columns of Table 6. We use the
ratio of 1987 U.S. imports from Mexico to total U.S. shipments in the same
industry as the dependent variable. The explanatory variables include factor
shares (reflecting the intensity of use of the various factors by the
different industries; see Harkness (19781), the U.S. effective tariff rate on
23
imports of goods in the industry, and the ratio of pollution abatement costs
(operating expenses) to total value added in the U.S. industry.16 1e also
report another regression that includes the average injury rate in the
industry as an additional independent variable. Since firm outlays for worker
compensation insurance are roughly proportional to injury rates (Krueger and
Burton, 1990). the injury variable proxies for one (large) component of the
cost to American manufacturing firms of U.S. labor protection laws.
We computed the factor shares as follows. Je took the payroll expenses
in an industry to represent a combined compensation for unskilled or "raw"
Labor and human capital. Payments to unskilled labor were defined as the
product of the number of workers in the industry and the economy-wide average
yearly income of workers in manufacturing with less than a high school
education. We formed a share by dividing this amount by value added, and
considered the remaining part of the total labor share to represent the
payment to human capital.11 Finally, we calculated the share of capital in
value added as the difference between one and the total payroll share.t619
16
The Cen5us survey did not include the apparel industry (SIC category 23).
For these observations and four others with missing data, we inserted the average
ratio of pollution abatement costs to value added for all manufacturing sectors
included in the survey.
For tvelve industries, this method gives a negative number as the
estimate of the share of human capital. To ensure that our results were not
sensitive to the choice of income for an unskilled worker, we also computed
factor shares using the income level for unskilled workers ($10,819) that wade
the minimum share of human capital our sample of industries equal to zero.
The
estimated import equations with these measures of factor shares look much the
same as for our original measures.
16 The import figures and the effective tariff rates were provided to us by
Greg Schoepfle of the U.S. Department of Labor. The tariff rates were estimated
by dividing the total duties collected on imports from Mexico in a two digit SIC
industry by the total value of imports in the industry. Data On shipments,
value added, employment, and payroll were taken from
the N.B.E.R. trade and
immigration data set (see Abowd, 1987). Since the trade data are classified
24
Factor intensities figure prominent].y in the determination of the
pattern of U.S. imports from Mexico. The ratio of imports to total US.
shipments is smaller in industries that are highly intensive in their use of
human or physicat capital. This means, of course, that Mexico exports to the
United States goods that have a relatively high share of unskilled Labor in
total factor cost. The coefficients on the physical. and human capital
variables both are statistically significant at the five percent level, but
the latter coefficient is quantitatively much larger. We estimate that a ten
percentage point increase in the share of human capital reduces the ratio of
imports to shipments by 0.52 percentage points, while a ten percentage point
increase in the share of physical capital lowers the import ratio by only 0.24
percentage points. Note that the mean ratio of imports to shipments across
all manufacturing industries is 0.69 percent.
We estimate that the import ratio rises with the share of pollution
abatement costs in U.S. industry value added, as would be predicted by a model
of environmental dumping. However, the impact of these regulatory costs is
according to the old (1972) SIC categories, we used manufacturing census data
that were bridged to this classification scheme. The average income for a worker
in the manufacturing sector with less than a high school education was calculated
from the 1987 Current PQvulation Survey by the authors. Injury rates by industry
were taken from the Eureau of Labor Statistics publication. Occunational Injuries
and Illness in the United States by Industry. In the few cases where an injury
rate was not available for a three digit industry, we used instead the average
injury rate for the applicable 2-digit industry.
' We are also experimenting with a second procedure for dividing the labor
share into component parts representing the shares of unskilled labor and human
capital. In this, we form factor share for different skill groups (workers with
a high school education, workers with some college education, and workers with
a college degree) by taking the product of the industry labor force, the fraction
of the industry's workers in the skill group in question, and the economy-wide
average income of workers in that skill group, and then dividing by value added.
If this procedure yields substantially different conclusions, we will report on
the results in a future draft of this paper.
25
both quantitatively small and statistically Insignificant Consider for
example an industry that has the Dean ratio of pollution abatement costs to
industry value added and another that has a ratio that is two standard
deviations higher. We estimate that the latter industry will have a greater
ratio of imports to U.S. shipments by about 0.05 percent, which is less than
1/20th of a standard deviation in the import variable. This
finding can be
understood from the fact that pollution abatement costs
average only 1.38
percent of value added across all manufacturing industries and rise to
only
4.85 percent in an industry that is two standard deviations above the mean.
The implied variation in competitiveness is small in comparison with that
which arises from cross industry variations in Labor costs, for example.
We note that the injury rate has a positive coefficient when it is
included in the import equation, but this variable too has a very small and
statistically insignificant effect. Finally the coefficient on the tariff
variable has the theoretically predicted negative sign (imports are smaller in
industries with high effective tariff rates), but also is not significant.2°
We turn next to the determinants of the pattern of U.S. imports from
Mexico under the offshore assembly provision (i.e., import category 207.00 in
the old TSUSA tariff schedule). U.S. trade law provides for duty free reentry
of American-made components embodied in imported final goods. In cases
where intermediate products are exported for further processing or assembly
abroad, the applicable tariff rate applies only to the foreign value added.
Nearly 44 percent of the value of Kexican exports to the United States
qualified for this customs treatment in 1987 (Schoepfle, 1991). We study
20
A simultaneity bias may exist here, insofar as many political-economic
theories of tariff formation predict that high tariff rates will endogenously
emerge in industries in which import penetration is great.
26
these imports separately, because much of the
output by maquiladora plants
enters the United States in this manner, and maquiladoras
are the source of
most of the item 807.00 U.S. imports from Mexico (Schoepfle.
