POLICY RESEARCH WORKING PAPER           3 064
Trade Policy, Trade Volumes,
and Plant-Level Productivity
in Colombian Manufacturing Industries
Ana M. Fernandes
The World Bank 
Development Research Group
Investment Climate
May 2003



I POLICY RESEARCH WORKING PAPER 3064
Abstract
Fernandes explores Colombian trade policy from                  and industry heterogeneity. This result is not driven by
1977-91, a period of substantial variation in protection       the endogeneity of protection nor by plant exit. The
across industries, to examine whether increased exposure        impact of tariff liberalization on productivity is stronger
to foreign competition generates plant-level productivity      for large plants and for plants in less competitive
gains. Using a large panel of manufacturing plants, she         industries. Qualitatively similar results are obtained when
finds a strong positive impact of tariff liberalization on      using effective rates of protection and import penetration
consistent productivity estimates, controlling for plant        ratios as measures of protection.
This paper-a product of Investment Climate, Development Research Group-is part of a larger effort in the group to
understand the links between trade and productivity. Copies of the paper are available free from the World Bank, 1818
H Street NW, Washington, DC 20433. Please contact Ana Fernandes, room MC3-363, telephone 202-473-3983, fax 202-
522-3518, email address afernandes@worldbank.org. Policy Research Working Papers are also posted on the Web at
http://econ.worldbank.org. May 2003. (42 pages)
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about
development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished The
papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Research Advisory Staff



Trade Policy, Trade Volumes and Plant-Level Productivity
in Colombian Manufacturing Industries
Ana M. Fernandes
The World Bank
JEL Classification Numbers: F13, D24, C14, 054.
Keywords: Simultaneity.and Production Functions, Trade Policy, Productivity, Colombian
Manufacturing, Endogeneity of Protection.
*I would like to thank Christopher Udry, Steve Berry, and Phil Levy for their invaluable guidance.
This paper has also benefited from the suggestions of Patrik Bayer, Garth Frazer, Pinelopi Goldberg,
David MacKenzie, Marc Melitz, Peter Schott, Christopher Timmins, and seminar participants at
Yale University, NEUDC 2001, UCLA, UC-San Diego, Georgetown U., Indiana U., U. Montreal,
Rutgers U., Syracuse U., U. Texas-Austin, the World Bank, and the 2003 Winter Meetings of the
Econometric Society in Washington, DC. I also thank the Colombian Statistical Institute (DANE)
for authorizing the use of the data and Mark Roberts for providing it. Financial support from Sub-
Programa Ciencia e Tecnologia do Segundo Quadro Comunitario de Apoio (Portugal) is gratefully
acknowledged. The findings expressed in this paper are those of the author and do not necessarily
represent the views of the World Bank. Please address correspondence to Ana M. Fernandes, The
World Bank, 1818 H Street, NW, Washington DC 20433. Phone: 202-473-3983. Fax: 202-522-3518.
E-mail: afernandesC0worldbank.org.






1. Introduction
This paper addresses a central question in trade and development economics: does in-
creased exposure to foreign competition generate gains in industrial productivity? Focusing
on the case of Colombia, we find that it does. A panel of manufacturing plants between
1977 and 1991 is used to investigate the link between trade policy and plant-level productiv-
ity. During these fifteen years, trade policy exhibited significant variation, with periods of
trade liberalization alternating with periods of increased trade protection. Moreover, these
changes in protection differed substantially across industries.
The empirical literature on the impact of trade liberalization on productivity provides
some, but not definitive, evidence of a positive effect of trade liberalization on productivity
levels and growth following three approaches. First; the macro-level approach utilizes cross-
country growth regressions, associating output growth with an aggregate measure of trade
openness.' Using measures of outward policy orientation across countries and over time is
plagued by difficulties (see, e.g., Pritchett, 1996; Rodrik and Rodriguez, 2001). Furthermore,
an aggregate measure of openness to trade cannot capture the differential incentives provided
by trade protection to different industries.
