NAFTA and Convergence in North America: High Expectations, Big Events, Little Time [with
Comments]
Author(s): William Easterly, Norbert Fiess, Daniel Lederman, Norman V. Loayza, Patricio
Meller
Source: Economía, Vol. 4, No. 1 (Fall, 2003), pp. 1-53
Published by: The Brookings Institution
Stable URL: http://www.jstor.org/stable/20065449
Accessed: 18/08/2010 14:24
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=brookings.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The Brookings Institution is collaborating with JSTOR to digitize, preserve and extend access to Economía.
http://www.jstor.org
WILLIAM
EASTERLY
NORBERT
FIESS
DANIEL
LEDERMAN
NAFTAand Convergence inNorth America:
High Expectations, Big Events, Little Time
Free Trade Agreement
(NAFTA) was formally
on 1 January 1994 by the United States, Canada, and
implemented
the
This treaty instantly gained global notoriety following
Mexico.
American
he North
T
in 1991, not only because
the initiative
negotiations
trade agreements
in history,
represented one of the most comprehensive
in establishing
free trade
but also because it seemed to be a breakthrough
initiation
of formal
and developing
countries. The
among developed
catch
high expectations were that trade liberalization would help Mexico
ratio
northern
The
of
Mexican
GDP
with
its
per capita to
up
neighbors.
in goods
and services
that of the United
States
did increase
after unilateral
trade reforms were
in the
in 1986 and also after the implementation
of NAFTA
implemented
other Latin American
aftermath of the so-called
tequila crisis. However,
also grew faster than the U.S. economy
after the mid-1980s,
a
to
it is not obvious
Costa
Rica.
Thus
lesser
extent,
and,
especially
in
Mexico
catch up with
that NAFTA was particularly
helping
important
the United States.1
economies
Chile
are with the World
is with New York University.
Fiess and Lederman
Bank.
Easterly
to Craig Burnside,
Gerardo Esquivel,
Andrew Harvey, Norman
We are grateful
Loayza,
William
Patricio
Ernesto
Messmacher,
Meiler,
Maloney,
Miguel
L?pez-C?rdova,
Guillermo
Maurice
Schiff, and Luis Serven for helpful discussions
Perry, Roberto Rigobon,
on earlier versions.
in this paper belong
to the
The opinions
expressed
the views of the World
Bank.
and do not represent
in that this econ
1. The experience
of Puerto Rico offers an interesting
counterpoint,
similar toMexico's
in the late 1950s and achieved
omy started with a level of development
an unprecedented
level of economic
and institutional
in
States
integration with the United
and
comments
authors
the fastest rates of economic
in the developing
1952. It subsequently
experienced
growth
An analysis of the Puerto Rican experience
is beyond
the scope
Latin American
economies.
of this paper as it would
data for many years prior to 1952, when
require the use of historical
the island became
for suggesting
a commonwealth
territory
of the United
States. We
thank Patricio
Melier
this analysis.
1
2
Fall 2003
ECONOM?A,
This paper assesses the extent to which these high expectations
It examines
trends and determinants
of income
be materializing.
seem to
and pro
across
as
in
observed
North
both
countries
well as
America,
ductivity gaps
were
within Mexico.
The high expectations
for NAFTA
supported by neo
classical growth and trade theories. The seminal work of Solow states that
grow faster than rich countries owing to the law of
returns, as long as production
diminishing
population
technologies,
are the same across countries.2 The neoclassical
growth, and preferences
trade model (the Stolper-Samuelson
theorem) similarly predicts that as the
capital-poor
countries
so will factor prices, including real
prices of goods and services converge,
across
Hence
income
levels
borders will also tend to converge as
wages.
economics
assumption of neoclassical
prices converge. A key simplifying
use the same production
technologies,
exhibiting either
or diminishing
returns to scale.
There is a lively debate about the evidence
the impact of
concerning
is that all countries
constant
on income convergence
across countries, as well as an
literature on economic
within countries.3 At least
convergence
of
the publication
Barro's early work, the economics
has
profession
trade liberalization
extensive
since
in cer
been aware that convergence might be conditioned by convergence
to cause economic
tain fundamentals
While
that are believed
growth.4
there is admittedly much uncertainty
about what these fundamentals
are,
can
as
in
the evidence of conditional
be
evidence
convergence
interpreted
that there are fun
favor of the neoclassical
growth model or as evidence
differences
that prevent income convergence.5
For Easterly and Levine, as well as Pritchett, the "big story" in interna
is that the rich grew richer while the poor got
tional income comparisons
in the levels
differences
poorer.6 Some studies focusing on cross-country
damental
are
(or GDP per worker) argue that these differences
Other
factors.7
factors
besides
different
institutional
largely explained by
fundamentals,
however, might impede economic convergence
among geo
even
areas
in
of
free
trade.
the
presence
graphic
of income per capita
2. Solow(1956).
3. On cross-country
On
within-country
see Slaughter
and Ben-David
(2001)
and
and Sala-i-Martin
(1995)
convergence,
convergence,
see Barro
(1996).
4. Barro
(1991).
5. Doppelhofer,
Mille,
6. Easterly
and Levine
7. Hall
and Jones
and Sala-i-Martin
(1999);
(2001);
Pritchett
Acemoglu,
(2000).
(1997).
Johnson,
and Robinson
(2001).
1996).
(2001,
Sala-i-Martin
3
William Easterly, Norbert Fiess, and Daniel Lederman
recent
More
across
cal differences
of growth with
theories
increasing
returns or technologi
in income levels and
predict divergence
flows might help international
Trade
tech
regions.8
is embodied
in goods and
when
technical knowledge
regions
rates across
growth
nology diffusion
services, and theories of technology diffusion via trade have been the sub
literature.9 A related literature focuses on the barri
ject of a fast-growing
ers that impede technological
in the levels
adoption to explain differences
of trade can thus facilitate con
of per capita income.10 The liberalization
even when production
differ across countries,
vergence
technologies
in convergence
of
tend to be detected
(divergence)
although this would
total factor productivity
(TFP) levels within industries across countries.11
to import production
if trade liberalization
allows poor countries
from advanced countries, productivity
levels might not con
technologies
Even
are different,
to the mismatch
owing
in poor countries
and the sophisticated
gaps within
technologies
imported from the rich countries. Productivity
even
across
countries
therefore
if
industries
trade facilitates
might
persist
verge if the factor endowments
labor skills available
between
technological
convergence.12
literature
The recently resurgent
on economic
transport
geography,
and
is not optimistic
costs,
scale,
knowledge
spillovers
on economic
about the impact of trade liberalization
For
convergence.13
transport costs will remain as barriers to trade and economic
example,
are removed.14 In addition,
if
integration even if all policy distortions
on
an
and
innovation
then
will
also
be
trade,
learning
depend
geography
to convergence
via technological
diffusion.15 These factors
impediment
economies
of
across countries.16 Economies
of scale
might hamper income convergence
some
and knowledge
make
spillovers might
geographic
regions more
of
than
others
because
the
cumulative
effects
of initial
prosperous
simply
such as the density
conditions
8. See
the pioneering
work
of economic
of Romer
Helpman(1991).
9. Eaton and Kortum
10. Parente
11. Bernard
(1999); Keller
and Prescott
(2000).
and Jones (1996).
(1986,
activity.17
1990);
Lucas
(2001).
12. Acemoglu andZilibotti (2001).
13. Krugman
(1991); Fujita, Krugman,
14. Eaton and Kortum
(2002).
and Venables
15. Keller (2002); Eaton andKortum (2002).
16. Redding
17. Ciccone
and Venables
and Hall
(1996).
(2001).
(1999).
(1988);
and Grossman
and
4
ECONOM?A,Fall2003
In the case of Mexico,
the Zapatista rebels took up arms in the southern
state of Chiapas on the day of NAFTA's
Later that year,
implementation.
was
to
in December
was fol
Mexico
forced
the
which
float
1994,
peso,
lowed by a deep banking crisis and severe recession. Domestic
investment
a sharp deterioration
underwent
before the Mexican
economy
began to
recover in late 1995.18 These big events coincided with the implementation
from a long-run perspective,
the post-NAFTA
Moreover,
is still short. This combination
of big events and a short experience
of NAFTA.
period
with NAFTA
increases the difficulty of empirically
identifying the impact
on income and productivity
the agreement
gaps in North America.
we
use
assess
to
various methodologies
NAFTA's
effect on
Nevertheless,
income and productivity
differences.
of
as follows. The next section uses
is organized
on the income
to
the
techniques
identify
impact of NAFTA
and the United States. To deal with the big-events
gap between Mexico
little-time problem, we apply two time series methods.
First, we follow
The
times
rest of
the paper
series
a structural time series exercise
in conducting
that might be able
Harvey
to separate transitory effects (such as the tequila crisis) from the long-term
effects expected from NAFTA.19 Second, following Bernard and Durlauf,
we apply cointegration
to see whether
there is an observable
analysis
between
the United States and Mexico.20
process of income convergence
to test for any structural change in the equilibrium
We do this recursively
GDP using quarterly data from 1960
condition between U.S. and Mexican
to 2001. We find that the debt crisis in the early 1980s and the tequila cri
sis temporarily
which
convergence,
interrupted a process of economic
resumed after 1995. Convergence
in the
trade liberalization
after Mexico's
might have been faster prior to the debt cri
economies
that other Latin American
also grew
late 1980s and after NAFTA
sis. However,
given
quickly during this period, we also provide
of the differences
between Mexico-specific
effects. These results indicate that Mexico's
and
1993 was
not that different
omy, but it was significantly
ous exception of 1995.
18. Lederman
19. Harvey
20. Bernard
and others
(2002).
and Durlauf
(2003).
(1995,
1996).
annual estimates
and Latin American
between
income
1986
performance
econ
the average Latin American
with the obvi
positive after NAFTA,
from
more
econometric
William Easterly, Norbert Fiess, and Daniel Lederman
The
across
section looks at the per capita income
subsequent
the extent to which
in 2000 and estimates
countries
observed
differences
income
differences.
This
5
differentials
institutional
exercise
explain
in using settlers' mortality
Johnson, and Robinson
Acemoglu,
as
instruments for currently observed differences
colonial times
follows
rates from
in institu
and Kraay.21 We find that the
tional quality, based on data from Kaufmann
can largely be
States and Mexico
the United
income gap between
the
We
then
institutional
variables.
gap plus geographic
by
explained
examine
States
the evolution
in Mexico
of the institutional
gap with respect
annual estimates
to the United
of Mexican
by, again, comparing
our
to the average Latin American
results
indicate
thatMex
effect;
more
not
than those of other Latin American
ico's institutions did
improve
effects
countries
in the post-NAFTA
convergence
period. Accelerating
institutions?NAFTA
effort to improve Mexico's
will
require a major
thus
is not
enough.
The following section studies the impact of NAFTA on TFP differentials
within manufacturing
industries across the United States and Mexico. Based
on a panel estimation of the rate of convergence
across twenty-eight manu
we
find
that
the
post-NAFTA period was characterized by
facturing industries,
a substantially faster rate of productivity convergence
than in previous years.
this time, however, we cannot say whether the productivity-convergence
result was due to increased imports of intermediate goods from the United
At
States
(as argued by Schiff and Wang),
competitive pressures
tial access to the U.S. market (as argued by L?pez-C?rdova),
and preferen
or increased
resulting from a variety of factors, including increased
research and development
(R&D) efforts and patenting aided by
the enhanced protection of intellectual property rights contained inNAFTA
(as argued by Lederman and Maloney).22
Mexican
innovation
domestic
The paper then looks at the impact of NAFTA on economic convergence
across Mexican
states. This issue is of particular interest to many Latin
in view of the proposed Free Trade Area of the Amer
American economies
icas (FTAA). This hemispheric
economic
integration would theoretically
lead to the establishment
of free trade and in some cases, such as in Central
America
to
and perhaps in the Southern Common Market
(MERCOSUR),
of economic
which
would
countries,
among
integration
deeper forms
21. Acemoglu,
and Robinson
and Kraay (2002a).
Johnson,
(2001); Kaufmann
Schiff and Wang
and Maloney
(2002); L?pez-C?rdova
(2002); Lederman
22.
(2003a).
6
ECONOM?A,Fall2003
a single economic entity. The unequal economic performance
of
a
states under NAFTA
thus
be
of
differential
might
prelude
under the FTAA or other proposed arrangements,
such as the Cen
resemble
Mexican
effects
tral American
Free Trade Agreement
(CAFTA). We test the conditional
across
on
Mexican
states, but focus exclusively
convergence
hypothesis
states grew faster
that might explain why some Mexican
initial conditions
than others during 1990-2000. We find that the initial skill level of the
density played an important role. We
interpret
that trade liberalization might
indirectly induce
even
across coun
if
induces
it
countries,
convergence
and telephone
population
these results as evidence
within
divergence
tries. The
research
related
final
section
a
and proposes
findings
on the questions
raised by our findings
the main
summarizes
agenda focusing mainly
to TFP convergence
in manufacturing.
Time Series Evidence
process is to separate
simple way to gain insight into the convergence
trends and cycles from the relative output gap between
the United States
a
and Mexico,
trend
in
the
output gap indicates con
whereby
decreasing
A
how
filter can create serious distortions,
vergence. The Hodrick-Prescott
as
can
the Baxter-King
band pass filter.23 We
therefore follow
ever,
Harvey and Trimbur and, in a later work, Harvey, who argue that trends
and cycles are best estimated by structural time series models.24 We esti
a bivariate
mate
between
allows
in which
time series model,
convergence
is captured through a similar-cycle
model
that
coun
across
to
be
correlated
the
driving
cycles
common fac
a direct link between cointegration,
structural
two economies
the disturbances
tries.25 Harvey
provides
tors, and balanced
growth models.26 He also shows that the balanced
growth model results as a special case of the similar-cycle model, when a
common
23. On
trend restriction
the distortions
associated
24.
filter,
band-pass
and Trimbur
Harvey
25.