1991). Thus, the
sectoral pattern of item 807.00 imports gives us an idea as to the pattern of
maquiladora activity. The
middle two columns of Table 6 reports estimates of an equation for
the ratio of the Mexican value added in imports entering under the 807.00 code
in 1987 to the total value added by the U.S. industry. We would
expect this
variable to be high in industries where maquiladoras are
especially active;
i.e. • those in which Mexico enjoys a competitive advantage. The independent
variables in these equations are the same as before, namely the factor shares,
the share of pollution abatement-costs in U.S. industry value added, and in
one set of estimates, the industry injury rate. The tariff variable was
omitted from these regressions because our effective tariff rate applies to
all imports from Mexico, and does not reflect the average tariff paid on
imports that entered under the offshore assembly provision. We use a tobit
model to estimate these equations in view of the fact that item 807.00 imports
are zero in 58 of the 136 Industry categories.- Since the import share cannot
be negative, a censored regression model is appropriate.
The foreign content of U.S. item 807.00 imports from Mexico is highest
in relation to total value added in the corresponding U.S. industry in sectors
that make relatively intensive use of unskilled labor. This is not
surprising, since many U.S. manufacturers attempt to outsource to maquiladora
operations the most unskilled labor intensive phases of the production
process. We find a negative association between the Mexican content of item
807.00 Imports as a fraction of total U.S. industry value added and both the
27
human capital share and the physical capital share in U.S. industry costs.
The coefficient on the physical capital variable is estimated to be nearly
twice as large in absolute value as that on the human capital variable
(reversing the ordering found for total imports from Mexico), and the former
coefficient is statistically significant at the five percent level whereas the
latter is statistically significant at only the ten percent level.
Again, we fail to find a significant positive relationship between the
size of pollution abatement costs (as a fraction of value added) in the U.S.
manufacturing industry and the scale of sectoral activity in Mexico. In fact,
the foreign content of item 807.00 imports from Mexico appears to be lower in
relation to the size of the U.S. industry in sectors where U.S. pollution
abatement costs are relatively high. The negative coefficient on the
abatement cost variable is found to be statistically significant at the five
percent level, although this may of course reflect a spurious correlation
between these costs and some omitted variable. When the U.S. injury rate is
included in the equation for item 807.00 imports, the estimated coefficient on
this variable is also negative. This would imply that relatively little
Mexican assembly activity takes place in those industries where the cost to
US. employers of workers' compensation insurance and other accident-related
costs are especially high. However, in this case, the estimated coefficient
is not significantly different from zero, so we cannot reject the hypothesis
that the association between injury rates and the pattern of Mexican assembly
operations is nil.
The final set of estimates recorded in Table 6 relate to the activity of
Mexican maquiladoras. The Instituto Nacional de Estadistica Geografia E
Informatica (INECA), a private Mexican research institute, has surveyed all
28
maquiladoras on the government's List of in-bond producers. Their
publication, Eiadistica de Ia Industria Maouiladora de Exoortacion. 1978UM.
provides data on value added by maquiladoras in eleven different
manufacturing industries. We developed a concordance of the available data to
a 2-digit SIC basis and sought to explain the ratio in 1987 of value added by
maquiladoras to value added in the corresponding U.S. industry.2' The
explanatory variables are defined as before, except that now we calculate
factor shares and pollution abatement costs as a fraction of value added at
the 2-digit SIC level.
The estimated coefficients from the models of maquiladora activity
confirm our findings for total U.S. imports from Mexico and for imports under
tariff item 807.00. Again we find that Mexican competitive advantage derives
from an abundance of unskilled labor, and that value added by maquiladoras
declines in relation to value added in the United States the greater are the
shares of human and physical capital in industry cost. Although neither
coefficient has been estimated precisely enough to allow for a clear rejection
of the hypothesis that the associations are zero, the magnitudes of the
coefficients are very much in line with what we have found before. Also, we
find no evidence in support of the hypothesis that U.S. regulatory costs
contribute to the explanation of the pattern of maquiladora activity. Neither
the coefficient on the abatement cost variable nor that on the injury rate
variable has the (positive) sign that one would expect if American firms are
investing in maquiladora plants primarily to avoid the high costs of
al We note that INECA withheld industry-leveL data for industries with
sixteen or fewer maquiladora establishments. Thus, our analysis treats as zero
the maquiladora activity in sectors where some small amount of production may
have taken place.
29
environmental and labor protection laws at home. Evidently, the costs
involved in complying with these laws are small in relation to the other
components of total cost that determine whether it is profitable to operate in
the United States or Mexico.
3 Resource Reallocation: Implications for Pollution
Our findings in Section 2 suggest that relative factor supplies govern the
pattern of trade between Mexico and its neighbors to the North. We might expect,
therefore, that trade liberalization will stimulate Mexican production in
unskilled labor-intensive industries, while the United States and Canada will
shift resources into sectors that make relatively heavy use of capital and
skilled labor. The removal by Mexico of barriers to direct foreign investment
can have the opposite effect on international patterns of specialization, if
foreign firms bring with them the factors that are scarce in Mexico. Then local
production may expand in (moderately) capital and skilled-labor intensive
sectors. The question that arises is, what are the environmental, implications
of these potential resource reallocations?
To answer this question fully, we need several pieces of information.
First, we need to know which sectors will expand in each country, and to what
extent. Second, we need to know the pollution-intensities of the various
industries. Finally, we must know how NAFrA will affect the production
technologies used in each location, so as to gauge any changes in pollution
generated per unit of output. Estimates of NAflA impacts on the production
structure in each country are available from computational modeling exercises.
Brown, Deardorff, and Stern (1991), for example, have predicted resource
movements for several different scenarios of policy change under a NAFTA.
30
Unfortunately, the remaining informational requirements pose more serious
difficulties. Concerning the pollution generated by different industries, the
United States collects data only on releases of toxic waste (and that of
questionable quality), while Mexico coLlects no such data whatsoever. And
although our findings in Section 1 suggest that production techniques might well.
change in response to a trade agreement that generates economic growth in Mexico,
it is difficult to assess how these changes will be distributed across
industries.