Second, the industry-level approach considers cross-industry regressions of Solow (1957)
residual total factor productivity (TFP) growth on trade policy variables or on demand
growth due to export expansion and import substitution.2 Having a single productivity
measure per industry, -however, ignores cross-plant heterogeneity, a stylized finding for de-
veloped and developing countries, which is useful to fully investigate the impact of trade
policy on productivity.
Finally, the micro-level approach uses regressions of either (i) firm output growth derived
in a Solow framework on an indicator variable for the period of trade reform, or (ii) plant
TFP measures on trade orientation in the plant's industry. Harrison (1994) and Krishna
and Mitra (1998) estimate the former for firms in Cote d'Ivoire and India, respectively.
However, the coefficient on the indicator variable for the trade reform period cannot isolate
the corresponding productivity gains since it also captures contemporaneous macroeconomic
shocks. Moreover, it ignores the variation in protection across industries. Pavcnik (2002)
1 See, for example, Dollar (1992), Edwards (1993, 1998), Harrison (1996), Levine and Renelt (1992), Sachs
and Warner (1995), and the references therein.
2 For example, Nishimizu and Robinson (1984) use export expansion and import substitution, while Kim
(2000) and Lee (1995) use trade policy variables. See Rodrik (1995) for a survey.



estimates the latter, identifying the effect of trade reform in Chile from the comparison of
consistent TFP measures of plants in import-competing and export-oriented, industries to
those of plants in nontraded industries over time as trade reform proceeds.3
Our study contributes to the micro-level approach by examining the impact of trade
policy on Colombian plant productivity following a two-stage estimation procedure. In the
first stage, we obtain new time-varying measures of plant productivity applying an esti-
mation method that addresses a fundamental simultaneity problem. The use of ordinary
least squares (OLS) for production function estimation assumes that regressors such as la-
bor are treated as exogenous variables. However, input choices are endogenous, depending
for example on managerial ability known to the plant's decision-maker, but unknown to the
researcher. Since input choices and productivity are correlated, OLS estimates suffer from a
simultaneity bias.4 Our methodology for production function estimation follows Levinsohn
and Petrin (2001). Under general conditions, a plant's demand for raw materials increases
monotonically with productivity, conditional on its capital stock. Hence, we use a nonpara-
metric estimate for the inverse raw materials demand function as the control for unobservable
productivity to correct for the simultaneity bias.
In the second stage, we estimate the link between trade policy and plant productivity in a
regression framework, relying on measures that exhibit significant variation across industries
and over time, rather than on a single change in trade regime as in previous studies. We
focus on nominal tariffs since they are direct price measures of trade barriers reflecting the
degree of government intervention and the changes in trade regime. Harrison (1996) and
Rodrik and Rodriguez (2001) argue that the use of tariffs, instead of indicator variables
for the liberalization period, is more useful to precisely identify the impact of trade policy
on productivity. However, trade policy (e.g., tariffs) is subject to a potential endogeneity
problem: the government may raise current trade policy barriers in response to lobbying
by firms in industries with relatively low productivity. We address this problem in our
estimation by (i) considering lagged trade policy measures, (ii) controlling for unobserved
3 Pavcnik also considers import penetration and an average tariff in a sub-period of her main sample.
However, since tariffs are uniform across Chilean industries, they are equivalent to year effects. Hence, her
analysis cannot exploit the variation in trade policy across industries, which is the most appealing feature
of the Colombian case.
4 This bias was initially discussed by Marschak and Andrews (1944). See Griliches and Mairesse (1995) for
a comprehensive survey. Variable inputs, which are easier to adjust, tend to have upwardly biased OLS
coefficient estimates. But with more than two inputs, not all biases can be exactly signed, as they depend
upon the degree of correlation between each input and the productivity shock.
2



industry fixed characteristics, and (iii) examining the economic rationales underlying trade
policy determination in Colombia.