Harvey
and Koopman
Harvey
27. Harvey
(2002).
and Carvalho
(1999)
King
26.
is imposed.27
with
see references
(2001);
(1997).
(2002).
Harvey
the Hodrick-Prescott
in Harvey
(2002).
(2002).
filter
and
the Baxter
and
7
William Easterly, Norbert Fiess, and Daniel Lederman
The analysis in this section is based on quarterly data on real per capita
over the period 1960:1 to 2002:4.
for the United States and Mexico
are
The per capita GDP figures
adjusted for purchasing power parity (PPP)
GDP
and are taken from the Penn World
Tables 5.6. We applied the following
a
to
create
data series. Quarterly GDP
quarterly PPP-adjusted
procedure
were
for
Economic Cooperation
and
obtained from the Organization
data
were
as
and
the
series
constructed
(OECD),
population
Development
averages of annual figures spread across four quarters.
quarterly moving
GDP
U.S. GDP data were seasonally
adjusted by the provider; Mexican
We
the
converted
data were seasonally
adjusted using X-12-ARIMA.
dollars using quarterly average nominal exchange
then deflated by the U.S. consumer price index
dollars. For the PPP adjustment of the quarterly series,
data into U.S.
Mexican
rates. Both
series were
(CPI) to 1995 U.S.
we estimated
rate bias following Summers and Ahmad, by
the exchange
the
GDP figures on an annual exchange
annual
PPP-adjusted
regressing
rate adjusted GDP series from the World Development
In a
Indicators.n
rate bias to our series of
exchange
GDP
per capita
figures.29
quarterly exchange-rate-adjusted
to the logarithms of quar
We then fit a similar-cycle
bivariate model
terly per capita GDP in the United States and Mexico.30 A model with two
final
step, we
the predicted
applied
appears to describe the data well, and the second cycle appears to
inMexico
around the 1980s.
capture large movements
two trends. This PPP-adjusted
1
shows
ratio
of
the
the
gap
Figure
cycles
until the setback of the 1980s associated
convergence
resumed around 1987, which coincides
debt crisis. Convergence
of the Mexican
unilateral
liberalization
economy
implemented
exhibits
with
the
with
the
in 1986,
the recession
of
also reflect the recovery
after
this might
although
1982-1984. The data also indicate that the tequila crisis represented a tem
porary setback. The downward
slope of the income gap is somewhat
steeper after the 1980s,
suggesting
that convergence
between Mexico
and
28. Summers andAhmad (1974);World Bank (2003).
29.
GDP
To
(yPPP)
estimate
the exchange
rate bias,
rate adjusted GDP
on exchange
Mexico:
United
we
regressed
(ye). Standard
log-transformed
PPP-adjusted
errors are in parentheses:
+ 1.111*^,
R2 =0.987;
yP?P =-0.2944
(0.1608)
(0.0200)
= -0.2944
+ 1.111*^,
R2 =0.992.
States:ym
(0.0121)
(0.1203)
30. Following Harvey (2002).
8
ECONOM?A,Fall2003
FIG U R E 1. TheU.S.-MexicoGDPper CapitaGap:Similar-Cycle
Modelwith Quarterly
PPP-AdjustedData, 1960 to 2002a
Ratio
_i_i_i_i_i_i_i_U
1965:1
1970:1
1975:1
1980:1
1985:1
1990:1
1995:1
2000:1
Source: Authors'calculations.
a. Thedotted line isthe ratioof the U.S./Mexicotrendcomponentsof GDPper capita;the solid line isthe observed ratio.
the United
occurred
States
to have
vergence appears
To investigate
at a faster rate after trade liberalization.
lost momentum
the speed of convergence
Con
however.
during 2000-2002,
the fol
further, we estimated
lowing model:
GAP,
= 0.162 +
0.935GAP,,-0.025NAFTA_GAPM
(0-013)*
(0.092)*
(0.032)**
+ 0.005 LIB _GAPM
(0.016)
R2 = 0.91
+1.083TEQUILA,,
income gap,
is the U.S.-Mexico
and where GAP
and
is a dummy for the 1994 tequila crisis (1994:4-1995:1),
are
for
unilateral
trade
and
dummies
Mexico's
NAFTA_GAP
LIB_GAP
both of
liberalization
and NAFTA
(1994:1-2002:4),
(1986:1-1993:4)
errors
are in
which are interacted with the lagged income gap. Standard
where
TEQUILA
William Easterly, Norbert Fiess, and Daniel Lederman
9
one asterisk means
at 5 percent, and
statistical significance
parentheses,
at 1 percent. We find that NAFTA,
two mean significance
but not unilat
a
on
had
the speed of
eral trade liberalization,
significant positive
impact
a
the half-life of one unit shock to the income
convergence. With NAFTA,
fallen from 2.6 to 1.8 years. The fact that unilateral
is
does not appear to be significant for income convergence
a
in
similar result later
the paper, when analyzing the
interesting. We find
on productivity
NAFTA
of
unilateral
liberalization
and
impact
growth.
gap appears
liberalization
to have
(ointegrationAnalysis
to Bernard and Durlauf,
between two or
long-run convergence
According
more countries
if the long-run forecasts
exists
of output differences
are
two
to have converged
In
zero.31
other
economies
said
words,
approach
if the difference
between
ditions, stability implies
convergence
ary. Absolute
relative
or conditional
them, ytt, is stable. If we abstract from initial con
two series is station
that the difference between
requires
that the mean
convergence
the two series has a constant mean.
of yt is zero, while
that the difference
between
requires
If two series are cointegrated, but with
are comoving
the economies
(that is, they
a vector different from (1,-1),
are driven by a common trend) but not necessarily
to identical
converging
levels of output. Cointegration
between economies
alone is therefore a
If a con
necessary, but not sufficient condition for absolute convergence.
stant is introduced into the cointegration
to test for
space, it is possible
zero. A
constant
to
absolute and relative convergence
the
by restricting
zero constant
supports absolute convergence.32 Following Fuss, we intend
vector of the form ( 1, -1 ) at the end
to interpret evidence of a cointegration
of the sample, together with a rejection of this vector parameterization
in
as
an
of
of
evidence
subsamples,
ongoing process
convergence.33
31.
32.
between
on
Bernard
and Durlauf
Introducing
stochastic
1996).
(1995,
into the cointegration
and deterministic
convergence,
a trend
it possible
to distinguish
space makes
a homogeneity
where
restriction
(1,-1)
to stochastic
and homo
convergence
with a trend corresponds
coefficients
a trend to deterministic
con
As we reject stochastic
(1,-1) without
convergence.
in favor of deterministic
in our data, we only report the findings
vergence
convergence
in the cointegration
con
based on a constant
space, which we view as a test of deterministic
the GDP
geneity
ditional
convergence.
Fuss (1999) postulates
= a + bx + u, then the results
y
33.
at the end of the period,
that if y and x are cointegrated
evidence
of the following:
provide
with
10
Fall 2003
ECONOM?A,
TABLE 1.
Cointegration Analysis for the United States and Mexico, 1960:4 to 2002:4
L-max
Eigenvalue
0.164429.64***
0.0171
2.85
Trace
H0:r
32.49***
0
1
2.85
p-rL-max90
10.29
2
Trace90
17.79
1
7.50
7.50
Source: Authors'calculations.
***Statisticallysignificantat the 1percent level.
A cointegration
GDP, with a con
analysis between U.S. and Mexican
stant and four lags in the cointegration
space over the full sample from
vector (see table 1). A
1960 to 2002, reveals one significant cointegration
to (1, -1) cannot be
the cointegration
space according
=
=
over
the full sample; this provides evidence
0.23)
1.45,p
rejected (%2[1]
=
in favor of convergence
1960-2002:
0.720, with
GDPUS GDPMX
during
a standard error of 0.082.34
restriction
of
The estimate
of the constant
in the cointegration
vector is greater than
zero, and the standard error for the constant is relatively small. We interpret
this as evidence of incomplete convergence,
in the sense that Mexico
is
toward the U.S.
level of income up to a point. That is, the
converging
observed process of convergence
is likely to lead not to absolute conver
gence, but rather to a constant income differential. The estimated constant
will reach a maximum
of about 40 to 50 percent of
suggests that Mexico
the U.S. per capita GDP. Whereas
the evidence applies to the whole period,
this process of conditional convergence may hold only for certain years.
Recursive
cointegration
analysis reveals that the (1, -1) restriction does
not hold in all subsamples
(see figure 2). The graph in figure 2 is scaled in
such a way that unity represents the 5 percent level of significance. A test
can
of convergence
statistic below one thus indicates that the hypothesis
not be rejected. We find strong evidence
for divergence
the
1980s
during
(debt crisis),
in spite of the fact that we estimated
the cointegration
vector
a = 0 and b =
a <>
1 indicates
that the series are converging;
0 and b = 1 indicates
that the two series are converging
a > 0 and b <
a<0
1 implies
and b > 1 implies
and b > 1 implies
a < 0 and ??< 1 implies
a>0
34. A
similar
dard error of 0.044.
result
that x converges
that y converges
divergence
toward
toward x;
(x falls behind
(y falls behind
divergence
is obtained
for annual
up to a constant;
y\
data: GDP?/S
y); and
z).
-
GDPMX
=
0.881,
with
a stan
11
William Easterly, Norbert Fiess, and Daniel Lederman
FIGURE 2. TraceTests for Cointegrationbetween U.S.andMexico (Log)QuarterlyGDP,
1960:4 to 2002:4 (RecursiveEstimates)
Multiples of the Critical-TraceTest Statistic
5 i?''V-"
Without tequiladummy
1974:1 1976:3 1979:1 1981:3 1984:1 1986:3 1989:1 1991:3 1994:1 1996:3 1999:1 2001:3
Source: Authors'calculations.
with
that properly
dummies
the key first and fourth quarters
identify
of
1982.35
the impact of the 1994 tequila crisis on the convergence
we
process,
analysis with and without
perform a recursive cointegration
a dummy for the tequila crisis. As shown in figure 2, which plots the coin
tegration trace test over time, the tequila crisis had an impact on the con
To
assess
vergence
The
process.
convergence
process
convergence
hypothesis
35.
The
debt crisis
relevant
tended
tests showed
that other dummy
variables
for the
specification
the estimates
of the cointegration
rank and coefficient
restrictions.
of three subsamples
finds a result similar to that reported above. A test
model
to bias
A
separate analysis
of the (1,-1) restriction
1961:01
can be rejected
to 1975:04
1976:01 to 1988:04
1989:01 to 2002:04
This
whole
supports
sample
of a crisis dummy reveals a resumed
the tequila dummy, the
1987 onward. Without
is rejected around the time of the crisis. This sug
inclusion
from
a similar
(figure
(%2(l)
=
in the following
1.12, p
subsamples:
0.29),
=
=
(x2(l) 8.86, p 0.00), and
= 0.61, p = 0.43).
(%2(\)
convergence/divergence
2).
=
pattern
as a recursive
analysis
over
the
12
Fall 2003
ECONOM?A,
conver
that the tequila crisis temporarily
interrupted an ongoing
in
the
late
1980s.
gence process
The evidence from time series analyses can be summarized as follows.
gests
Structural
time series modeling
and recursive cointegration
analysis both
of
and
between
Mexico
and the
convergence
identify periods
divergence
United States during 1960-2002.
Both econometric
evi
find
techniques
that the tequila crisis only temporarily
interrupted a convergence
the estimates of structural
process that started in the late 1980s. However,
in
the
of
the
income gap
coefficient
U.S.-Mexico
changes
autoregressive
seems to be faster than in the rest
indicate that the speed of convergence
dence
of NAFTA.
of the sample only after the implementation
seems to have a limit.
of convergence
In any case,
this
process
Mexico'sPerformance
RelativetoOtherLatinAmericanCountries
in the region may
in figure 3, other economies
As highlighted
relative to the United
just as fast as or even faster than Mexico
have grown
States after
process of conver
identify the Mexico-specific
we compared Mexico's
level of development,
per
the per capita income gap relative to the United
the late 1980s. To better
toward
gence
formance
the U.S.
States with
in closing
the equivalent
reformed
their economic
NAFTA.
This
of Latin American
countries
that
performance
not
of
but
did
the
benefits
policies
enjoy
statistical
difference
testing for a significant
a
countries and the
for group of Latin American
involved
the year effects
to Mexico.
The dependent
variable was the (log)
year effects
specific
ratio of per capita GDP of the countries relative to the United States. The
that
test was conducted with two samples of Latin American
countries
between
include Mexico: Group
with nine countries.36
36.
Costa
of twenty-two
countries,
and Group
2,
in figure 4.37 Mexico's
year effects are statisti
different from the rest of Group 1 at a 10 percent confi
results are shown
The
cally
1, consisting
significantly
The
Rica,
twenty-two
Dominican
Group
Republic,
1 countries
Ecuador,
are Argentina,
El Salvador,
Bolivia,
Guatemala,
Brazil,
Chile,
Guyana,
Trinidad
Colombia,
Haiti, Hon
and Tobago,
Peru,
Panama,
Nicaragua,
Paraguay,
are Argentina,
The nine Group 2 countries
Brazil, Chile, Colom
and Venezuela.
Rica, Mexico,
Peru, Uruguay,
37. The estimated model was yc,t - c + ?, Dt + ?,MEXA
DMEX, where y is the log of
and DMFX is
the per capita GDP ratio with respect to the United
States, Dt is a year dummy,
aMexico
4
dummy.
Figure
plots ?,,MEX ?,.
duras,
Jamaica,
Mexico,
and Venezuela.