In this section we draw upon the detailed estimates of Brown, Deardorff,
and Stern (BBS) to derive some possible environmental implications of the
resource reallocations that would result from a NAFTA. After describing the SOS
modeling exercise, we turn to two issues. First, we discuss the model's
predictions about each country's demand for the services of utilities. Second,
we use the information available in the U.S. EPA's Toxic Resources Inventory to
analyze how a NAFTA might affect releases of hazardous waste by the manufacturing
sectors in the three partner countries.
The BbS estimates are based on a computable general equilibrium model of
the economic interactions between the United States, Canada, Mexico, a group of
31 other major trading nations, and an abbreviated fifth region comprising the
rest of the world. The model aggregates production into 23 categories of
tradable goods and six categories of non-tradable goods and services. The
industries are treated either as perfectLy competitive or monopolistically
competitive, with the latter set exhibiting economies of scale. Output in each
sector is produced from intermediate inputs and an aggregate of capital and
Labor. The authors allow for varying degrees of substitution between capital and
labor in the different sectors, but treat labor as a homogeneous input. The
31
latter assumption is unfortunate in view of our
finding in Section 2 that human
capital endowments play a central role in determining the
bilateral trade pattern
between the United States and Mexico.
Since it is not clear what policy measures will be
included in a NAFTA, BDS
explore a number of alternatives. In one scenario they
assume the removal of all
bilateral tariffs between the United States, Canada and Mexico, and an easing of
U.S. quantitative restrictions that generates a twenty five percent increase in
U.S. imports of agricultural products, food, textiles, and apparel from Mexico.
In a second scenario they allow for these same forms of trade liberalization and
also a relaxation of restrictions on direct foreign investment in hexico. The
liberalization of investment is assumed to result in an (exogenous) ten percent
increase in Mexico's capital stock. It should be noted that in both of these
cases, the estimated impacts include not only the removal of existing barriers
between Mexico and its trade partners, but also the ultimate implementation of
the policy changes that comprise the already concluded Canada-U.S. free trade
agreement.
It is well known that many pollutants, such as sulfur dioxide, nitric
oxide, nitrogen dioxide, and carbon dioxide, are byproducts of electricity
generation, especially when fossil fuels are burned in the process. Thus, an
important determinant of the net effect of trade liberalization on air pollution
will be the induced change in the demand for electricity. Unfortunately, the
disaggregation in the EDS model does not allow us to identify the likely impacts
of a NAFTA on electricity use, inasmuch as the model treats electricity as a
component of the broader category of utilities. It is interesting to note,
nonetheless, that BDS predict a decline of 0.56 percent in output by the Mexican
utilities sector in response to a hypothetical agreement involving an elimination
32
of tariffs and an increase in U.S. import quotas. This prediction can be
understood from the fact that the scenario generates output contractions in ten
of twenty one Mexican manufacturing sectors, and expansions are anticipated
primarily in labor-intensive industries such as food, textiles, apparel, leather
products, and footwear, which presumably are not the sectors that use energy most
intensively. ut just as Mexico will shift resources into activities that
require relatively little energy input, the United States and Canada may be
expected to do the opposite. The model predicts modest increases of 0.07 percent
and 0.09 percent, respectively, in production by utilities in the United States
and Canada.
Quite different conclusions emerge from the scenario that attempts to
capture the effects of a potential liberalization of investment flows. The
exogenous ten percent increase in the Mexican capital stock that is taken to be
the outgrowth of an easing of restrictions on foreign investment effects an
expansion of every manufacturing industry there, with the implication that demand
for utilities must rise. The experiment generates an increase in utilities
output in Mexico of 9.31. percent. Presumably, air quality would deteriorate in
this case, unless the associated thcome growth gave rise to political demands for
more stringent standards and tougher enforcement.
We turn next to toxic waste. Our focus on this form of pollution is
primarily a reflection of data availability. As far as we know, this is the only
type of pollution for which data on releases are collected at the firm level.
U.S. law requires all manufacturing firms with ten or more employees that use at
Least 10,000 pounds of one or more of over 300 chemicals to report their annual
chemical releases to the EPA. Lines of business are recorded along with
information on releases, enabling aggregation of the data to the industry level.
33
Several pitfalls in the use of the data contained in the Toxic Release
Inventory (TRI) should be noted.22 First, since the report reflects releases
rather than exposures, it cannot reflect the great differences that exist in the
rates at which different chemicals are dispersed or transformed. Second,
releases are measured only in terms of weight, and no effort is made to account
for the fact that some high volume releases are relatively benign, while other
lower volume releases may create great health risks. Third, many toxic chemicals
are omitted entirely from the inventory. Fourth, the inventory does not reflect
emissions by firms outside the manufacturing sector or by federal facilities.
Fifth, the EPA conducts relatively little verification of the information it
receives and makes relatively little effort to ensure compliance with the
reporting requirements. Finally, there is no way to know whether the
relationships between toxic waste and industry outputs in Canada and Mexico
mirror those in the United States, or whether there are great differences across
countries in the industry-specific relationships between hazardous waste and
quantities of output.
We have used the predictions of the BDS model and the data on the industry
breakdown of toxic releases by U.S. manufacturing firms contained in the TRI to
generate Table 7. The table contains estimated impacts on total toxic releases
by the manufacturing industries of Mexico, United States, and Canada in response
to two scenarios for a NAFTA. As before, the scenarios distinguish a NAFTA that
effects only trade liberalization and one that also includes investment
liberalization measures in Mexico. In constructing the table we have assumed
that an industry-specific fixed coefficient characterizes the relationship
between the amount of toxic release and the quantity of output produced. We have
22The following discussion is based on U.S. GAO (199la).