The main findings of this paper are as follows. First, our production function estimates
reveal the importance of correcting for the endogeneity of input choices with respect to pro-
ductivity across industries. Second, we provide strong evidence supporting the hypothesis
that Colombian plants' productivity is negatively affected by trade protection. Lagged nom-
inal tariffs have an economically and statistically significant negative impact on productivity,
even after controlling for, plant characteristics, for industry heterogeneity, and for variation
in the real exchange rate (RER). Changes in Colombian tariffs during the sample period are
a response to fiscal and external imbalances, which suggests that the negative impact of tar-
iffs does not reflect causality running from productivity to trade policy. Third, the negative
impact of tariffs is not driven by the exit of less productive plants under trade liberalization,
rather it reflects within-plant productivity gains. We provide evidence suggesting that these
gains are associated with an increase in (i) skilled labor intensity of production, (ii) imports
of intermediate inputs, and (iii) investments in machinery at the plant level. Fourth, we
allow for heterogeneity in the impact of trade policy on productivity according to plant size
and find that the negative impact of tariffs on productivity is stronger for larger plants. To
the best of our knowledge, this finding has not been previously provided in the literature.
Finally, we introduce cross-industry heterogeneity in the impact of trade policy on produc-
tivity and estimate a stronger negative impact of tariffs for plants in industries with a lower
degree of domestic competition. Our main findings are robust to the use of effective rates of
protection (ERP) and import penetration ratios as measures of trade protection.
Trade liberalization may affect plant productivity through several mechanisms. First,
as imports expand, the ensuing competitive pressure results in higher productivity if do-
mestic firms eliminate X-inefficiency or slack and use inputs more efficiently.5 In contrast,
infant-industry arguments sustain that selected protection allows for productivity gains in
industries where learning-by-doing is important. Second, trade liberalization may boost
plant productivity by allowing for an increased access to imported intermediate inputs of
5 Vousden and Campbell (1994) examine the efficiency of a firm with internal informational asymmetries and
show that trade protection induces slack, by reducing competition. Extending the technology ladder model
in Grossman and Helpman (1991b), Holmes and Schmitz (2001) show that for an entrepreneur, lowering
the tariffs protecting his/her industry makes it less attractive to engage in nonproductive activities (wast-
ing efforts blocking competitors' potential innovations) and more attractive to pursue productive activities
(engaging in research).
3



higher quality and/or broader variety (e.g., Grossman and Helpman, 1991a), through export
activities, the exposure to technologies embodied in imported final goods and the access
to imported capital goods corresponding to technologies that were previously unavailable.
Third, trade liberalization may influence the incentives to invest in technological innovation.
In fact, Colombian government agencies attribute weak industrial productivity in the early
1980s to existing trade protection mechanisms that reduced those incentives (Zerda, 1992).
Goh (2000) focuses on the opportunity cost of technological effort (the foregone profits due
to the delay in the commercialization of output) and shows that trade liberalization in-
creases a firm's incentives to engage in productivity-enhancing technological effort since it
reduces its profits for any productivity level. In contrast, Rodrik (1991) and Traga (1997)
find that lower trade protection or higher import competition reduce a firm's investments
in productivity-enhancing technological upgrading, when the incentives to invest depend on
the firm's output or market share reduced by trade liberalization. In sum, the theoreti-
cal literature delivers some disparate predictions: trade liberalization may result in either
productivity gains or losses.6 Hence, empirical evidence is essential to inform the debate.
This paper is organized as follows. In Section 2, we present the empirical methodology.
In Section 3, we describe the data. In Section 4, the production function and productivity
estimates are presented. In Section 5, we estimate the impact of trade policy on plant
productivity. In Sections 6 and 7, we examine the differential impact of trade policy on
productivity according to plant size and to the degree of domestic competition in the industry,
respectively. Section 8 concludes.