Uruguay,
bia, Costa
William Easterly, Norbert Fiess, and Daniel Lederman
FIG U R E 3.
1960 to 2001
13
Per CapitaGDPRelative to the United States, Selected Economies,
GDPper capita (PPP)/U.S.GDPper capita
1970
1965
1975
1980
1985
1990
2000
1995
Source: Loayza,Fajnzylber,
and Calder?n(2002);PennWorldTables5.0;World Bank (2003).
1982 onward. In other words, the annual observations
in figure 4 are significantly
different from zero only after 1982.
With respect to the smaller comparator group, Mexico's
annual effects are
and 1999-2001.38 However,
these differ
also different during 1982-1994
dence
level from
shown
ences
tended to be significantly
richer
simply reflect the fact that Mexico
than other regional economies
these
The
real
is
years.
during
question
whether
richer than other Latin American
grew significantly
these
which
should
be reflected in upward move
years,
during
shown in figure 4. This only
of the country-effects
differentials
Mexico
economies
ments
occurs
after
1995 with
group of Latin American
occurred in 1986-1993.
38. Wald
American
tests
for
and Caribbean
respect to both comparator groups. For the larger
and Caribbean economies,
this might also have
of the difference
significance
are not reported.
effects
between
Mexico
and
average
Latin
14 ECONOM?A,Fall2003
FIGURE 4.
Mexico Year EffectMinus RegionalYear Effect9
u_i_i_i_i_i_i_i_i_i_u
1961
1965
1969
1973
1977
1981
1985
1989
1993
1997
2001
Source: Authors'calculations.
a. Log (GDP
per capita/U.S.GDPper capita) (PPP).Theexcludedyear is 1960.See tableA1 inthe appendixfor summarystatisticsfor
data used and definitionof the groups.
did not catch up to the United States significantly
countries (the eight included in Group 2)
about the possibility
thatMexico's
unilateral reforms spurred
The fact thatMexico
faster than other middle-income
raises doubts
to a greater extent than reforms in
In contrast, the post-NAFTA
period
Mexico-U.S.
income gap, which declined
with the United States
convergence
countries such as Chile or Costa Rica.
is characterized
by a declining
included in both samples. This
faster than for the average Latin economies
result is consistent with previously discussed estimates of the acceleration
of convergence
sections identify
only after 1994. The following
of
the
U.S.-Mexico
constraints
convergence
process.
lying
the under
IncomeGaps and Institutional Gaps
the role
in the introduction, a substantial literature highlights
in per
in producing
differences
variation
institutional
cross-country
As discussed
of
William Easterly, Norbert Fiess, and Daniel Lederman
FIGURE 5.
15
InstitutionalGaps inNorthAmerica,2000 to 2001
Variable ranges from -2 to+2 for all countries
ICanada
Voiceand
accountability
Political
Government Regulatory
stability
effectiveness
Ruleof law
Control
of
quality
corruption
Source: Kaufmannand Kraay(2002a).
trade liberalization
and the institutional harmo
capita income.39 Despite
nization
intellectual
(for example,
requirements
imposed by NAFTA
and
environmental
investor
standards), obvi
property rights,
protection,
ous institutional gaps remain between the United States and Mexico.
Fig
ure 5 draws on data from Kaufmann
and Kraay to show the gaps along six
In 2000-2001,
Mexico
its North
clearly
lagged behind
all
institutional
dimensions,
partners along
corrup
especially
tion and rule of law. If these institutional differences
persist, absolute
dimensions.40
American
as predicted by neoclassical
will proba
economics,
even if trade is completely
liberalized. These types
are difficult to identify with time series
to convergence
of impediments
as
in the previous section, mainly because
such
those
analyses,
presented
can
be rooted in history and tend to vary little over time.
institutional gaps
income convergence,
bly never materialize
The experience
of Puerto Rico
on how
medium-term
perspective
39.
Hall
40.
Kaufmann
and Jones
(1999); Acemoglu,
and Kraay (2002a).
(recall figure 3) can provide a useful
institutional convergence
might affect
Johnson,
and Robinson
(2001).
16 ECONOM?A,Fall2003
ter
convergence. When Puerto Rico became a commonwealth
of
the
United
States
in
not
free
in
it
trade
1952,
ritory
goods
gained
only
and factors of production,
but also some of the political and regulatory
economic
institutions
available
incentives
for
in the United
In addition, firms received tax
in the island. Consequently,
the
States and Puerto Rico narrowed
States.
setting up operations
gap between mainland United
over the next 50 years, especially
significantly
income
compared with the income
and other Latin American
countries. The remainder of this
gaps of Mexico
the role of institutional
section estimates
gaps in maintaining
long-run
income gaps.
DataandMethodology
To investigate
of Acemoglu,
of exogenous
the impact of institutional gaps, we follow the methodology
involves using a set
Johnson, and Robinson.41 This basically
to
related
variables
characteristics
geographic
(namely,
variables,
latitude, and
dummy
landlocked-country
dummy,
a
for oil and commodity
constructed
trade
share indi
exporters),
cator that takes into consideration
countries'
size and geographic
factors,
regional
dummies
an indicator of ethnolinguistic
fractionalization,
the Kaufmann-Kraay
indicators of institutional
and a composite
index of
as
2000-2001
from
quality
variables of per capita income (in PPP-adjusted U.S. dollars)
explanatory
as of 2000.42 Table A2 in the appendix contains the summary statistics for
our data set. Our methodology
is two-stage
least squares (2SLS).
Since the indicators of institutions
and the corresponding
composite
we need to find
to the level of development,
index can be endogenous
instruments for this variable. Also, the institutional variables are measured
with error, as explained by Kaufmann
and Kraay and Acemoglu,
Johnson,
and Robinson. A priori, it is difficult to say which effect will predominate,
since
the endogeneity
problem
whereas
improves institutions,
duce an attenuation bias.
could bias the estimates
the measurement
upward
error problem
if income
could pro
rates
show that the (log) mortality
Johnson, and Robinson
Acemoglu,
can
a
of settlers
be good instrument for current institutions. These authors
on
a
literature linking the importation of political and
rely
long historical
41.
42.
and Robinson
Johnson,
(2001).
Acemoglu,
The trade share indicator
is from Frankel
and Romer
the average
of the six individual
components.
(1999);
the composite
index
is
William Easterly, Norbert Fiess, and Daniel Lederman
17
institutions to the extent to which colonies were settled by their
sources for the extraction
as opposed
to becoming
colonizers,
European
Where
commodities.
of high-priced
settled, they imported
Europeans
had incentives not to
"good" institutions. At the same time, Europeans
settle in places where the climate and other historical factors reduced life
economic
It thus seems logical
expectancy.
centuries
teenth and nineteenth
to use settler mortality
rates in the eigh
as instruments
for institutions
in the
present.
Results
2, 3, and 4 present our results. Table 2 presents the 2SLS
effects of the key variables on the (log) PPP-adjusted
per capita
of 2000. Table 3 shows the first-stage regressions,
in which the
index of institutional quality is the dependent variable. Table 4
Tables
ordinary least squares (OLS) regressions,
corresponding
the assumption
that institutions are exogenous.
which
estimated
income as
composite
shows the
depend on
In the five specifications
shown in table 2, the instrumented composite
index of institutions is positively
and significantly correlated with income.
In fact, across the four models
the relevant coefficient
is quite stable, rang
1.35 to 1.94. The only other robust explanatory
variable is the
oil
which
with
for
appears consistently
exporters,
dummy
positive and sig
nificant coefficients.
The Frankel-Romer
trade openness
indicator is not a
of income per capita: virtually
identical results
significant determinant
ing from
were
obtained
for
average
share.43 These
we
when
1965-1990
used the Sachs-Warner
policy openness
instead of the Frankel-Romer
constructed
index
trade
can be interpreted as an indication either that the
of countries
is mainly determined by the
long-run level of development
or
of
domestic
institutions
that
the
correlation
between
the instru
quality
results
ments
used by Frankel and Romer to estimate the exogenous portion of the
trade-to-GDP
ratios (the so-called geographic
gravity variables) and the
rates is so high that it is quite difficult to really identify
settlers' mortality
the marginal
effects of institutions and trade separately.44
results for the first-stage OLS regressions
in table 3 show that the
are
rates
(log) settlers' mortality
good predictors of institutional quality in
2000. The mortality variable is always statistically
significant and has the
The
43.
Sachs
and Warner
44.
Dollar
and Kraay
(1995).
(2003).
18
ECONOM?A,
TABLE 2.
Fall 2003
Two-Stage LeastSquares Regressions of LogGDPper Capita2000a
(1)
Explanatory variable
Institutional index
Net oil exporters
Net commodity exporters
1.35**
1.39**
1.40**
(0.19)
(0.20)
(0.20)
Americas
Log constructed trade share
(Frankel-Romer)
1.37***
(0.25)
0.87***
0.69**
0.72**
0.73**
0.71***
(0.18)
(0.21)
(0.20)
(0.21)
-0.16
(0.13)
-0.21
-0.16
(0.16)
-0.12
-0.16
(0.16)
-0.10
-0.16
(0.16)
-0.14
(0.42)
(0.59)
(0.35)
(0.38)
(0.38)
0.98
0.45
0.59
0.60
0.55
(0.73)
(0.38)
(0.43)
(0.43)
(0.48)
0.70
0.53*
0.61*
0.62*
0.59
(0.53)
(0.30)
(0.33)
(0.33)
(0.38)
0.43
0.26
0.27
0.28
0.26
(0.43)
(0.24)
(0.27)
(0.27)
(0.30)
-0.04
(0.12)
0.02
0.00
(0.09)
(0.10)
Ethnolinguisticfractionalization
Landlocked
(5)
(0.30)
0.22
East Asia and the Pacific
(4)
1 94***
(0.18)
South Asia
0)
(0?53)
-0.22
Africa
(2)
0.00
0.00
0.00
(0.00)
(0.00)
(0.00)
-0.05
0.26
(0.28)
(0.39)
-0.02
Latitude
(0.01)
0.72
R2
0.84
0.84
0.83
0.84
Source: Authors'calculations.
* *
**
*
statisticallysignificantat the
Statisticallysignificantat the 10 percent level; statisticallysignificantat the 5 percent level;
1percent level,
a. Robuststandarderrorsare inparentheses.
of the OLS and 2SLS estimates of
expected negative sign. A comparison
the institutional coefficient
shows that the OLS estimates are significantly
suffer from attenuation
lower. These results suggest that OLS estimates
bias owing
errors afflicting the institutional variable.
results shed light on the
these econometric
The last bar
between the United States and Mexico.
to measurement
6 illustrates
Figure
income gap observed
how
in the log of PPP-adjusted
on the right is the income gap (the difference
as
1.2. The penultimate
of
is
which
2000,
per capita GDP)
approximately
income gap (from column one of table 2).
estimated
bar shows the model's
show the marginal
effects of the statistically
significant
status of
on the (log of) the U.S.-Mexico
income gap. Mexico's
a net exporter of oil tends to reduce the income gap by about 0.88. In con
The
other bars
variables
trast, the first six bars on the left side of the graph show the contribution of
each institutional dimension. The sum of the individual institutional con
William Easterly, Norbert Fiess, and Daniel Lederman
19
First-StageRegression for Institutional Index9
TABLE 3.
variable
Explanatory
(1)
-0.17**
Logmortality
(0.07)
Oilproduction
dummy
-0.37**
dummy
Commodity
-0.37*
(0.08)
-0.45**
-0.18**
(0.08)
-0.45**
(0.20)
(0.18)
0.03
0.00
0.00
(0.16)
(0.16)
(0.20)
(0.18)
(0.18)
EastAsia and the Pacific
-0.52
Americas
-0.35
(0.34)
(0.33)
-0.65*
(0.30)
-1.00*
(0.34)
-0.52
(0.33)
-0.35
-0.69**
(0.34)
-1.07**
(0.41)
-0.45
(0.45)
-0.35
(0.24)
(0.24)
(0.26)
0.04
0.04
0.05
(0.11)
(0.11)
(0.12)
Ethnolinguistic fractionalization
-0.43*
R2
-0.42**
-0.18**
0.04
-1.00***
Latitude
(0.08)
(0.18)
Asia
South
Landlocked
-0.18**
0.04
(0.30)
Log constructed trade share
(Frankel-Romer)
(0.07)
(0.18)
-0.65**
Africa
(2)
(3)(5)
(4)
-0.17*
-0.43*
-0.69**
(0.34)
-1.12**
(0.39)
-0.48
(0.44)
-0.36
(0.26)
(0.18)
-0.69**
(0.34)
-1.12***
(0.39)
-0.48
(0.44)
-0.36
(0.26)
0.00
0.00
(0.00)
(0.00)
-0.43*
-0.45*
-0.45**
(0.20)
(0.20)
(0.22)
(0.22)
0.02*
0.02*
0.02**
0.02*
(0.22)
0.02**
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
0.62
0.62
0.63
0.63
0.63
Source: Authors'calculations.
**
***
*
Statisticallysignificantat the 10 percent level; statisticallysignificantat the 5 percent level;
statisticallysignificantat the
1percent level,
a. Robuststandarderrorsare inparentheses.
tributions
bit more
is about 2.5, but gaps in rule of law and corruption seem to be a
errors in
important than the other institutions. The measurement
each category probably make this last observation
ever, since we cannot be sure that these institutional
In any case,
States and Mexico
the others.
different
from
between
the United
less meaningful,
how
are
gaps
significantly
the large income gap observed
is readily explained
by institu
were not an oil exporter,
it would
tional features. Moreover,
if Mexico
be
it
than
the full model predicts a
poorer
probably
actually is. Finally,
log ratio of U.S. over Mexican GDP per capita of about 0.62, which trans
lates into a 0.54 ratio of Mexican
GDP per capita over the U.S. GDP per
capita.