34
calculated this coefficient using output and total release data for the United
States in 1989.23 To the extent that releases per unit of output are uniformly
higher or lower in Mexico or Canada than they are in the United States, the
estimates in Table 7 for these countries can be adjusted upward or downward to
reflect these percentage differences. But lacking data for Canada and Mexico,
we cannot address the possibility that these countries have higher or lower
pollution coefficients than the United States in some industries but not in
others.
The table tells a similar story to the one told by the utilities sector.
A liberalization of trade in the absence of increased capital flows causes Mexico
to shift resources toward industries that, on average, generate less pollution
than its current producers. In particular, the EDS model predicts contraction
by the industries producing chemical products and rubber and plastics products,
both of which generate great quantities of waste per unit of output. The
beneficial environmental effects of these resource flows are largely offset by
an expansion of the electrical equipment industry, but a small positive net
impact remains. The reallocation of resources in the United States and Canada
differ from those in Mexico, as. these countries enjoy comparative advantage in
a complementary set of activities. The model predicts an expansion of the
chemical products industry in the United States, and of the primary metals
industry in Mexico, with the implication that aggregate chemical releases by
manufacturing enterprises will rise in both of these countries.
Again, the scenario that assumes a ten percent increase in Mexico's capital
23The output data, which were provided to us by Dnisilla Brown, are those
that were used in the calibration of the BDS model. Since these data are
reported on an ISIC basis, we were forced to reclassify some industries in order
to make them compatible with the SIC-based TRI data.
35
stock has quite different implications for Mexico. As noted before, such growth
in the capital stock causes output to expand in every Mexican manufacturing
industry. If the relationship between waste and output remains unchanged, then
total chemical releases must rise.
While our estimates in this section must be taken with a large grain of
salt, they suggest conclusions that accord well with intuition. Since Mexico
enjoys comparative advantage in a set of activities (agriculture and labor intensive
manufactures) that on the whole are "cleaner than the average, the
composition effect of trade liberalization way well reduce pollution there. On
the other hand, a NAFTA will cause the United States and Canada to specialize
more in physical and human capital-intensive activities, to the possible (slight)
detriment of their local environments. On the global level, a net benefit may
derive from the movement of the dirtier economic activities to the more highly
regulated production environments.
4 Conclusions
Environmental advocacy groups have pointed to several risks that might be
associated with further liberalization of trade between the United States and
Mexico. While they raise a number of valid concerns, our findings suggest that
some potential benefits, especially for Mexico, may have been overlooked. First,
a more liberal trade regime and greater access to the large U.S. market is likely
to generate income growth in Mexico. Brown, Deardorff and Stern (1991). for
example, estimate potential short run welfare gains to Mexico of between 0.6 and
1.9 percent of CDP. We have found, through an examination of air quality
measures in a cross-section of countries, that economic growth tends to alleviate
pollution problems once a country's per capita income reaches about $4,000 to
36
$5,000 US. dolLars. Mexico, with a per capita GOP of $5,000, now is at the
critical juncture in its development process where further growth should generate
increased political pressures for environmental protection and perhaps a change
in private consumption behavior. Second, trade liberaLization may well increase
Mexican specialization in sectors that cause less than average amounts of
environmental damage. Our investigation of the determinants of Mexico's trade
pattern strongly suggests that the country draws comparative advantage from its
large number of relatively unskilled workers and that it imports goods whose
production requires intensive use of physical and human capital. The asymmetries
in environmental regulations and enforcement between the United States and Mexico
play at most aminor role in guiding intersectoral resource allocations, But
since it would appear that labor-intensive and agricultural activities require
less energy input and generate less hazardous waste per unit of output than more
capitaL and human capital-intensive sectors, a reduction in pollution may well
be a side-benefit of increased Mexican specialization and trade.
Our findings must remain tentative until better data become available. Ve
have been unable to use any information about the pollution situation as it
currently stands hi Mexico, since environmental monitoring there has been
unsystematic at best. Furthermore, the kinds of pollutants that we can examine
are limited by data availability (eg. , there are no reliable data on emissions
of carbon dioxide in different countries). Still, one lesson from our study
seems quite general and important. The environmental impacts of trade
liberalization in any country will depend not only upon the effect of policy
change on the overall scale of economic activity, but also upon the induced
changes in the intersectoral composition of economic activity
and in the
technologies that are used to produce goods and services.
37
References
Abowd, John M. 1990. "The NBER Immigration, Trade, and Labor Markets Data
Files. National Bureau of Economic Research Working Paper No. 3351, May
Bennett, Burton C. , Kretzschmar, Jan C., Akland, Gerald C. , and de Koning,
Henk W. 1985. "Urban Air Pollution Worldwide." Environmental Science and
Technology 19: 298-304.
Brown, Drusilla K., Deardorff, Alan V., and Stern, Robert K. 1991, "A North
American Free Trade Agreement: Analytical Issues and a Computational
Assessment." Mimeo. Institute of Public Policy Studies, University of
Michigan.
Gregory, Michael. 1991. "Sustainable Development vs. Economic Growth;
Environmental Protection as an Investment in the Future." Statement on Behalf
of Arizona Toxics Information before the International Trade Commission,
Hearing on Probably Economic Effect on U.S. Industries and Consumers of a Free
Trade Agreement Between the United States and Mexico.
Harlcness, Jon. 1978. "Factor Abundance and Comparative Advantage." American
Economic Review 68: 784-800.
Kelly. Mary E. and Kamp, Dick. 1991. "Mexico-1.LS. Free Trade Negotiations
and the Environment: Exploring the Issues." Mimeo. Texas Center for Policy
Studies, Austin TX, and Border Ecology Project, Naco AZ.
Kormondy. Edward J. • ed. 1989. International Handbook of Pollution Control.
New York and Westport. CT: Greenwood Press.