2. Empirical Methodology
We estimate consistent production function estimates combining parametric and non-
parametric techniques as in Levinsohn and Petrin (2001). This methodology builds upon
that proposed by Olley and Pakes (1996), where investment corrects for the simultaneity
bias resulting from the correlation between plants' input choices and privately known pro-
ductivity, but is more suitable for developing countries' datasets. A technical condition for
Olley and Pakes' methodology - investment is a monotonic function of productivity con-
ditional on capital - is verified only for plants with positive investment (Pakes, 1994). In
6 The exploitation of scale economies is another mechanism by which trade liberalization could lead to
productivity gains. In our framework, we capture intra-plant gains unrelated to scale economies since these
are embodied in the production function coefficients and productivity estimates.
4



Colombia, small -plants invest infrequently, so they would be systematically eliminated from
the estimating sample. Using investment to correct for simultaneity would in fact result in
biased production function estimates and possibly unrealistic annual variability in estimated
plant productivity. In our estimation procedure, we use raw materials to control for the si-
multaneity bias.7 Our dataset allows us to precisely measure raw materials u-sage (hereafter
simply 'raw materials') that is easily adjustable to productivity shocks.'
We assume that a plant decision-maker maximizes expected profits from a Cobb-Douglas
production function under uncertainty. Plant-level heterogeneity is allowed in the form of
plant specific productivity shocks. The timing of decisions of plant i in industry j in year t is
as follows. The decision-maker initially observes current productivity w tt and then chooses
variable inputs labor Q ., raw materials mij and energy e i to be combined with the quasi-
fixed input capital kit for production of output yj . We assume that plants' decision-makers
decide whether or not to exit the industry before observing w,. The unbalanced nature of
our panel dataset (described in Section 3) controls in part for the selection bias that arises
if this assumption does not exactly hold in our sample. The estimating equation for plant i
in industry j in year t is as follows (with output and inputs in logarithms):9
Yi= a + f311l + 3ee 3 + Pmmmit + Okki3k  + w,i + eJ,                (1)
where w i, privately known to the decision-maker and correlated with1 3', e,3, and mj3 gener-
ates the simultaneity bias and iti, unknown to the decision-maker, represents unpredictable
mean-zero shocks to productivity realized after input choices are made. A set of production
function parameters is obtained for each industry j to allow for technological differences
across industries.
The plant's variable input demands derived from profit maximization depend on cap-
7 The simultaneity bias could also be addressed by fixed plant effects estimation. However, doing so would
impose the particularly restrictive assumption of no dynamics in plant productivity. Alternatively, instru-
mental variables (IV) estimation would correct for simultaneity but would require the use of plant-level
instruments correlated with input choices and uncorrelated with productivity, such as input prices, which
are not reported for Colombian plants. Lagged inputs cannot serve as instruments, since our framework
allows for serial correlation in plant productivity.
8 Raw materials usage is measured as purchases of raw materials plus the net change in inventories. By
accounting for changes in inventories, this measure captures the current demand for raw materials that is
undoubtedly correlated with current productivity. Furthermore, most plants report positive raw materials
usage in all years (a requirement to belong to the estimating sample), whereas that is less prevalent for other
intermediate inputs.
9 We choose an output over a value-added specification since the separability of the production technology
in intermediate inputs, required for value-added to be a valid production index, is not verified for our data.
Within industries, the ratio of intermediate inputs to output exhibits high variability.
5



ital and on privately known productivity. We invert the raw materials demand function
m  =mt (w , kiJ) to obtain a productivity function imposing the following monotonicity as-
sumption: conditional on capital, the demand for raw materials increases with productivity.10
The productivity function wit =       i/t(mj, kij) depends only on observable variables. Eq. (1)
can be rewritten in the partially linear form (hereafter omitting industry superscript j):
Yit =   'llit + 03eeit + (t (mit, kit) + cit,                (2)
where
(t (mit, kit) = a + /3mmit + fkkit + Wt (mit, kit).
Since E[eit I mit, kit] = 0, the difference between Eq. (2) and its expectation, conditional on
raw materials and capital is given by:
yit - Et yit I mit, kit] = 81(lit - E[ lit mit, kit]) + 8,(eit - Et eit I mit, kit ]) + eit.  (3)
Eq. (3) is estimated by OLS (with no constant term) to obtain consistent parameter esti-
mates for the variable inputs that do not correct for simultaneity, labor and energy. The
conditional expectations are obtained by locally weighted least squares (LWLS) regressions
of output, labor and energy on (rnit, kit) .