It is perhaps
a coincidence
that this is more
or less the limit to the
convergence
process estimated with the cointegration
analysis above.
thus hamper convergence
in North America.
Institutional
gaps might
This does not mean that NAFTA,
in particular, did not have an effect on
20 ECONOM?A,Fall2003
TABLE 4.
OLSEstimates of LogGDPper Capita2000a
variable
Explanatory
(1) (2)
Institutional index
1.10***
Oil production dummy
(0.11)
(0.11)
(0.11)
0.58***
0.59***
0.60***
0.57***
(0.16)
(0.16)
(0.20)
(0.17)
(0.17)
-0.15
-0.57**
(0.29)
(0.33)
EastAsiaandthePacific
(0.24)
-0.02
(0.20)
-0.03
-0.56*
(0.29)
0.18
0.19
(0.38)
0.29
0.29
(0.24)
0.05
0.030.02
(0.21)
(0.22)
(0.11)
-0.12
-0.14
(0.16)
(0.15)
-0.57*
-0.56*
(0.30)
(0.24)
(0.30)
0.12
(0.36)
0.24
(0.24)
(0.22)
(0.22)
(0.00)
(0.00)
(0.36)
0.01
0.01
-0.01
(0.09)
(Frankel-Romer)
(0.16)
(0.28)
0.00
0.12
(0.32)
0.160.25
(0.22)
SouthAsia
Log constructed trade share
-0.14
(0.13)
-0.65**
Americas
1.08***
0.51***
(0.13)
Africa
(5)
1.11***
1.11***
(0.11)
-0.17
Commodity dummy
(4)(3)
1.11***
(0.09)
(0.10)
0.00 0.00 0.00
Ethnolinguisticfractionalization
(0.00)
Landlocked
-0.18
(0.17)
Latitude
-0.01
-0.20
(0.19)
(0.00)
No.observations
6161
68
68
61
Source: Authors'calculations.
*
**
***
Statisticallysignificantat the 10 percent level; statisticallysignificantat the 5 percent level;
statisticallysignificantat the
1percent level,
a. Robuststandarderrorsare inparentheses.
institutional
gence
was
Our time series analyses
convergence.
suggest that conver
in fact present after NAFTA. Was
this due to institutional
convergence?
PerformanceinMexico versustheRestof theRegion
Institutional
of the impact of institutions on the level of devel
that
institutions
tend to change little over time, and thus
opment presumed
the
instrument
that
Johnson, and Robinson?
by Acemoglu,
proposed
The previous
namely,
the
estimates
settlers'
lysts expected
45.
Acemoglu,
mortality
that NAFTA
Johnson,
rate?is
would
and Robinson
appropriate.45
exert direct
(2001).
However,
some
and indirect pressures
ana
on
William Easterly, Norbert Fiess, and Daniel Lederman
FIGURE 6.
21
TheContributionof InstitutionalGaps to the U.S.-MexicoIncomeGap
GDP
andObserved
percapita/Mexico
percapita
Explained
(log)U.S.GDP
of
Total
AccountabilityPolitical GovernmentRegulatoryRuleof law Control
corruption institutional
stability effectiveness quality
gap
Explained Observed
gap (2000)
gap
Source: Authors'calculations.
to improve its institutions.46 The direct pressures came from spe
cific elements of the trade agreements,
including those related to investor
intellectual property rights, labor, and the environment, which
protection,
Mexico
explicitly
focus
could
on Mexico's
have
enforcement
emanated
pressure
States regarding Mexico's
from
of its own
the political
laws. The
debate
indirect
in the United
to implement
its commitments.
Our
over
rare
but
view is that institutions probably change little
time, although
uncommon
or
events
other
in
institutions
might
profound changes
political
46.
NAFTA.
An
anonymous
reviewer
ability
suggested
that this might
be
the strongest
impact
from
22
ECONOM?A,
Fall 2003
FIGURE 7. Mexico Year EffectsRelative to Regional Year Effects,
Institutional Index(ICRG)
Source: Authors'calculations(see text).
a. Theexcludedyear is 1984.See tableA1 inthe appendixfor summarystatisticsfordata used and definitionof the groups.
change
the quality of public institutions.47 We therefore
toMexican
institutions before and after 1994.
pened
To test whether Mexican
institutions
changed more
analyze what hap
than those of other
similar to those con
countries, we estimated regressions
in
4.
The
the
income
gaps presented
figure
dependent variable was
cerning
institutional
the country's
the difference
between
indicator,
composite
Latin American
of three indexes of institutional quality provided by the Inter
composed
national Country Risk Guide (ICRG) and the U.S. value of this index. The
index was constructed using factor analysis of ICRG's bureaucratic qual
ity, law and order, and absence of corruption variables. These data cover
1984-2001.
Again, for the comparisons we used the Group 1 and Group 2
samples (Group 1 includes Cuba in this analysis). Figure 7 shows the
to use the Acemoglu
instrument while
also
sound contradictory
follow
that the
might
change over time, it does not necessarily
on the level of per capita income are uninfor
is useless
and that the regressions
contains
various com
mative.
The exogenous
quality most probably
portion of institutional
to more
recent
some related
to long-term
historical
and some related
heritage
ponents,
over time might
not be
in institutions
innovations.
This implies that the variation
exogenous
47.
itmight
Although
that institutions
believing
instrument
William Easterly, Norbert Fiess, and Daniel Lederman
InstitutionalChanges in LatinAmerica
TABLE 5.
or
Country
NAFTA
Before
(1984-93)
group
NAFTA
After
(1994-2001)
Change
-1
Mexico
Argentina
Brazil -1
23
-1
80 -1.46
0.34
49 -1.05
0.43
-1.57
.00
Chile -1
55 -0.73
Colombia -1
-0.57
0.82
80
-1.91
South America -1
,68 -1.59
CentralAmerica -2,
Andean countries
-1
-0.11
0.09
51 -1.61
0.90
,98 -1.60
0.39
,83 -1.53
0.30
LatinAmerican countries -1
Source: Authors'calculations,based on data from International
CountryRiskGuide.
year effects for the whole period were not statistically
from the first group of Latin American
countries, but they were
seems to
from
different
the
1994. Mexico
after
group average
statistically
results. Mexico's
different
relative to the regional average during this period,
have underperformed
or stable negative difference
is reflected in a declining
between
which
Mexico
and the average regional effects.
improved its institutions relative to the United
though Mexico
in the post-NAFTA
period, the results in figure 7 are due to the fact
in the region also improved their institutions without
that other countries
Even
States
benefiting from NAFTA. Table 5 shows the changes in the gap relative to
index before and after
institutional
the United States of the composite
that improved their institutional gap the most after
1994. The countries
1994 were Chile and the Central American
group, whereas Mexico's
norm
was
rather
the
for
the
whole
improvement
region. Moreover, Mex
took place after 1999 and thus was probably related
ico's big improvement
to the political
transition, as was the case in Chile and Central America.
data are consistent with the findings of Lederman,
and
Loayza,
has a positive
effect in
Soares, who find that political democratization
terms of reducing corruption
in a large sample of countries.48 NAFTA
alone is unlikely to contribute to the institutional development
of Mexico
areas covered
outside
the specific
the
agreement.
by
Consequently,
These
fully
stationary.
exogenous
The
component
historical
of
instrument
the level
of
this issue.
Rigobon for highlighting
48. Lederman,
and Soares
Loayza,
the
can
still be valid,
institutional
(2002).
an
since it captures
however,
are grateful
to Roberto
index. We
24 ECONOM?A,Fall2003
Mexico's
tions need
policy efforts to combat
to be pursued further.
and improve general
corruption
institu
Productivity Gapswithin Industries,
across the United States and Mexico
IfNAFTA
trade liberalization helped technological
adoption and modern
we should observe an acceleration
inMexico,
in the rate of TFP
within industries. To
between
the United States and Mexico
convergence
ization
examine
between
this channel
the United
of convergence,
and Mexico
States
we
calculated
in manufacturing
TFP
differentials
sectors. The fol
the data, methodologies,
and econometric
discuss
lowing paragraphs
on TFP convergence.
the impact of NAFTA
results concerning
Dataand TFPEstimates
measure
the
in total factor productivity
(TFP) following
and Diewert, which is used in
approach suggested by Caves, Christensen,
context by Keller.49 They calculate a multilateral
the cross-country
(bilat
We
differences
eral in our present
case) and flexible
InTFPd,
(1)__
TFP
index of the following
form:
=
(inYcit -ta?;)-a~t(lnLat
-ETZ^)
-(\-Gcit)(\nKclt-\nKu),
c is the country index (Mexico and the United States), i represents
industries, and t is time, y is total output, L is labor, and K is capital stock,
labor share of output. The Caves, Christensen,
while a is the cost-based
of the log output, labor, and cap
and Diewert approach entails de-meaning
where
ital series, using the geometric
averages of both countries. The resulting
index in each country and industry is based on a vector of outputs and
inputs that are common to both countries. An intuitive reading is that this
index tells us what the productivity
level in each country and industry
TFP
be if they had the same labor cost shares.
and factor shares come from
Data on production
would
United Nations
49.
Caves,
Industrial Development
Christensen,
and Diewert
(1982);
Organization
Keller
(2002).
the OECD
(UNIDO)
and the
and cover
William Easterly, Norbert Fiess, and Daniel Lederman
25
at the three-digit
industries
International
manufacturing
twenty-eight
code.50
The
Classification
Standard Industrial
(ISIC)
output data were
deflated using the U.S. industry deflators from Bartelsman,
Becker, and
(2000). The capital stock data were constructed using the permanent
rate per year, based
assuming a 5 percent depreciation
inventory method,
on fixed investment data from UNIDO,
and were deflated using the PPP
investment price levels from the Penn World Tables 6.O.51 Tables A3 and
Gray
in the appendix contain summary statistics
for Mexico
and the United States, respectively.
A4
for the industry-level
data
EstimationStrategy
To assess
how the rate of (log) TFP convergence
changed after the imple
we estimated an autoregressive
model with struc
of NAFTA,
with
coefficient
and
in
the autoregressive
tural change
industry fixed
mentation
effects
and year effects:
=
(2) ylJt a, + y, + ? yM + ?XW?,-i
where
i=
Wand
1, 2,...,
As mentioned,
maximum
number
t=
1,2,...,
+ ?Ata + elV,
T.
number of industries is N = 28, and the
=
is T
25. In the context of the fixed-effects
our maximum
of years
to control for industry-specific
is designed
effects, a,, by
could
leftand
both
the
variables
produce a
right-hand-side
de-meaning
to
the
correlation between
the
coefficients
bias in the estimated
owing
estimator,
which
lagged mean
of y and the contemporaneous
error, e/r The bias is inversely
to T. Also, as mentioned,
there are no good data on Mexico's
proportional
unit price for industry-level
output, such that the use of the U.S. deflator
error that is endogenous
a
to (that is,
measurement
might have introduced
efforts. This is a concern because
is affected by) the trade liberalization
and thus the
trade reforms reduced the prices of capital goods inMexico,
are biased upward after liberalization. We
TFP estimates
for Mexico
therefore
used
the Arellano-Bond
got our data
from the OECD.
50. We
from UNIDO,
differences
which,
estimator
in turn, received
to estimate
the Mexico
directly
51. Output
and capital
inputs were
PPP deflator series from the Penn World
and U.S.
the
data
in constant
1987 prices. The investment
expressed
Tables and the industry deflators
from Bartelsman,
end in 1996. We applied the average growth rate of the investment
and Gray (2000)
Becker,
PPP deflator for the available
years
to the rest of our sample
ending
in 1999.
26 ECONOM?A,Fall2003
model
in equation 2.52 This estimator helps reduce the influence of the
errors by using lagged levels of the TFP
biases induced by measurement
to instrument the changes
differentials
in these differentials.
Hence we
for unobserved
effects. Time effects, yt, are
industry-specific
for by the inclusion of year dummy variables.
In equation 2, the autoregressive
coefficient,
?, provides an indication
A coefficient of less than 1 can be interpreted
of the speed of convergence.
as evidence of convergence
in TFP levels between
the United States and
also control
controlled
was
associated
with
gence, then the estimated
able should be negative.
coefficient
of the corresponding
Mexico.
If NAFTA
an acceleration
of TFP
conver
interactive
vari
Results
6 reports the results from the Arellano-Bond
differences
estimator
to
2
for
the
model
additional
controls
suggested by equation
plus
applied
the potential
effect that Mexico's
unilateral
liberalization
(from 1985)
Table
focuses on the
The second model
have had on TFP convergence.
are
not affected
these
data
in
for
labor
since
gap
comparisons,
productivity
a
for
fixed
investment
deflator
the
the
lack
of
Mexican
by
twenty-eight
industries. In both cases, the models pass the specification
manufacturing
tests, indicating that the instrument set is adequate and there is no serial
might
are not biased owing
to
that the coefficients
suggests
error in the output series. Also,
in both cases, NAFTA was
associated with a faster rate of manufacturing
convergence,
productivity
as indicated by the highly
of the
and negative
coefficients
significant
correlation.
This
measurement
differen
dummy variable interacted with the lagged productivity
tial. The TFP results (column 1, table 6) imply that the half-life of a unit
to 0.7 afterward.
shock to the TFP gap fell from 1.6 years prior to NAFTA
NAFTA
The corresponding
change for labor productivity
from 2.5 to 1.7 years. These results are consistent
change in the degree
cussed above.
of persistence
(column 2, table 6) was
with the estimates of the
of the U.S.-Mexico
income
gap dis
that the NAFTA
results strongly
In sum, the econometric
suggest
a
in manufac
was
with
associated
faster
convergence
significantly
period
are
that the trade agreement
tempted to postulate
turing TFP levels. We
52.