Krueger, Alan B. and Burton, John F., Jr. 1990. "The Employers' Costs of
Workers' Compensation Insurance: Magnitudes. Determinants, and Public Policy."
Review of Economics and Statistics 72: 228-260.
L.ave, Lester B. and Seskin, E. P. 1970. "Air Pollution and Human Health."
Science 169: 723-733.
Leonard, H. Jeffrey. 1988. Pollution and the Strunle for World Product.
Cambridge: Cambridge University Press.
Leonard, Rodney E. and Christensen, Eric. 1991. "Economic Effects of a Free
Trade Agreement between Mexico and the United States.' Testimony on Behalf of
Community Nutrition Institute before the International Trade Commission
Hearing on Docket No. 332-307.
Low, Patrick and Safedi, Raed. 1991. "Trade Policy and Pollution." Mimeo.
International Trade Division, World Bank.
KcGuire, Martin C. 1982. "Regulation. Factor Rewards, and International
Trade." Journal of Public Economics 17: 335-354.
38
National Wildlife Federation. 1990. "Environmenta]. Concerns Related to a
United States-Mexico-Canada Free Trade Agreement. Mimeo. National Wildlife
Federation, Washington DC.
New York Tines. 1991, "Facing Environmental Issues." Sept. 22. Sec. 3: 5,
Ortman, David E, 1991. "On a Comprehensive North American Trade Agreement."
Testimony on Behalf of Friends of the Earth, National Wildlife Federation, and
the Texas Center for Policy Studies before the Subcommittee on Trade,
Committee on Ways and Means, U.S. House of Representatives.
Pethig, Rudiger. 1976. "Pollution, Welfare, and Environmental Policy in the
Theory of Comparative Advantage." Journal of Environmental Economics and
flanatement 2: 160-169.
Schoepfle, Gregory K. 1991. "Implications for U.S. Employment of the Recent
Growth in Mexican Maquiladoras." Frontera None 3: 25-54.
Siebert, Horst. 1977. "Environmental Quality and the Gains from Trade,'"
Kyiclos 30: 657-673.
Summers, Robert and Heston, Alan. 1991. "The Penn World Table (Mark 5): An
Expanded Set of International Comparisons, 1950-1988," Ouarterly Journal of
Economics 106: 327-368.
Task Force on the Environment and the Internal Market. 1990. 1992: The
Environmental Dimension. Bonn: Economica Verlag.
The Economist. 1991. "Smog City." May 18: 51.
Tobey, James A. 1990. "The Effects of Domestic Environmental Policies on
Patterns of World Trade: An Empirical Test." Kykios 43: 191-209.
United States Department of Commerce. 1988. Manufacturers' Pollution
Abatement Capital Exoenditure and Operatint Costs. Washington DC: Bureau of
the Census.
United States Environmental Protection Agency. 1982. Air quality Criteria
for Pnticulate Matter and Sulfur Oxides. Research Triangle Park, NC: ILS.
EPA.
United States General Accounting Office. 1991a. "Toxic Chemicals: EPA's
Toxic Release Inventory is Useful but Can be Improved." Report No. GAO/RCED-
91-121, Washington DC.
United States General Accounting Office. 199lb. "U.S. - Mexico trade:
Information on Environmental Regulations and Enforcement," Report No,
GAO/NSIAD.91-227, Washington DC.
United States General Accounting Office. 199lc. "U.S. - Mexico Trade: Some
U.S. Wood Furniture Firms Relocated from Los Angeles Area to Mexico," Report
No. CAO/NSIAD-91.19l, Washington DC.
39
Walter, Ingo. 1982. "Environmentally Induced Industrial Relocation to
Developing Countries," In S.J. Rubin and T.R. Graham, eds. Environment and
Trade. New Jersey: Allanheld, Osmun, and Co.
World Health Organization. 1984. Urban Air Pollution. 1973-1980. Geneva:
WHOWorLd
Resources Institute. 1988. World Resources: 1988-59. New York: Basic
Books.
Yohe, Gary W. 1979. The Backward Incidence of Pollution Control - Some
Comparative Statics in General Equilibrium." Journal of Environmental and
Economic Mana2ement 6: 187-198.
20
ii
01
Figure 1: Histograms of Air Pollutants
• Ii
02
a.
—
.06
0'
0Z
6 760 ida od, ado an'.. I I no eo'o
g 0•t Coic N.tot
Histogram of 95th PctL. at Susaenaea Partocle,
I,
0S -
36 ado 'so iáa 240
—. nt Cta IC MO lit
riiS!Qgram ii MeOan Gaily 50—2 Measurements
'd. 34. oo .4.,4o4o 710 .4, .4. amo
0 tor cta,c — lit
Histogram of gstr, Pui o' Daily S0—2 Measurements
L
u
.12
.6
06
.0l
.02
Sine - - a
I.
Ii —
ii
06 —
.04 -
02 —
a—
It —
U —
0—
s_ .
34. 400100140200 740 ado ida .4, ida 34.360640
930'. pittetia.
Histogram of g5, PctL of Dark Matter
36 100 00 260 00 300
ma,, 0.rc.at00.
Histogram o M000an of 0ar Matter
.4, .4. .4,
—0 pIr Ctoic MOte
Histogram of Mecian of Suspenood Particle,
Figure 2: Fitted Cubic and Unrestricted Dummies
SO—2 vs. GDP Per Capita
0 4000 8000 6000 7000 6000 000 10000 11000 12000 3000 40CC '000 '6000 I 70C0
CDP Per Capita
Dark Matter vs. GDP
1000 2000 300
0'
U-,
130
120 -
110 -
TOOO
90-
80—
70-
60-
0' 50-J 40-
30-
0 0-
-10-x
-20-
—30o
—40—50
-
—50 —70
-
—80 -
80
E
40
20
0
2- -20
—40
-60
—80
—'00
—120
- —140
—160
—150
aci —00
—223
—240
4loG
2000 3000 4000 8000 6000 7000 6000 9000 0000
CDP Per Capito
Suspended Particles vs.