Cycles in Colombian manufacturing output growth affect input demands through, e.g.,
variation in the ratios of input to output prices. While the raw materials demand function
does not explicitly depend on plant-level input and output prices, it is allowed to differ across
two periods: (i) 1977-1983, a period of slow output growth, and (ii) 1984-1991, a period of
faster output growth."2 Consequently, the productivity function wt(.) and the function t (.)
also differ across two periods. In particular, qt (.) is obtained from a LWLS regression of (yit
- 3llit - /eeit) on (mit, kit) estimated separately across the two periods.
To estimate (0m,13k) consistently, we assume that productivity follows a first order
Markov process as in, e.g., Hopenhayn and Rogerson (1993) and Olley and Pakes (1996):
'OA sufficient (but not necessary) condition for this monotonicity assumption to hold is perfect competition
in input and output markets. The estimation is also valid under some types of imperfect competition in
output markets (e.g., Cournot oligopoly with linear demand functions (Levinsohn and Petrin, 2001)). Note
that the raw materials demand function is industry-specific, not plant-specific.
IlFor example in the case of output, we estimate a weighted linear regression of Yit on a second order
polynomial in (mit, kit) using data in the neighborhood of a data point (mit, kit). The intercept from this
regression is an estimate for the expected value of Yit conditional on (miit, kit). See Fernandes (2002) for
further details. .
l2For brevity, a t subscript indexes mt(.), wt(.), Ot(.). However, these functions are allowed to differ only
across the two periods, not across years. Also, note that the break years for the productivity function wt(.)
differ from the break years for the trade regimes later described in Section 3.2.
6



Wit = E[wit I Wit-,] + (it, where (it, the unexpected productivity shock, is independent and
identically distributed. Our estimation strategy is based on the identification assumption
that capital may adjust to expected productivity but does not adjust to the unexpected pro-
ductivity shock. The following moment conditions are obtained by taking the expectation
of Eq. (1) conditional on, respectively, capital and lagged raw materials, and replacing wit
by its Markov process:
E[yit - 0I3lit -  eit - 0 m.Mit - fkkit - E[wit I Wit-,] I kit]=E[eit + (it I kit] = 0,  (4)
E[yit - #,lit - feit - smMit - Okkit - E[wit I wit-l] I mit-,]=E[eit + (it I mit-1] = 0.  (5)
Since E[cit I kit] = 0, Eq. (4) indicates that capital in year t is uncorrelated with the
unexpected productivity shock in year t. Given that E[eit I mit-1] = 0, Eq. (5) indicates
that raw materials in year t - 1 is uncorrelated with the unexpected productivity shock in
year t.03 The residuals +it  6t are obtained using the estimated coefficients (:1X 13e)) some
initial values (,8*, 0k), and a non-parametric estimate for expected productivity E[wit
witA. 4 A generalized method of moments (GMM) criterion function weights the plant-year
moment conditions by their variance-covariance matrix. The estimation algorithm starts
from initial OLS estimates, iterates on the sample moment conditions to match them to
their theoretical value of zero, and then reaches final parameter estimates."5 The standard
errors for the parameter estimates are bootstrapped.16
The plant-level Hicks-neutral TFP residual is defined as PTit = Wit + eit and represents
the efficiency in transforming inputs into output, through learning-by-doing, adopting newer
and better methods of production, improvements in managerial practices, worker training,
among others. It may incorporate unobservable changes in factor utilization, since costs
rise when plants operate below capacity. In fact, the use of raw materials to correct for
simultaneity parallels its use to correct for unobserved variation in factor utilization (e.g.,
13Eq. (5) identifies the parameter /3 since mi*t- and mit are correlated under the Markov structure.