Arellano
and Bond
(1991).
27
William Easterly, Norbert Fiess, and Daniel Lederman
on Manufacturing TFPConvergence3
The Effectof NAFTA
TA B LE 6.
(V
(2)
0.65*
0.76***
Explanatory variable
Log productivity differential (f-1)
x Logproductivity
NAFTA
differential
(f-1)
x Logproductivity
LIB
differential
(f-1)
-0.28*
-0.09***
-0.03
0.04
0.25
0.39
0.32
0.87
Specification test
Sargan overidentification test (p value)
Second-order serial correlation test [p value)
Summary statistic
No. observations
462
28
No. industries
482
28
Source: Authors'calculations.
***
Statisticallysignificantat the 1 percent level.
a. Thedependent variable incolumn 1 isthe logTFPdifferential(UnitedStates andMexico); in column2 it is the logoutput per
workerdifferential(UnitedStatesandMexico).Thefiguresreportedare first-stepestimatesof regressionsrunusingArellano-Bond
gen
eralmethod ofmoments. Thesampleperiod is 1980to 2000. Yeardummiesare not reported.
had an important positive effect on Mexican
manufacturing
with firm-level
evidence
results are consistent
provided
TFP. These
by L?pez
data presented by Schiff and Wang.53 How
and industry-level
the former study argues that this effect was related to preferential
C?rdova
ever,
access
but not to
import competition,
imports of intermediate goods. In contrast, the study by Schiff and Wang
benefited from imported intermediate goods from the
argues thatMexico
on the extent of R&D efforts
in the United
United
States, depending
market
to the United
States
States. Our results seem to indicate
and
brought something to the
table that was not necessarily
liberalization,
accomplished
by unilateral
but we have not speculated about the exact channels of influence. In our
view,
this issue remains
that NAFTA
an open question
for future research.
Initial Conditions and Divergence within Mexico
reviewed
the times series evidence
income conver
concerning
TFP
and
the
evidence
between
the
gence
convergence
panel
concerning
we now turn to the impact of NAFTA within
United States and Mexico,
If geography
Mexico.54
and initial conditions
play an important role in
Having
53. L?pez-C?rdova (2002); Schiff andWang (2002).
54.
This
section
is based
on Esquivel
and others
(2002).
28
ECONOM?A,
economic
income
Fall 2003
then NAFTA
convergence,
across Mexican
differentials
might
states.
have had a notable
impact on
It is standard practice in the analytical work on economic
growth to
a
set
of
in
determinants
of
entities
potential
growth
geographic
using econometric
techniques.55 Both Esquivel and Messmacher
apply this
examine
approach
to the case
of Mexico.56
Here
we
use
the same
standard
variables that deter
approach, but we focus on a small set of policy-related
state. The following
in each Mexican
mined
initial conditions
paragraphs
describe the data and methods
used to address these questions.
Data andMethodologies
to explain
the growth rate of state GDP per capita during
constant
1993 prices).57 This is the period during which
(at
trade liberalization and NAFTA must have been felt, and it is sufficiently
We
want
1990-2000
long that the cumulative growth rate during this whole period could reflect
such as
medium-term
rather than just short-lived conditions
phenomena,
the economic
crisis of 1995. Figure 8 shows the evolution of the ratio of
per capita GDP
Federal District
in a selection
of northern
and southern
states relative
to the
since 1940. The big story is,
(the capital of the Republic)
was
that
the
Federal
richer
and
District
stayed richer for the last
again,
or
so.
to
states
of
in
None
these
catch up significantly
sixty years
managed
has existed
terms, despite the fact that free trade within Mexico
for a long time. Also, it looks like the 1990s were characterized
by a slight
states
of
the
southern
the
northern
and
catch-up by
continuing divergence
absolute
states relative
to the Federal District.
factors might explain why some states grew more than others?
the role of geography
the issues raised by the literature concerning
one set of
costs in hampering convergence,
and transport or coordination
What
Given
variables
key explanatory
munications
infrastructure,
55.
See
the textbook
by Barro
indicators of transport and com
encompasses
which we measured by the kilometers of paved
and Sala-i-Martin
(1995).
56. Esquivel (1999);Messmacher (2000).
de M?x
from El Colegio
57. The data were graciously
Esquivel
by Gerardo
provided
in
of oil revenues, which
ico, Mexico
City. The GDP series were adjusted for the allocation
and Information,
Institute of Statistics,
the original
series (from the National
Geography,
to different
in practice
allocated
states, although
INEGI) had been periodically
they are
shares.
allocated
to
according
probably
population
William Easterly, Norbert Fiess, and Daniel Lederman
FIGURE 8.
29
Ratio of State GDPper CapitaRelative to the FederalDistrict, 1940 to 2000
State / FederalDistrict 6DP per capita
i_i_i_i_i_i_i_i_i_i__
1940
1950
1970
1960
1980
1988
1990
1993
1995
2000
Source: Authors'calculations.
highways per worker and telephone density.58 We also used the distance
from the U.S. border as an additional explanation
of economic growth to
assess the argument that being far from the United States was an impedi
ment
to growth.59
It is conventional
wisdom
that the level of education of the adult popu
lation might be related to the growth rate. Hence, we also examine
the
impact of educational attainment in 1990 as an explanation of growth rates
58.
with respect to the surface area of
because we would
need to know
however,
meaningful
territory. In any case, when we used the ratio
over surface area of each state, the results are virtually
identi
The coverage
of paved
state. This measure might
the surface area of economically
each
of paved roads or highways
cal to those discussed
herein.
59.
distance
The distance
from
from
the major
roads
the U.S.
city
could
also be
in each
be measured
imprecise,
border was measured
state
to the closest
in two alternative
major
city near
ways:
(1 ) by the
the border, plus the
30 ECONOM?A,Fall2003
In this way we can be sure that
the subsequent period 1990-2000.
cause
not
of
We also experimented
with
did
the
level
education.
growth
instead of the years of schooling.
literacy rates of the adult population
during
they receive
argued that poor states grow slower because
resources
to
One
such
finance
their
argument,
growth.
public
do not provide
sufficient
is that private capital markets
for example,
to
various
for
of
the
types
lagging regions owing
financing
development
It is often
insufficient
to private financing related to insufficient
information about
areas
to
of firms operating in those
pay back loans. However,
of obstacles
the capacity
it is also possible
that large public sectors can be a drain on economic
the local labor markets
(for example,
raising wages
growth by distorting
above what private enterprises can pay) or by raising the costs of capital
that would otherwise have gone to the private sector (that is, the so-called
To assess these alternative
of public expenditures).
at
the
size
of
the public sector, measured
look
the
of
arguments
impact
on the growth
as the share of public employment
in total employment,
states.
rates of Mexican
crowding
out effect
we
To
assess
Oaxaca?had
and
the really poor states?Chiapas,
Guerrero,
for
their prospects
that hampered
characteristics
a
included
dummy variable that identifies these states.
whether
other
we
development,
Finally, we included
tional convergence
the initial
level of per capita GDP
to test the condi
hypothesis.
Results
tech
7 reports some of our results, based on standard statistical
least
report results based on ordinary
niques. The first two columns
an
alterna
from
results
the
columns
and
fourth
and
third
report
squares,
Table
regressions, which is less sensitive to outliers. The
the initial per capita
of conditional
convergence;
in all four exer
coefficient
and statistically
significant
tive technique, median
table shows evidence
GDP
has a negative
It thus seems
that poor states do grow faster if they have similar poli
cies to the rich states.
cises.
The other explanatory variables, except the variable that identifies the
states (Chiapas, Guerrero, and Oaxaca),
also seem to be important
for growth, and they are generally
statistically
significant. As expected,
southern
distance
of the latter to the border
tal city of each
state
to the closest
itself;
major
and
U.S.
(2) by
city.
the geographic
distance
from
the capi
William Easterly, Norbert Fiess, and Daniel Lederman
TABLE 7.
31
Potential Determinants of Growthof State GDPper Capita, 1990 to 2000a
variable
Explanatory
(2)
(1)(3)
(4)
1990
InitialGDPper capita,
(in natural logarithm)
Initialeducation (years of schooling
-0.15**
-0.15**
-0.14**
-0.12**
(-2.35)
(-2.32)
(-3.95)
(-2.09)
0.24
0.22
over15yearsofage),1990
ofpopulation
0.08*
Telephone
density,1990
(1.38)
(1.09)
(3.40)
0.08*
(1.91)
(1.93)
Public employment (log of share
1990
total employment),
of
0.27*
0.05**
(2.86)
(1.86)
0.05
(1.39)
-0.12**
-0.12*
-0.07*
-0.09
(-2.13)
(-1.98)
(-1.97)
(-1.54)
-0.01
Not included
Not included
States of Chiapas, Guerrero, and Oaxaca
(dummy variable)
0.27**
-0.021
(-0.02) (-0.33)
Summary statistic
0.31 0.28
Adjusted/?2
Pseudo/?2
No. observations
0.21
0.21
32 32 32 32
Source:Authors'calculations.
*
**
Statisticallysignificantat the 10percent level; statisticallysignificantat the 5 percent level.
a. The regressionsestimate the effect of a 1percent increaseinthe correspondingvariableon the cumulativeGDPgrowth rateper
capita, 1990-2000. Columns1and 2 are estimatedusingOLS;columns2 and 3 are estimatedusingmedian regressions.A constantwas
were
includedinthe regressions,but itscoefficientsare not reported.NumerousadditionalspecificationsinOLSandmedian regressions
estimatedusingthe followingexplanatoryvariables:(a) literacyrates insteadof yearsof education;(b)two alternativemeasures of dis
tance fromthe UnitedStates insteadof and inadditionto the Chiapas,Guerrero,andOaxacadummy;(c)paved roadsand double-lane
highwaysover surfaceareaor perworker insteadof telephonedensity; (d)the shareofmanufacturingGDPover totalGDPin1988;and
(e) urbanizationrates.See the text fora discussionof the alternativeresults.Finally,t statisticsare inparentheses.
effect on growth. However,
estimates
density has a positive
over
roads
and
two-lane
roads
worker
surface
(or
per
paved
using paved
correlated with growth
area) reveal that these variables were negatively
telephone
that building
there is no evidence
during the period.60 Hence
suggesting
more roads will lead to higher growth in the future. This result might be
due to the existence of economically
infrastructure that does
unnecessary
not serve a useful purpose
The results concerning
for existing economic
activity.
the role of distance from the U.S.
border (not
was
a
not
that
this
variable
statistically
significant
reported here)
to economic
impediment
growth in most exercises,
although the coeffi
the distance variables
cient is always negative.61 However,
introducing
not
the statistical
the direction
drove down
of the
(but
significance
indicate
60.
District,
These
which
OLS
has
had relatively
high
61. We estimated
note
one
four models
with
a sample excluding
the Federal
to high population
density and which
used
the two distance
variables
discussed
above
in foot
were estimated
In only
via OLS and two via median
regressions
regressions.
was
at the 10 percent
the distance
variable
these four models
level,
significant
37. Two
of
results did not change when we
low paved roads per worker owing
rates of growth.
32 ECONOM?A,Fail2003
variables. This evidence
indi
effects) of the other explanatory
that the states located farther from the United States suffer from low
estimated
cates
of education
levels
and telephone
which
density,
hamper
their growth
prospects.
at the beginning of the period has no statistically
on
important impact
growth in the OLS estimates. This result might be due
to the fact that human capital can migrate to dynamic regions, and thus this
variable does not have any discernible
it
impact on the states for which
level of education
The
was
calculated
literacy rates were used instead of educa
coefficient was positive and statistically
the estimates based on median
force
regressions
in 1990. When
tional attainment,
the estimated
significant. Moreover,
fully show that educational
between
attainment
does
initial GDP
matter.
The
correlation
and initial education
per capita,
telephone density,
of the impact of education rather difficult.
might make the identification
The share of public employment
had a negative effect on economic
shows
9
the
correlation
between
these two vari
activity. Figure
simple
it is negative. This negative correlation might be due to some obser
the median
vations that appear in the lower right of the chart. However,
are
to
influ
which
less
be
estimates,
likely
disproportionately
regression
ables;
enced by strange observations,
also show that this variable had a negative
effect on economic
growth although it is not statistically
significant, after
of the southern states (see
for other unobserved
characteristics
controlling
the fourth column
of table 7).
in growth
To be sure that these explanations
of the observed differences
we conducted
rates across Mexican
states are not misleading,
additional
in which we controlled
for the share of manufacturing
exercises
produc
in Esquivel and others, the
tion over total state GDP in 1990. As discussed
southern
states have never had a high share of manufacturing
production,
as a whole
some manufacturing
industries (and some
of manufac
grew quite rapidly in the 1990s.62 The performance
and for the country
services)
turing relative
to natural resource or agricultural
industries could have
been due to changes in relative prices. For example, the international price
of coffee began to decline in the late 1980s. Our statistical analyses indi
although
fications.
other
62.
several
of the other
were
These
results
between
variables
explanatory
are due to the correlation
explanatory
variables.
Esquivel
and others
(2002).
in these speci
also not significant
and the
the distance
variables
William Easterly, Norbert Fiess, and Daniel Lederman
33
Relation between Growthand Public Employment inMexican States, 1990sa
FIGURE 9.
Cumulative growth of income per capita, 1990-2000 (percent)
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
Public employment, 1990 (percent of total)
Source: Authors'calculations.
a. y=-2.2719x+0.0787;tf2 = 0.1282.
cated that the qualitative nature of the OLS results presented in table 7 are
share of production.
by the inclusion of the manufacturing
the inclusion of the share of
in the relevant median regressions,
However,
not affected
affected the sign of the education and public
production
variables,
employment
although none of them were statistically
signifi
cant. This influence of manufacturing
production on the estimated effect
manufacturing
of education
tion between
a negative
could stem from a positive correla
and public employment
education and manufacturing
(which is 0.5) and
production
correlation with the share of public employment
(which is,
seems to
In other words, manufacturing
-0.5).
production
coincidentally,
in states with either high levels of education or low levels
be concentrated
of the high mobility
of new cap
of public employment.