1000 2000
GDP
CDP Per Copto
E
(-3
0'
(N
0U)
Figure3
010002000300040005000600070008000900010000110001200013000l40005000160001/000
CDPPerCapita
SO—2vs.GDPPerCapita30
25
20
15
10
5
0
—5
—10
—15
ID
U
a'
D
a)
a
S
a
a
COPPerCapita
DarkMatter
Figure4
vs.GDPPerCapita400
350
300
250
200
150
100
50
0
010002000300040005000600070008000'30001000011000
Figure5
SuspendedParlielesvs.GDPPerCapita
ID
U
a'
D
U,
4,
0
L
a
0
00)
VC
4,
a0,
U)
0
100
200
—300
—400
—500
—600
01000200030004000500060007000800090001000011000120001300014000150ev160ev11000
GOPPerCapita
Figure 6: Fixed Effects Estimates — 50th Percentile
10 -
0-
—10
—20 -
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 1000011000120001 30001400015000
'5000 7006
GOP Per Capita
30
SO—2 vs.
U
C-I
0•1J)
GDP Per Capita ________
—30 -
—40
—50
45
E 40
35
0'
25
. 20
15
1000 2000 3000 4000 5000 6000
GDP Per Capita
7000 f000 9000 10000
Suspended Particles vs. GDP Per Capita
'000
D
U
0'
0
I)
U
0
a-
0
2)
0C
I)
ai-fl
-d
U)
500
450
400
350
300
250
200
'50
100
50
0
1000 2000 3000 4000 5000 6000 7000 8000 9000 100001100012000' 300014000 1500016000 7000
GDP Per Capita
Table I.
Descriptive Statistics on Air Pollution in Urban Areas
Pollutant Mean Std. Dev. Median WHO Standard
Sulphur Dioxide 33.08 33.11 26.2 40-60
50th Percentile
Sulphur Dioxide 117.17 11271 87.0 100-150
95th Percentile
Dark Matter 42.22 41.92 29.5 50-60
50th Percentile
Dark Matter 127.47 101.45 102.0 100-150
95th Percentile
Suspended Particles 146.62 126.79 91.0 60-90
50th Percentile
Suspended Particles 301.01 268.01 187.0 150-230
95th Percentile
Icotes: Pollutants are measured in pg per cubic meter. World Health
Organization standards listed for the 50th percentile are for the annual
average measure, and those listed for the 95th percentile are for the 98th
percentile of daily measures. Sample size is 1,370 for sulphur dioxide,
1.021 for suspended particles, and 506 for dark matter.
TABLE 2
flw Dcterminants 01Sulphur Dioxide Air Pollution
Random Effects Esümates
(Standard errors in parnitheses)
Variable 50th Percentile 95th Percentile
(t) (2) (3) (4) (5) (6)
Per Capita CD? 4.70 — 3.28 --
S2,- $3,999 (3.94) (13.67)
Per Capita CD? 6.43 — 23.45 — -
$5,999 (4.19) (14.57)
PerCapitaCDP -4.15 — — -15.03 —
S6.- 87.999 (5.31) (18.42)
Per Capita CD? 3.91 20.80 .- —
- $9,999 (4.55) (15.79)
Per Capita CD? -9.33 — — -3.58
$10,000 - $1 1.999 (4.40) (15.26)
Pa Capita CD? -19.07 -3.58
$I2,-$I3,999 (5.74) (19.82)
Per Capita CD? -I I.82 — — -24.62 — --
$I4,-$I5,999 (5.11) (17.61)
PerCapicaGDP -10.28 — -11.35
- $11,999 (6.63) (22.34)
Per Capital CD? — 7. 14 I I.22 — 12.02 22.18
($I,s) (2.50) (2.64) (8.66) (9.22)
Per Capita CDP - -1.l2 -l.44 -1.68 -2.42
squared (0.34) (0.34) (1.53) (1.18)
Per Capita CD?- — 0.041 0.047k — 0.055 0.068
nihed (0.013) (0.013) (0.043) (0.044)
Coasl -8.68 4.68 -5.73 -52.62 -46.79 -45.21
(239) (2.32) (2.32) (8.17) (7.94) (8.03)
Ceclial City 9.45 8.94 8.94 35.4r 34.33 3531
(1.83) (1.82) (1.86) (5.76) (5.73) (5.11)
Industrial 1.58 1.32 1.74 10.95 10.39 9.65
(1.95) (1.94) (2.01) (6.06) (&04) (6.12)
Residential "5.30' -5.7t' 497 .4.08 -5.45 -3.23
(1.96) (1.95) (2.01) (6.10) (6,08) (6.14)
Pop. density 5,54* 4.09
l0.31
4)74* 35.43' 4911'(l0,Ls. ml.) (2.65) (2.57) (4.64) 19.27) (8.97) (16.30)
Year -1.69' '1,19' •i,77* 4.38' -5.32' 512
(0.36) (0.34) (0.35) (1.21) (1)8) (1.22)
Communist i1.59' 11.47' 12.64' 88.05' 88.04' 90,77'
(3.88) (3.83) (3.86) (13.47) (13.32) (13.61)
Trade Intensity —
-15.47' —
.4096'
(4.38) (15.36)
p-value for .11 .000! .000! .000! .06 .07 .007
GD)' v.riables
Pet Capita GDP at which $4107 $5257 —
$4,635 $6182
pollution reaches rcak (1.327) (1.179) (3.309) (2,963)
556 554 579 5.593 5.575 5.555
368 378 323 5.431 5.541 5.232
.158 'ISO .166 .132 .125 .138
Notes: Equations also include an intercept. a dummy to indicate that the type of area is unknown, and a dummy to indicate that
the measurement device is a gas bubbler, o is the estimated variance of the common-to ciiy component of the residuals
and q is the estimated variance of the idiosynctatic component of the residual. Sample size is 1.370 for columns I, 2.