14This estimate is obtained as a LWVLS regression of (wit + eit) *  it - 6 -it - /eeit -  -*mit - Pkkit on
Witl=  (mit-1, kit-,) - /3mmit-I - 6*kit-l Note that the parameter Ca cannot be separately identified
from the estimated E[wit I wit-ij.
15A derivative optimization routme is complemented by a grid search, given the existence of multiple local
minima for the GMM criterion function in some industries. The minimizers resulting from grid search are
used as starting values in the derivative optimization routine to reach more precise (Pm, 63k) values.
16For any industry, our bootstrap procedure consists of sampling randomly with replacement plants from the
dataset, matching in any year the number of plants in the original sample. Each plant is taken as a block if
it is randomly selected (i.e., all its observations are included in the bootstrap sample), since the estimation
requires the use of lagged inputs. We obtain estimates (/3,/3e,/3m,/3k) for 100 different bootstrap samples.
The standard deviation of a parameter estimate across bootstrap samples constitutes its standard error.
'7



Basu, 1996). Considering this broader concept of productivity is useful to examine whether
plant productivity is affected by trade policy. Using the consistent production function
coefficients, plant TFP is estimated by Arit = Yit - Bljt - I3eeit - /3mmit - ,Bkklt. This TFP
measure is associated with a particular industry's technology, so it is not comparable across
industries. We follow Aw, Chen and Roberts (2001) and Pavcnik (2002) to obtain a relative
TFP measure, comparable across years and industries. For each plant in an industry, relative
TFP is the difference between the plant's estimated TFP and the TFP of an average plant in
the industry in 1977 (obtained combining average logarithmic output and inputs in 1977 with
the estimated production coefficients). Hereafter, prit denotes the relative TFP measure.
Our analysis of the impact of trade policy on plant productivity is based on the following
specification that pools plants across all industries and years:
prit = 10 +-At + 31TP t3L1 + (32)'Xitl + I   +u it               (6)
where At is an indicator variable for sample years, TPti'1 is a measure of lagged trade protec-
tion, Xj is a matrix of plant characteristics affecting productivity, I i is an indicator variable
for the plant's 3 or 4-digit industry and u12 represents other omitted factors influencing pro-
ductivity. Part of the heterogeneity in plant productivity is deliberately left unexplained in
Eq. (6), as our interest lies in a specific factor affecting productivity: trade protection. Note
that ,i3 is negative if trade liberalization is associated with an increase in plant productivity.
3. Data: Production and Trade Policy
3.1. Production Variables
This paper uses plant-level panel data from the Colombian Manufacturing census pro-
vided by DANE (National Statistical Institute) for the years 1977 through 1991. Plants
with less than 10 employees are: (i) included in the census in 1977-1982, (ii) excluded in
1983-1984, and (iii) included as a small proportion after 1985.17 The census covers exten-
sively formal industrial production in all 3-digit ISIC (revision 2) industries in Colombia.
'7The census does not indicate whether a plant is a single-plant firm or belongs to a multi-plant firm. It
is unlikely, however, that this biases the estimated effect of trade liberalization on productivity. If, for cost
efficiency purposes under liberalization, firms replace less productive plants by more productive new plants
keeping the same identification number in the census, then a positive effect of liberalization on productivity
would be largely due to the exit of less productive plants. This is not a concern for our study since DANE
registers any new plant with a new identification number. Also, if plants belonging to industrial groups have
better access to domestic and especially to imported inputs, they may derive less productivity gains from
liberalization, but that does not bias the estimated effect of liberalization.
8



The unbalanced nature of the panel allows the identification of entering and exiting plants
each year. For each plant and year, the survey collects data on production and sales rev-
enues, value added, input use (labor categories, raw materials, electricity and fuels which
are aggregated into an energy input), inventories of output and raw materials, investments
(buildings, machinery, transportation, office, land), exports (1981-1991), 3 and 4-digit ISIC
industry code and year of start-up operations."8 Some plants in the original sample (102,911
observations) are eliminated due to, e.g., incomplete series, zero values for output or inputs,
clear reporting errors, or an ambiguous industrial classification. The final sample has 97,107
observations corresponding to an average of 6,474 plants per year.