The combination
ital and the relative irre ver sibility of past investment probably makes
sensitive
activities particularly
capital-intensive
in a state, such that manufacturing
environment
to the initial economic
is implicitly
capturing
34 ECONOM?A,Fall2003
things such as the rule of law, instability,
by the state.
crime, or excessive
intervention
thus suggests that hope for the southern states is not lost:
of conditional
and some key policy
convergence,
the
of
economic growth observed
variables help explain
patterns
Our evidence
there is some evidence
sensitive
across Mexican
states during
(measured by telephone
associated with economic
positively
In particular, communications
density) ismore likely to have been
activity than paved roads or high
1990-2000.
infrastructure
ways. Also, the evidence does not support the idea that increasing the size
of the public sector can be a force for economic
However,
convergence.
seem to have had important effects
the big story remains: initial conditions
on economic growth within Mexico
in the 1990s. States that were initially
better prepared to reap the benefits of NAFTA
grew faster during this
period, while
the poor states of the south fell further behind.
Conclusions and Final Remarks
and sources of convergence
the dynamics
analyzed
and the United States. Time series analyses of the con
facts about the U.S.
process
interesting
stylized
produced
and
identified
and
convergence
process
periods of convergence
This
paper has
between Mexico
vergence
Mexican
suffered a major setback in the 1980s as a
convergence
divergence. While
result of the debt crisis, the tequila crisis only temporarily
interrupted a
when
Mexico
that
started
in
the
late
1980s
convergence
process
opened its
we only found evidence of incomplete convergence,
economy. However,
sense
in the
that the constant in the cointegration
space was greater than
toward a constant income dif
is converging
zero, indicating that Mexico
ferential
between
of about 50 percent of the U.S. GDP per capita. The comparison
relative income effects and average Latin Amer
annual Mexican
toward the United States
ican effects indicated thatMexico's
convergence
was especially
after
1995.
important
in institutional fea
evidence showed that differences
The cross-country
income
from history play an important role in producing
of
effects
much
estimated
estimates
larger
produced
on incomes than the OLS estimates,
thus indicating that mea
error is an important source of attenuation bias in these rela
tures inherited
gaps. The
institutions
surement
2SLS
William Easterly, Norbert Fiess, and Daniel Lederman
35
tionships.63 The use of historical instruments for current institutional qual
ity is also interesting on its own, since institutions tend to persist over time
for a long time.
and thus might remain a source of income divergence
research could yield additional
of institutional quality.
determinants
about the role of political institutions
Future
practical insights if it focuses on the
In particular, further understanding
the quality of gover
in determining
what
could
types of reforms may
help identify
policy
overcome
Recent
of
research
the weight
history.
along these lines has
help
already proved fruitful.64 Yet little is known about how accountability
can help improve national institutions.
In the case of North
mechanisms
nance and economic
in the long run might
economic
international
convergence
to
on
to
the
standards
of its neighbors.
catch up
Mexico's
capacity
depend
In fact, the econometric
indicated that the model with institu
analyses
America,
and trade predicts an income gap of the Mexico-U.S.
tions, geography,
similar
GDP per capita ratio of about 54 percent, which is coincidentally
to the incomplete
estimated
convergence
using cointegration
analysis.
institutions did not improve signifi
the quality of Mexican
Furthermore,
cantly more than those of other Latin American
NAFTA period.
countries
during
the post
within manufacturing
The analysis of TFP convergence
industries pro
the impact of NAFTA. The evi
results concerning
duced more optimistic
in the rate
dence indicates that NAFTA was associated with improvements
these
the United States and Mexico. While
of TFP convergence
between
results are broadly consistent with other studies, these studies contradict
in terms of the channels through which NAFTA
is thought to
have improved Mexican
TFP.65
manufacturing
Namely,
L?pez-C?rdova
each other
access to the U.S. market (for example,
the
argues that it was preferential
tariffs faced by Mexican
exporters to the United States) and import pene
tration, but not imports of inputs from the United States. Schiff and Wang
were due to the R&D content of imported
argue that TFP improvements
inputs. We can also think of other alternative hypotheses.
One possibility
is that NAFTA,
either through its demanded
ment in the protection of intellectual property rights or through
63. This is consistent
with previous
studies,
including Acemoglu,
son (2001) and Kaufmann
and Kraay
(2002b).
64. Persson
and Soares
(2002); Lederman,
(2002).
Loayza,
65. L?pez-C?rdova (2002); Schiff andWang (2002).
Johnson,
improve
increased
and Robin
36 ECONOM?A,Fall2003
international
competition
incentives
tries), provided
Meza
and
Mora,
patenting.
and exporting
indus
(for import-competing
in private R&D efforts and
for improvements
as well as Lederman and Maloney,
find that the
increases in
post-NAFTA
period was, in fact, characterized
by significant
R&D expenditures.66
inventors improved
Patenting activity by Mexican
as
this
well.
Yet
the
significantly during
existing literature remains
period,
this particular force toward convergence.
An examination
of
on
of patent
these issues would require empirical work
the determinants
a
on
across
focus
the
of
trade
with
countries,
impact
policies
ing
special
silent about
policies. Much work remains to be done in this area,
an emerging
is
literature.67 Lederman
and Maloney
show
there
although
that, in fact, the protection of intellectual property rights tends to increase
and
innovation
relative to GDP in a broad panel of countries and that these
are
in the sense that they tend to rise with improve
expenditures
cyclical
ments
in short-term growth.68 It is thus very likely that NAFTA
helped
its
intellectual
Mexico
its
innovation
property
rights
improve
through
R&D
efforts
regime and by helping Mexico
hand, Lederman and Maloney
recover
after the tequila crisis. On the other
also show that the emerging manufacturing
sectors under NAFTA
telecommunications
(namely, road vehicles,
equip
are not yet characterized
and appliances)
by significant
improve
in patenting
the presence of significant
suggests
activity, which
related to the lack of linkages between R&D per
efficiency
problems
ment,
ments
formed
the public
by
and higher-education
sectors
and the productive
sector.69
showed
Our study of growth patterns within Mexico
during 1990-2000
states grew faster. We
that initial conditions
determined which Mexican
interpret this evidence as showing that trade liberalization might be asso
economic
ciated with
initial conditions.
divergence
In the Mexican
within
countries owing
telecommunications
to differences
in
infrastructure
case,
and human capital were especially
important. In addition, it is commonly
and
that the poor states suffer from poor public institutions
understood
states
faster
The
have
grown
poor
during this
might
instability.70
political
period
if they had been
adequately
68.
(2002); Lederman
Porter, and Stern (2002).
and Maloney
Lederman
(2003b).
69.
Lederman
70.
Esquivel
66. Meza
67.
and Mora
Furman,
and Maloney
(2003a).
and others (2002).
prepared
and Maloney
to reap the benefits
(2003a).
of free
William Easterly, Norbert Fiess, and Daniel Lederman
37
in North America might not materialize
trade. Economic
convergence
under free trade or under any trade regime as long as fundamental differ
ences in initial conditions
some of these
persist over time. Fortunately,
fundamentals
should be sensitive
to policy
changes.
Appendix: Supplementary Data
TABLE A1. SummaryStatistics for Data Used for EconometricResults on Institutional
Gapsand IncomeGaps (Figures4 and 7)
No.
Variable
Sample3
Group1
observations
ICRG
variables"
GDP
Log(country's
GDP
percapita/USA
Standard
Mean
deviation
Minimum Maximum
414
-0.4069638
0.558766
-1.75361
923
-1.715673
0.579324
-3.65967
162
-0.1312372
0.4356544
-1.00386
378
-1.328616
0.3673385
-2.19757
0.6972296
-0.3095284
per capita)
2
Group
ICRG
variables"
GDP
(country's
Log
GDP
percapita/USA
percapita,PPPadjusted)
0.6972296
-0.3095284
a. Group 1:Argentina,Bolivia,Brazil,Chile,Colombia,CostaRica,Cuba,DominicanRepublic,Ecuador,ElSalvador,Guatemala,
Guyana,Haiti,Honduras,Jamaica,Mexico,Nicaragua,Panama,Paraguay,Peru,Trinidadand Tobago,Uruguay,andVenezuela;Cuba is
not includedinthe GDPsample.Group2: Argentina,Brazil,Chile,Colombia,CostaRica,Mexico,Peru,Uruguay,andVenezuela.
b. Weighted averageof the ICRG
variables(absenceof corruption,lawand order,and bureaucraticquality).
TABLE A 2. SummaryStatistics for Data Used inAnalysisof InstitutionalGaps
and IncomeGaps
No.
Variable
observations
Standard
Mean
68
0.1323529
Landlocked
Openness (Sachs andWarner, 1995)
Log constructed trade share
deviation
Minimum Maximum
0.3413936
0
63
0.2252768
0.3423797
0
68
2.721456
0.7672238
0.94
68
Latitude
6.318064
1
1
4.586000
(Frankel-Romer)
Ethnolinguisticfractionalization
EastAsiaandthePacifie
Oilproduction
Commodity
Institutional
per
LogGDP
dummy
dummy
index
capita
19.691030
1
61
46.377050 29.430240
68
Africa 0.3382353 0.4766266
68Asia
0.0588235 0.2370435
South
68
0.0735294 0.2629441
0
68
Americas
0.3970588 0.4929263
68
0.2647059 0.4444566
68
0.6764706 0.4713010
68
-0.1134657
68
4.588946
Logmortality
68
7.794468
0.7704978
1.2550750
1.1091530
-41.81407
61.06258
90
0
0
1
1
1
0
0
0
-1.978333
2.145931
5.252923
1
1
1
1.585833
7.986165
10.031100
38 ECONOM?A,Fall2003
TABLE A3.
Mexico: SummaryStatistics of Variables and Data Used for TFPConvergence
Analysis, by Industry
(Log)
Industrycode3
311
313
314
321
322
323
324
331
332
341
342
351
352
353
354
355
356
361
362
369
371
372
381
382
383
384
385
390
output
15.77
15.08
13.65
14.35
13.11
12.52
12.86
11.91
12.49
14.61
13.29
14.98
15.09
13.23
12.72
13.66
14.00
12.08
13.81
14.41
15.38
14.31
14.24
14.02
14.64
15.95
12.15
12.21
(Log)
Labor
(Log)
Obs
labor
25
25
25
25
17
7
17
25
17
25
17
25
25
7
25
25
17
17
25
25
25
25
25
25
25
25
17
17
12.98
13.87
12.72
13.68
10.36
11.50
12.50
13.41
11.33
11.44
10.01
10.87
11.19
11.70
9.85
11.77
10.49
10.55
12.08
14.35
11.38
11.67
12.48
14.16
12.89
13.49
10.49
11.94
Obs
capital
9.84
12.44
11.69
12.90
11.83
12.70
10.13
9.04
11.86
13.12
12.05
14.36
12.59
14.84
11.34
12.73
12.08
12.58
11.78
11.97
12.57
13.02
13.15
14.22
9.76
10.19
10.34
10.86
Source: UnitedNations Industrial
DevelopmentOrganization(UNIDO),
a. See tableA5 fora listof the industriesby code.
Obs
share
Obs
25
25
25
25
17
7
17
25
17
25
17
25
25
7
25
25
17
17
25
25
25
25
25
25
25
25
17
17
0.06
25
25
25
25
17
7
17
25
17
25
17
25
25
7
25
25
17
17
25
25
25
25
25
25
25
25
17
17
0.10
0.04
0.16
0.17
0.08
0.19
0.13
0.14
0.08
0.15
0.09
0.11
0.07
0.06
0.14
0.12
0.14
0.15
0.10
0.07
0.06
0.12
0.11
0.13
0.07
0.10
0.16
William Easterly, Norbert Fiess, and Daniel Lederman
TABLE A4. United States: SummaryStatistics of Variables and Data Used for TFP
ConvergenceAnalysis, by Industry
Labor
(Log)
(Log)
(Log)
Industrycode3
output
Obs
labor
Obs
capital
Obs
share
Obs
19.47
25
17.06
25
18.08
25
25 0.09
17.50
25
15.19
25
16.74
25
25 0.10
16.85
18.14
17.64
25
25
25
14.03
16.45
16.12
25
25
25
15.37
17.21
15.86
25
25
25
250.06
25 0.18
25 0.22
15.35
25
13.66
25
14.11
25
25 0.19
15.32
25
13.81
25
14.72
25
25 0.22
17.64
25
15.92
25
16.79
25
25 0.18
17.27
25
15.86
25
15.61
25
250.24
18.46
18.57
18.67
25
21
25
16.58
17.21
16.36
25
21
25
18.15
17.48
18.54
25
21
25
250.15
21
0.26
250.10
18.46
25
16.36
25
17.34
25
25 0.12
18.62
25
14.86
25
17.90
25
25 0.02
16.58
21
13.88
21
15.12
21
210.10
16.99
25
15.45
25
16.20
25
250.21
17.95
14.72
25
25
16.32
13.56
25
25
16.93
14.03
25
25
25 0.19
25 0.32
16.64
25
15.14
25
16.15
25
17.62
69
25
15.97
25
16.92
0.19 25
71
18.09
17.74
18.73
25
25
25
16.43
15.69
17.25
25
25
25
18.15 0.19 25
16.97
25
17.62
25
19.31
25
17.78
25
18.21
25
25 0.22
19.15
25
17.60
25
18.07
25
250.22
19.66
18.21
17.25
?90
25
25
25
17.88
16.81
15.71
25
25
25
18.43
25
16.98
25
16.10 0.21 25
Source: UnitedNations Industrial
DevelopmentOrganization(UNIDO),
a. See tableA5 fora listof the industriesby code.