4 and 5; sample size is 1,301 for column, 3 and 6.
Slatistically significant at .05 level for. two-tailed t-test.
TABLE 3
The Daeruünanis of Disk Miner Potlutioc
Raadom Effects Estimates
(Standard cryw, in parentheses)
Variable 50th PaceniiI 95th Percentik
(I) (2) (3) (4) (5) (6)
PcrCapiiaODP 50.50'
99.21* --
$2,000-$3,999 (11.77)
(31.77)
Per Capita GD? 5825 — 118.05' --
$4.000 - $5,999 (12.63) (33.99)
Per Capita GDP 43.49' —
II 1.74* --
$6.- $7,999 (13.10) (35.23)
Per Capita GD? 21.26 —
39.85 —
$8.-$9,999 (13.07) (35.12)
Per Capita GDP 22.27 - 30.27 --
SI0.-$II,999 (13.04) (35.17)
Per Capita CD? 27.29 — 16.0! — —
$12,000- $13,999 (28.64) (74.95)
Per Cap.iai CD? 79.33' 30.52 — 173.84' II). 19'
(S!,s) (13.04) (16.79) (34.32) (47.45)
Pa Capita CD?. -- -12.38' -5.08' -25.62' -16.45'
squared (2.05) (154) (5.38) (7.16)
Per Capita GOP- —
0.56' 0.233 — 1.09' 0.68'
cubed
(0.10) (0.121) (0.26) (0.34)
Desert 40.19' 42.1 I' -10.12 I 13.03 I 18.14' 44.27'
(8.61) (&49) (14.21) (23.24) (22.67) (41.85)
Coast -21.29 .21.75* -17.68' -35.85' -41.44' '334'
(4.94) (4.55) (4.52) (13.00) (11.87) (12,50)
Cential City 9.55' 8.26' 10.95' 32.41' 20.32* 21.44'
(4.05) (3.96) (3.86) (9.38) (9. IS) (9,50)
lndustnal 0.25 -0.55 -0.13 -5.95 -637 -5.92
(4.23) (4.18) (4.06) (9.62) (9.51) (9.74)
Residential .10.60* .11.56* .709' -21.64' 2251' IS.20
(3.98) (3.93) (3.87) (9.12) (9.01) (9-34)
Pop. dcosiuy I.32 I.35 2.92 0.70 0.94 475
(IO/sq. mi.) (0.37) (0.36) (1.17) (0.98) (0.95) (3.17)
Year -0.26 -0.60 -0.59 0.41 -0.1.5 1.27
(0,58) (0.56) (0.58) (1.51) (1.4$) (139)
Comn,isnisi -2O.44 -20.93 -14.44 -0.98 -9.73 756
(7.91) (7.54) (8.51) (20.87) (19.78) (23.39)
Tndt lntcnsuiy —
-6.51 —
-1L05
(6.35) (17.84)
rvilua for alt .OI .OcOt .I .0®t
GDP variabI
Per Capica GDP. which $4,721 $4240 $4,910 $4,971
pollution niches peak (771) (2,180) (973) (2,105)
598 953 864 4,865 4.822 4,772
192 186 166 2.183 2,040 2.108
.345 .352 .210 . .315 .315 .22!
Notes: Equations also thclude an intercep and a dummy to indicaca thai the type ot' ares is unknown. Sasrq,le sirs is 506 for
columns 1, 2. 4 and 5; sawple sirs is 457 for coIu 3 and 6.
Statistically signiticanc ai .05 level for. two-tailed I-test.
TABLE 4
m (or Paiticta PolIt.üoe
Random Effects Esiim.tc.
(SI.adasd enon in pareotheses)
V.nable 'th f'nccttile 95th Peitcocjic
(I) (2) (3) (4) (5) (6)
Per Capita GOP -102.4' —
-19L4'- 53999 (11.9) (25.4)
Pet Capita GOP nc.i• - -247.4'
$5999 (13.5) (29.0)
PerCapitaGDP -101.9' —
-134.2 -- $7999 (31.8) (69.2)
Per Capita GOP 201.2' — — 348.8' —
.59,999 (22.4) (47.1)
PerCapitaGDP .flI.t' -425.0' —
$l0.-S1I.999 (13.0) (27.7)
PerCapita GOP -187.9' — -368.1' - —
512.000 — $13S99 (14.6) (31.6)
PetCapitaGO? -183.1' — -366.2' $14,-
$15,999 (12.5) (27.1)
Per Capita CD? -184.5 — — -388.2' — —
$l6.- $17.999 (15.5) (32.9)
Per Capita GOP :71.97S -72.54' — -144.31' -146.42'
($I.s) (8.02) (8.49) (17.81) (17.14)
Per Capita GOP- — 6.03' 5.88' t2.65' 12.31'
sçnsed (1.01) (1.10) (2.28) (2.28)
Pet Capita GDP- -0.157' 4:143' — 0353' '0.328'
cvbod (0.043) (0040) (0.085) (0.084)
Deane 189.6' 213.3' 289.4' 354.7' 409.8' 489.5'
(19.0) (18.6) (34.3) (403) (39.6) (68.9)
Coast 0,42 4.47 2.90 -9.78 0.22 -4.39
(7.73) (7.58) 0.41) (16.41) (16.0!) (15.43)
Cect..J City it.tt• iz.• 14.76' 32J2' 36.31' 39.IV
(5.24) (5.20) (4.28) 00.47) (10.40) (10.36)
Industrial 13.13 15.2P 11.15 44.2V 47555 40.48
(5.79) (5.77) (5.91) (11.55) (11.49) (I 1.66)
-12.l7 -9.67 -9.45 -1.19 3.83 2.02
(5.83) (4,78) (5,90) (11.62) (11.51) (11.66)
Pop. density 5.20 4.64 -17.55 41.3$ '43.97 '74.43
(t0.X0/sq. ml.) (9.99) (9.64) (12.63) (21.84) (20.95) (26.92)
Year -2.06 -2.27 -2.26 -1.41 •1.65 4.27
(1.20) (1.14) (1.15) (2.58) (2.43) (2.41)
Communist 90.52 I08.37 107.81 22I.51 256.46 251.l2
(13.21) (12.6$) (12.54) (28.68) (V.28) (26.52)
Trade Inteo4ty
— 12.39 6.49
(15.17) (31.23)
p-value tar .11 .0031 .CCOI .0001 .0001 .0001 .0031
GDP v.tiâlea
3.379 3,350 3.482 12.950 12.793 13,085
3,353 3,754 3.0Th 18,043 17.218 14.894
.581 .577 .574 .569 .582 .590
Nolas: Equaziocs also include an intenq,(. • dummy to indicate the type of a, unknown, and two dwmics to indicate the kind
of iniziumeot used to measure suspended particulate matter. Sa.wle sir. is 1.021 for columns I. 2. 4 and 5: sample site
is 971 foe colua3and6.