250.23
25
25
250.13
25 0.23
250.17
250.25
25
39
40 ECONOM?A,Fall2003
TABLE A 5. Listof Codesand IndustriesUsed in
TFPConvergenceAnalysis
?SIC
Code
Industry
311 Foodproducts
313 Beverages
314
Tobacco
321Textiles
322 Wearing apparel, except footwear
323 Leather products
324 Footwear, except rubberor plastic
331 Wood products, except furniture
Furniture, except metal
341 Paper and products
332
342 Printing
andpublishing
351 Industrial chemicals
352 Other chemicals
353 Petroleum refineries
354 Miscellaneous petroleum and coal products
355 Rubber products
356 Plastic products
Pottery, china, earthenware
362 Glass and glass products
369 Other nonmetallic mineral products
361
371
Ironand steel
372 Nonferrous metals
381 Fabricatedmetal products
382 Machinery, except electrical
383 Machinery, electric
384 Transport equipment
385 Professional and scientific equipment
390 Other manufactured products
Comments
Bill Easterly, Norbert Fiess, and Daniel Lederman
V. Loayza:
in North
have written a serious and comprehensive
study on convergence
after NAFTA. The authors approach the subject from many dif
America
ferent perspectives,
perhaps to make up for the little time available for
Norman
a definitive
of NAFTA's
evaluation
aftermath. The paper's
conducting
to making
has contributed
is the extent to which NAFTA
main question
to
that
of
the
United
and
income
closer
States
Canada.
Mexico's
per capita
To provide an answer that would address the various aspects of the ques
and microeconomic
the authors examine macrodata; use time
and panel econometric
and consider
series, cross-sectional,
techniques;
cross-state
both cross-country
and (Mexican)
evidence. This may seem
but there is a rationale for each exercise. Microeconomic
excessive,
(firm
tion,
level) data can resolve aggregation biases and concentrate on productivity
in specific
industries. Macroeconomic,
and
time-series,
convergence
can
common
events
for
control
evidence
inter
cross-country
taking place
for comparison,
and thus help us under
nationally, provide a benchmark
stand the effects of the unique Mexican
experience with NAFTA. Finally,
cross-state evidence allows an evaluation of the differing effects
Mexican
on Mexico's
of NAFTA
regions,
a necessary
undertaking
given
this coun
try's large size and diversity.
to the paper's emphasis on income convergence
A possible objection
should consider other more
could be that a proper evaluation of NAFTA
such as trade volumes
and
aspects of the agreement,
prices, foreign investment flows, capital costs, and innovation trends. This
is unwarranted, however, on considering
that this paper is part
objection
more generally
of a larger research project that evaluates NAFTA
and
relevant
or direct
and other Latin American
countries.
implications for Mexico
are
this
in
from
the
volume
collected
project
being
resulting papers
Lessons from NAFTA, edited by Daniel Lederman, William Maloney,
and
Luis Serven.
draws policy
The
41
42
ECONOM?A,
The
authors
NAFTA
Fall 2003
arrive
at a nuanced
correctly
reflects
income
closer
success.
to that
are preventing Mex
this
partners faster. I believe
and limitations of NAFTA
the achievements
of the U.S., but institutional and governance
to its North American
ico from converging
conclusion
on NAFTA's
conclusion
to bringing Mexico's
has indeed contributed
factors
on income
I
convergence
up to this point. At the end of my comments,
offer additional evidence
supporting it.My criticism of the paper resides
not in its conclusions,
but in some of its methodology.
Isa DynamicProcess
Convergence
The
the issue of convergence
from two differ
ent methodological
In their firm-productivity
and cross-state
standpoints.
as
a
transitional
convergence
analyses, they regard
dynamic,
phenomenon.
authors
implicitly
address
To examine
vari
it, therefore, they estimate dynamic
(lagged-dependent
most
treatment
This
is the
of convergence
for
able) models.
appropriate
countries. Conversely,
when
the authors turn to their cross
developing
country analysis,
The econometric
as a steady-state phenomenon.
they regard convergence
to
this perspective
is the estimation of static
counterpart
models, based on the comparison of output levels via cointegration
analy
sis or cross-country
how
regressions. This is of only limited usefulness,
ever, for countries that are rapidly evolving.
The first consideration
is whether cointegration
analysis can help deter
extent
to Bernard
mine
the
of income convergence.
and
According
two countries exists if the long
between
Durlauf,
long-run convergence
run forecast of their output difference
is stable.1 The challenge
for imple
this concept is how to assess the long-run stability of the income
menting
to use cointegration
choose
difference.
and Lederman
Fiess,
Easterly,
can
to
if the
said
be
U.S.
and
Mexican
income
be
analysis:
converging
per capita output series cointegrate
between
is the case, the stationary difference
countries'
vides
a measure
denotes
of the extent
of convergence,
a (1,-1) vector. If this
the two income levels pro
with
in which
convergence.
The problem with this approach is that it requires
be stable over
ference between
the two countries
whereas
stable
a zero difference
absolute
that the income
the concept of convergence
only requires
in the long run. The cointegration
approach would
1. Bernard
dif
the sample period,
that this difference be
and Durlauf
(1995).
be appropriate
if
William Easterly, Norbert Fiess, and Daniel Lederman 43
FIGURE 10.
Steady-State and TransitionalConvergence
the two countries
had already arrived at their steady states, but it is incor
if they are at different points on the path toward their
rectly
as Mexico
and the United States are bound to be. The
long-run positions,
first panel of figure 10 represents the convergence
process that is implied
restrictive
in cointegration
tries are always
analysis:
the income
the two coun
between
differences
stable over the sample period. The second
convergence
process, in which the income
(stochastically)
panel represents amore general
difference
is allowed a transition
period and stability occurs only toward
and the United States for the sample period
are probably not
the income differences
(1960-2002),
the end. In the case of Mexico
under consideration
stable, but declining
(as in the left portion of the second panel). Therefore,
as statistically,
as
well
analysis may not be
conceptually
cointegration
for analyzing
appropriate
I now
turn
to
convergence
income-level
regressions.
in this case.
After
years
of
cross-country
to rely on output-level
com
it has become fashionable
growth regressions,
so
across
to
the
level
of
is
different
development
parisons
explain why
countries. Acemoglu,
and Easterly and Levine are
Johnson, and Robinson
two of the most influential papers of this literature.2 An evaluation of this
approach should start by asking what is likely to explain output differences
among countries. The answer depends on how the world distribution of
output across countries behaves over time. If this distribution has achieved
across countries
would
be
output differences
in
that
the
is, highly persistent country
very long run,
explained by factors
such as political
and social institutions
and economic
characteristics
its
steady
state,
then
can be identified with
power relations. In this case, output differences
measures
if the world distribution
of the extent of convergence.
However,
of output across countries is changing over time?following,
for instance,
2. Acemoglu,
Johnson,
and Robinson
(2001);
Easterly
and Levine
(2003).
44
ECONOM?A,
FIGURE 11.
Fall 2003
U.S.andMexican IncomeDifferences inthe Transitiontoward the
Steady State
U.S. income
Mexican income
*t-1
a dynamic
transition pattern?then
among countries
output differences
would be explained not only by such long-run factors, but also by eco
nomic policies,
international and domestic
shocks, and, most importantly,
the
In this case, we would be interested in measuring
initial conditions.
con
extent
of
of
rather
than
the
(a dynamic concept),
convergence
speed
(the static counterpart).
vergence
Consider
the stylized paths of output over time for the United
to understand
in figure 11. If one wants
the output
Mexico
States and
difference
in the steady state (Tss), it is best to focus on
if both countries are evolving
dynami
long-run factors only. However,
not
tell the whole story with regard to out
these
factors
do
long-run
cally,
between
the two countries
at, say, time t. Initial conditions,
put differences
represented by the output
in the previous period, t - 1, are likely to be fundamentally
difference
current differences
between Mexico
and the
in explaining
important
United
States.
How
do these points translate into econometric
Given
specifications?
across
not
of output
countries has
reached a
that the world distribution
steady
state, a static output
(Y) regression
is misspecified:
William Easterly, Norbert Fiess, and Daniel Lederman 45
(1) I?
=
7Z?+e?.
It should be replaced by a dynamic regression
tial conditions
(Yit_A) and shocks and policies
institutional
factors
that takes into account
(X), in addition
ini
to long-run
(Z):
(2) ^^a^+?X.+yZ^e,,,
the subscripts / and t represent country and time, respectively.
in natural logs, then
If, as is standard, output per capita is expressed
as
a
can
2
be
rewritten
dynamic equation
growth regression, which can be
where
estimated
cross-country
using pooled
(3) YUt Y^ =(a-
and time-series
data:
\)YU_X+ ?*,r + yZ, + e,,.
EvidencefromGrowthRegressions
a return to growth regressions,
I now use this method
that NAFTA has indeed had a pos
evidence
advocated
Having
to offer circumstantial
ology
itive effect
on Mexico's
The exercise
growth performance.
supports
conclusion
that Mexico
has approached
Fiess, and Lederman's
in terms of per capita income after NAFTA
started.
Easterly,
the U.S.
The
like to present is taken from a recent paper on
in
Latin
America
and the Caribbean, written by C?sar
growth
Pablo Fajnzylber,
and myself.3 There, we estimate a growth
and
using panel data on a worldwide
sample of countries
evidence
economic
Calder?n,
regression
I would
nonoverlapping
five-year periods spanning 1960-99. We consider a large
of
variety
growth determinants, which we group into categories related to
transitional convergence,
cyclical reversion, structural reforms (including
institutional
and external conditions. We
factors), stabilization
policies,
control
for unobserved
effects and the likely endogeneity
country-specific
of the explanatory
variables. We use the estimated parameters
to explain
the growth changes experienced by individual Latin American
countries in
recent decades. This model can be applied to account for the change in
to 1996-99,
that is, roughly
growth rates from 1991-95
NAFTA. Table 8 shows the results for Mexico.
3. Loayza,
Fajnzylber,
and Calder?n
(2002).
before
and after
46 ECONOM?A,Fall2003
TABLE 8.
Mexico: Determinants of the Change inGrowth Rates, 1991-95 to 1996-99
Percentage points
Growthdeterminant Projected contribution to change ingrowth rate
Transitional convergence
Cyclical reversion 1.23
0.03
Structural reform0.66
Stabilization policies
External conditions
0.31
0.06
Total projected change
Actual change 3.88
Growth premium (actual
2.29
projected)
1.59
The actual change in growth rates inMexico
before and after NAFTA
3.88 percentage points, while the projected change was only 2.29 per
a growth premium of 1.59 per
thus experienced
centage points. Mexico
for a
centage points that we cannot explain despite having accounted
was
set of growth determinants,
initial conditions,
comprehensive
including
financial development,
trade
inflation rates,
burden,
volume,
government
rate misalignment,
real exchange
financial crises, terms of trade shocks,
infrastructure facilities, and world conditions.
Still, this growth premium may not be particular
could be attributable to either a feature of the model
to other countries,
particularly
in Latin America.
toMexico,
but rather
or an event common
To dismiss
this possibil
TABLE 9. LatinAmericanCountries:Difference between Actual and Projected Change in
GrowthRates, 1991-95 to 1996-99
Premium
Country
-2.04
Argentina
Bolivia
0.36
Premium
Country
Honduras
Nicaragua0.99
0.01
Paraguay
-3.12
Brazil
-2.08
Chile
-2.54
Colombia
1.27
Costa
Rica
0.30
Ecuador
El Salvador -2.79
Mean -1.20
Median
-1.30
-1.61
1.59
Mexico
Peru
Uruguay
-1.75
-0.94
Venezuela -3.86
William Easterly, Norbert Fiess, and Daniel Lederman 47
we per
ity and verify that the size of this premium is unique toMexico,
the change in growth rates between
form the same exercise of explaining
countries (see table 9).
1991-95 and 1996-99 for fifteen Latin American
Mexico
has the largest growth premium of all the countries in the sample,
only by Costa Rica. Sixty percent of the countries had
negative growth residuals, with the typical country in the region having an
shortfall in the growth rate of more than one percentage point.
unexplained
followed
closely
Granted, this evidence
circumstantial.
Given
on the beneficial
the little
itmay be the most
event, however,
impact of NAFTA
time available
for judging
telling macroeconomic
is indirect, or
such a large
evidence
at our
disposal.
Fiess, and Lederman
present the following
Easterly,
are
between
scheme, (i) There
the United
per capita income differentials
on
States and Mexico,
its
Mexican
(ii) NAFTA,
impact
through
positive
a
should
of
income,
convergence
per capita
(iii) The
generate
growth,
Patricio
Meiler:
authors
test the existence
of convergence
income differentials
Mexico's
per capita
those of selected Latin American
To
accelerate
the evolution of
by comparing
vis-?-vis
the United States with
countries;
Mexican
little difference
is found,
(iv)
Mexican
convergence,
they suggest
improving
the paper also addresses a quite different
(v) Finally,
on regions within Mexico.
the differential effect of NAFTA
institutions,
issue,
namely,
in the title its main problem: the "big events, lit
The paper recognizes
tle time" effect. Can a long-run phenomenon
like income convergence
an
to
event
like NAFTA, which has so far
really be measured with regard
had only a marginal
impact? In a nutshell, a free trade negotiation
implies
are divided
the following. Goods
to the
into three categories
according
zero:
at
to
tar
which
its
will
be
reduced
the
fast
tariffs
whose
group,
speed
iffs are reduced
to zero at the time of the signature of the free trade agree
in which
the medium-speed
tariffs are reduced to zero
category,
ment;
over three to four years;
years. The fast category
takes more than four
set, which
a marketing
it includes
ploy, because
and the slow
is really
not much can
those goods that already have a zero tariff. Consequently,
in
first
three
The
the
dataset
in
used
the
years.
paper covers only
happen
three years after NAFTA was signed. Breaking
the annual data into quar
ters increases
span. The
the number of observations,
but it cannot increase the time
to trying to measure the economic
task of the paper is equivalent
48 ECONOM?A,Fall2003
of America
in Spain prior to the year 1500. The
impact of the discovery
a long-run phenomenon
time period is simply too short to measure
like
convergence.
A more
to
involves
the paper's
reference
suggestion
could test their methodology
with the Puerto
to check
Rican case, that is, they could use the first five years (1960-1965)
the income convergence
forecast forty years later. In the first five years of
Puerto Rico.
the Puerto
respect
constructive
The
Rico
to the U.S.
omy forty years
to the data, even
as the United
authors
the per capita income differential
(with
experience,
income level) went from 0.30 to 0.36. It took the econ
to reach half the U.S.
though Puerto Rico
States.
per capita income level, according
had the same institutional framework
NAFTA
constitutes
the paper's main explanatory
factor, yet the authors
do not use a trade theoretical framework. The theoretical framework usu
the
ally used to link trade and the labor market is the Stolper-Samuelson
orem. A key mechanism
on
is the
for explaining
the trade effect
wages
behavior of prices. There are several papers examining
the U.S.-Mexican
evolution
of
wage differential
during the 1990s; there is no comparison
in this
the empirical
results of this literature with the results obtained
paper.
The empirical methodology
used in the paper is the so-called dynamic
that
econometric
data for many
is,
convergence,
regressions
combining
countries.
Income per capita is the left-hand-side
variable, and on the
right-hand side appear all sorts of ad hoc and arbitrary variables; there is
no limit to the number of variables
included, and the empirical measure
ment of most of them is highly questionable.
This has become a standard
in the literature, but I have serious doubts that this type of
procedure
research generates anything useful. Institutions, for example, have become
a key explanatory
how
variable today, yet I have problems understanding
on
not
and
it
is
clear
the right-hand side of the regressions,
they appear
how
they are measured.
important trade-related
in the
issue involves Mexico's
competition
own
that
Chinese exports com
U.S. market. In my
research, I have found
pete with Mexican
exports in the U.S. market. How does this fact influence
An
the Mexican-U.S.
convergence?
NAFTA
What
FTA
with
the so-called
certainly
tequila crisis.
helped Mexico
to Mexico
in 1994 if there had not been a
would have happened
agreement?
Perhaps Mexico
would
have followed
a path similar
to
William Easterly, Norbert Fiess, and Daniel Lederman
49
the external crisis of 1982. The macroeconomic
that following
stability of
to the Mexican
main contribution
the 1990s may represent NAFTA's
economy.
in the paper. On the one hand, institu
Finally, there is an inconsistency
the main explanatory
factor of why Mexico
tions are considered
has not
achieved faster convergence with the United States. On the other hand, the
reveals that some regions have
convergence
analysis of regional Mexican
a
than others. However,
all regions have the same
had higher convergence
type of (Mexican) institutions.
rates?
convergence
How,
then, could
some regions have higher
50 ECONOM?A,Fall2003
References
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. "The Colonial
Origins of Comparative Development: An Empirical Investigation." American
Economic Review 91(5): 1369-401.
Acemoglu, Daron, and Fabrizio Zilibotti. 2001. "Productivity Differences." Quar
terly Journal of Economics 116(2): 563-606.
Arellano, Manuel, and Stephen Bond. 1991. "Some Tests of Specification for
Panel Data: Monte Carlo Evidence and an Application to Employment Equa
tions." Review of Economic Studies 58(2): 277-97.
Barro, Robert J. 1991. "Economic Growth in a Cross Section of Countries." Quar
terly Journal of Economics 106(2): 407-44.
Barro,
J., and Xavier
Robert
Sala-i-Martin.
1995.
Economic
Growth.
McGraw
Hill.
Bartelsman, Eric J., Randy A. Becker, and Wayne B. Gray. 2000. "NBER-CES
Industry Database." Cambridge, Mass.: National Bureau of
Manufacturing
Research.
Economic
Mimeographed.
1999. "Measuring Business Cycles:
Baxter, Marianne, and Robert G. King.
Approximate Band-Pass Filters for Economic Time Series." Review of Eco
nomics and Statistics 81(4): 575-93.
Ben-David,
-.
1996.
Dan.
International
"Trade
Economics
2001.
"Trade
and Convergence
279-98.
and
Liberalization
Journal of International Economics
Andrew
Bernard,
-.
Countries."
Journal
Income
A
Convergence:
Comment."
1995.
in International
"Convergence
10(2): 97-108.
Output." Journal of Applied Econometrics
1996. "Interpreting Tests of the Convergence Hypothesis."
Econometrics 71(1-2): 161-73.
Andrew
Bernard,
and Charles
B.,
Economic Journal
Caves,
Douglas
W.,
106(437):
Laurits
of
55(1): 229-34.
N. Durlauf.
and Steven
B.,
among
40(3-4):
I. Jones.
1996.
Journal of
and Convergence."
"Technology
1037-44.
R. Christensen,
and W.
Erwin
1982.
Diewert.
lateral Comparisons of Output, Input, and Productivity
Index Numbers." Economic Journal 92(365): 73-86.
Using
"Multi
Superlative
Ciccone, Antonio, and Robert E. Hall. 1996. "Productivity and theDensity of Eco
nomic Activity." American Economic Review 86(1): 54-70.
Dollar, David, and Aart Kraay. 2003. "Institutions, Trade, and Growth: Revisiting
the Evidence." Policy research working paper 3004. Washington: World Bank.
Doppelhofer,
Gernot,
Ronald
I. Mille,
and Xavier
Sala-i-Martin.
2000.
"Determi
nants of Long-Term Growth: A Bayesian Averaging of Classical Estimates
Approach (BACE)." Working paper 7750. Cambridge, Mass.: National Bureau
of Economic
Research.
William Easterly, Norbert Fiess, and Daniel Lederman
Easterly, William, and Ross Levine. 2001. "It's Not Factor Accumulation:
ized Facts and Growth Models." World Bank Economic Review
51
Styl
15(2):
177-220.
-.
2003.
nomic
Eaton,
Journal
Development."
Jonathan,
Theory
2002.
How
Crops:
of Monetary
1999.
"International
Economic
International
Influence
Eco
3-39.
50(1):
Technology
Review
40(3):
Diffusion:
537-70.
and Trade." Econometrica
Geography,
"Technology,
Endowments
Economics
Kortum.
and Samuel
and Measurement."
-.
and
Germs,
"Tropics,
70(5):
1741-80.
1999. "Convergencia regional en M?xico,
1940-1995."
Esquivel, Gerardo.
Trimestre Econ?mico 66(264): 725-61.
Esquivel, Gerardo, and others. 2002. "Why NAFTA Did Not Reach the South."
Washington: World Bank, Office of the Chief Economist for Latin America
and the Caribbean. Mimeographed.
Frankel, Jeffrey, and David Romer. 1999. "Does Trade Cause Growth?" Ameri
can Economic Review 89(3): 379-99.
Paul R. Krugman, and Anthony
Fujita, Masahisa,
Cities,
Economy,
and
Regions
International
J. Venables.
Trade.
MIT
1999. The Spatial
Press.
Furman, Jeffrey L., Michael E. Porter, and Scott Stern. 2002. "The Determinants
of National Innovative Capacity." Research Policy 31(6): 899-933.
Fuss, Catherine. 1999. "Convergence among Industrialized Countries: A Time
Series Investigation." Brussels: National Bank of Belgium.
Grossman,
Gene
and Elhanan
M.,
1991.
Helpman.
Innovation
and Growth
in the
Global Economy. MIT Press.
Hall, Robert, and Charles I. Jones. 1999. "Why Do Some Countries Produce So
Much More Output perWorker thanOthers?" Quarterly Journal of Economics
114(1): 83-116.
Harvey,
Andrew
of Economic
C.
2002.
"Trends,
edited
Growth,
Cycles
by Norman
In The Challenges
and Convergence."
and Raimundo
Soto.
Santiago:
Loayza
Central Bank of Chile.
Harvey,
Andrew
and V.
C,
mies." University
Harvey, Andrew C,
Series Models."
edited by C. Heij
Carvalho.
2002.
"Models
for Converging
Econo
of Cambridge. Mimeographed.
and Siam J. Koopman. 1997. "Multivariate Structural Time
In System Dynamics
in Economic and Financial Models,
and others, 269-98. Chichester: Wiley and Sons.
Filters for
Harvey, Andrew C, and T. Trimbur. 2001. "General Model-Based
Extracting Cycles and Trends in Economic Time Series." Discussion paper
0113. University of Cambridge, Department of Applied Economics.
Kaufman,
Daniel,
and Aart
cators for 2000-01."
Bank.
Kraay.
2002a.
"Governance
Matters
II: Updated
Indi
Policy research working paper 2772. Washington: World
52 ECONOM?A,Fall2003
-.
2002b.
"Growth
without
Governance."
World
Washington:
Bank
Insti
tute. Mimeographed.
2001. "International Technology Diffusion." Working
Keller, Wolfgang.
8573. Cambridge, Mass.: National Bureau of Economic Research.
-.
2002. "Geographic Localization of International Technology
American Economic Review 92(1): 120^42.
paper
Diffusion."
Krugman, Paul R. 1991. Geography and Trade. MIT Press.
Lederman, Daniel, Norman Loayza, and Rodrigo Soares. 2002. "Accountability
and Corruption: Political Institutions Matter." Washington: World Bank, Office
of the Chief Economist for Latin America and the Caribbean. Mimeographed.
Lederman,
-.
F. Maloney.
and William
Daniel,
2003a.
"Innovation
in Mexico:
NAFTA Is Not Enough." Washington: World Bank, Office of the Chief Econ
omist for Latin America and the Caribbean. Mimeographed.
2003b. "R&D and Development." Washington: World Bank, Office of
the Chief Economist for Latin America and the Caribbean. Mimeographed.
Lederman, Daniel, William F. Maloney, and Luis Serven. 2003. "Lessons from
NAFTA." Washington: World Bank. Mimeographed.
Lederman, Daniel, and others. 2003. "The Recovery of Mexican Investment after
the Tequila Crisis: Basic Economics, Confidence Effects or Market Imperfec
tions?" Journal of International Money and Finance 22(1): 131-51.
Pablo Fajnzylber,
Norman,
Loayza,
in Latin
Growth
America
and
2002.
and C?sar Calder?n.
the Caribbean:
Stylized
Facts,
"Economic
and
Explanations,
for Latin
Forecasts." Washington: World Bank, Office of the Chief Economist
and
America
the Caribbean.
J. Ernesto.
L?pez-C?rdova,
ductivity: An Empirical
Inter-American
Lucas,
1988.
"On
"NAFTA
Bank.
y
otras
Manufacturing
reformas
Pro
Data." Washington:
Mimeographed.
the Mechanics
ofMonetary Economics 22(1): 3-42.
Messmacher, Miguel. 2000. "Desigualdad
TLCAN
and Mexico's
Investigation Using Micro-Level
Development
M.
Robert
Mimeographed.
2002.
estructurales."
of Economic
regional en M?xico.
Research
Journal
Development."
working
El efecto del
paper
2000-4.
City: Bank of Mexico.
Meza, Liliana, and Ana Bel?n Mora. 2002. "Why Mexican Firms Invest inR&D."
Washington: World Bank, Office of the Chief Economist for Latin America
and the Caribbean. Mimeographed.
Mexico
Parente,
Stephen
L.,
and Edward
C. Prescott.
2000.
Barriers
to Riches.
MIT
Press.
Persson, Torsten. 2002. "Do Political Institutions Shape Economic Policy?"
Econometrica 70(3): 883-906.
Pritchett, Lant. 1997. "Divergence, Big Time." Journal of Economic Perspectives
11(3): 3-17.
William Easterly, Norbert Fiess, and Daniel Lederman
53
Redding, Stephen, and Anthony Venables. 2001. "Economic Geography and
International Inequality." Discussion paper 495. London School of Economics,
Center
Romer,
Performance.
for Economic
Paul.
1986.
"Increasing
-.
1990.
Returns
and Long-Run
Growth."
of Polit
Journal
1002-37.
ical Economy 94(5):
"Endogenous
Journal
Change."
Technological
Econ
of Political
omy 98(5): S71-102.
Sachs,
Jeffrey
and Andrew
D.,
Warner.
"Economic
1995.
Reform
and
the Process
of Global Integration." Brookings Papers on Economic Activity 7(1995): 1-95.
Sala-i-Martin, Xavier. 1996. "The Classical Approach to Convergence Analysis."
Economic Journal 106 (July): 1019-36.
Schiff, Maurice, and Yanling Wang. 2002. "Regional Integration and Trade
Related Technology Diffusion: The Case of NAFTA." Washington: World
Bank, Office of the Chief Economist for Latin America and the Caribbean.
Mimeographed.
Slaughter, Matthew
A
vergence:
J. 2001. "Trade Liberalization
Difference-in-Differences
Analysis."
and per Capita Income Con
Journal
of
International
55(1): 203-28.
Solow, Robert M. 1956. "A Contribution to the Theory of Economic
Quarterly Journal of Economics 70(1): 65-94.
Economics
Summers,
Robert,
and
Sultan
Ahmad.
1974.
"Better
Estimates
Growth."
of Dollar
Gross
Domestic Product for 101 Countries: Exchange Rate Bias Eliminated." Dis
cussion paper 297. University of Pennsylvania.
World Bank. 2003. World Development Indicators. Washington.