SiausücalIy si2nificsas at .05 level For. iwo-tailed Hess.
TABLE $
Rsadom Effects Estimates of lb. Determinants of Air
PollutionIncluding Fixed Site Effects
(Standard en on us putnibeses
so2 Dark Miner Susocoded Paslicics
Variable 30 pci! 95 pet! 50 pcti 95 pctl 50 pci! 95 pelt
Per Capita GDP 12.54 -8.28 24.35 77.89 63.44• 142.18
(SI,s) (4.69) (15.11) (17.70) (49.71) (13.94) (30.30
Per Capita CDP - -l.74 0.10 -4.15 -12.10 -2.84 -6.21
squared (0.52) (1.67) (2.73) (7.66) (1.49) (3.23)
Pet Capita GDP - 0.05 -0.03 0.19 0.51 0.04 0.07
cubed (.02) (.06) (0.13) (0.36) (0.05) (0.11)
Yea, -l.05 -3.51 -0.91 -1.12 -3.60
(0.28) (0.90) (0.42) (1.17) (0.82) (1.77)
No. of site
dummies 239 239 87 87 161 16L
p-nine for 0.I 0.118 0.479 0.250 0.I 0.003L
GOP variables
161 1.951 141 1,070 919 3679
97 982 93 755 512 2473
A' 0.76 0.77 0.87 0.82 0.91 0.91
Notes: Sample size is 1.370 for 506 for dark master and 1.021 for suqsoded p.iticles. The SO2 equations also include a
dummy for the measuring device, and the suspasded paiticles equations include two dummies for meansang device.
•SatisticsUy significantae .05 level Fort two-tailed t-te.
TABLE 6
The Detconinant. ol the P.11cm of U.S. Import.
from Mexico and of Value Added by Maqwladors
lndcpcndcnu Total US Imports from Mexico Mexican Value Added in 807 Imports Maquiladora Value Aalded as
Vujahics . Frsct,on of US thinmcnt, ai Fraction of US Value Added Fractioc of US Value Added
(1) (2) (3) (4) (5) (6)
Coostant .028 •.027* .019 .021 .010 .015
(.008) (.009) (.006) (.007) (.007) (.007)
Human Capital Share -.053 ..053a -.016 -.015 -.015 -.016
(.016) (.016) (.010) (.010) (.012) (.011)
Physical Capital Sitar. -.024 -,023 .026 -.027 -.011 -.016
(.010) (.011) (®9) (.010) (.010) (.011)
PollwionAbstemern Coda .014 .012 .I65a ..151. -.0t5 -.077
Fraction of US Industry (.060) (.061) (.073) (.074) (.098) (.090)
Value Added
TantfRiie -.002 00l - — (.028)
(.029)
Injury Rate -. -.011 — -.028
(.020) (.016) (.020)
.127 .127 .095 .095 .236 .392
Sample Size 135 135 136 136 19 19
Mean of Dependent Variable .0069 .0069 .0022 .0022 .0012 .0012
Notcs Standard won ate in parenthesis. Columns (I) and (2) are 01$ estimates. Columns (3). (4). (5) and (6) an maximum
Iikclibood estimates of Tobit models.
flodicstes statislically signiñcait at .05 level.
TABLE 7
Estimated Impacts of NAFTA 00 Toxic Releases
by Manufacturing Enterprises
(Pounds in Thousands)
Industry Trade Liberalization Only Trade Sc Investment Liberalization
Maico ILLS. Canada Mexico Canada
Food&Tobacco Products 5 17 40 129 IC 39
Textile Products 22 499 39 169 487 42
Apparel 6 lB 28 IS 57 28
Lumber & Wood Products -13 86 -88 76 74 -88
Furniture 128 210 176 257 148 178
Paper Products 55 1,198 -953 611 1.198 -971
Printing&Publishing -10 52 -49 56 40 -48
Chemical Products -1,430 12j98 -1,178 4.047 12,408 -1,180
Petroleum & Coal Products 62 93 19 29 70 19
Rubber & Plastic Products -484 693 2,072 289 636 2,(2
Leather Products 37 -26 181 1S8 -38 182
Stone, Clay. Glass & Concrete 0 124 152 107 112 152
Primary Metals -15 -1,59! 5,437 2.166 -1,452 5,494
Fabricated Metals -1 558 254 259 530 254
Nonelectrical Equipment -61 587 -103 62 569 -104
Electrical Equipment 1,445 -1,101 217 1,490 -1,091 217
Transportation Equipment -178 -594 1,435 354 -594 1.424
Misc. Manufacturing 171 32 338 192 97 332
Total -261 13,053 8,017 10,466 13,261 8,032