A large degree of plant heterogeneity is found in size, location, age, output and inputs.
Standard deviations of output and inputs across plants are more than twice the size of means
in all industries. The distribution of plant size is relatively stable over time, with plants
with less than 50 employees representing more than 70% of manufacturing in any year. The
median plant age increases from 10 years in 1977 to 14 years in 1991. The major industries
are food, apparel, textiles, printing, nonmetallic minerals and metal products. While the
distribution of plants across industries is relatively stable over time, a large number of plants
enter into and exit from the various industries. Average annual entry and exit rates in our
sample are 11.4% and 9.8%, respectively. Nevertheless, entrant and exiting plants represent
a small percentage of total output and are much smaller than incumbents.
3.2. Trade Regimes in Colombia
From 1977 to 1991, Colombian trade policy underwent significant swings, making it an
interesting case to identify the adjustment of plant productivity to changes in protection.'9
Three trade regimes can be clearly identified:
(i) First liberalization period (1977-1981): the government liberalized import barriers by
lowering tariffs and increasing the proportion of items under the free import regime, as a
response to exchange rate pressures from an increase in world coffee prices, high foreign
'8Capital stock measures are constructed by the perpetual inventory method for each plant and the five
types of capital. Nominal variables in current pesos are converted to 1980 pesos by the corresponding price
deflator. Specific price indexes from the Colombian Central Bank are used to deflate the different types
of capital and intermediate inputs. We thank Mark Roberts at Pennsylvania State Univ. for providing
output price indexes at the 3digit ISIC level used to deflate plants' nominal sales (adjusted for changes in
inventories) and generate a measure of output. This procedure has some limitations (Klette and Griliches,
1996) but has been widely used in previous studies in the absence of plant-specific price data.
I9See Garay (1991), Garcia (1991), GATT (1990) and World Bank (1984, 1989, 1991).
9



borrowing and illegal drug trade.
(ii) Protection period (1982-1984): trade restrictions were significantly increased. All tariffs
increased three times, most items were transferred to the prior-licensing list and some imports
were prohibited. A strong real exchange appreciation that hurt producers in traded sectors
and a world recession motivated this increase in protection.
(iii) Second liberalization period (1985-1991): a gradual shift to trade liberalization occurred.
Initially, the major goal was the administrative rationalization of the structure controlling
imports. Subsequently, significant reductions in tariff rates and dispersion were pursued and
unrestricted tariff items represented an increasing fraction of total imports.
During the entire sample period, protection was characterized by a large dispersion in
tariff levels and a cascading structure: lower tariffs on raw materials and intermediate inputs
not produced domestically (e.g., industrial chemicals and nonferrous metals) and higher
tariffs on consumer and finished products produced domestically (e.g., apparel and furniture).
Also, an import licensing system was in place, whereby each item in the tariff code was
classified into a free-import, a prior-licensing, or a prohibited import list.
3.3. Measures of Trade Protection
A challenge to an examination of the link between trade protection and productivity is
that trade protection cannot be described by a single measure. We consider three different
measures of trade protection.20 Tariffs and ERP reflect the degree of government intervention
and protection at the industry-level. Volume measures such as import penetration and
export orientation reflect how important foreign consumers and producers are to domestic
producers. In the case of Colombia, the different measures are consistent in indicating the
relative openness of industries and the evolution of protection over time. For example, high
tariffs and ERP are associated with low import penetration into 3 and 4-digit industries.
Also, the levels and changes of 3-digit tariffs and ERP are highly correlated.
Table 1 shows the average nominal tariffs for 3-digit industries across the three trade
2OTariff levels at 3 and 4-digit ISIC levels were obtained from Jorge Garcia at the World Bank and from
DNP (National Planning Department). In two of the common years across these sources, the value of tariffs
differs, but the differences are negligible, except for printing and transport equipment. ERP at the 3-digit
ISIC level were obtained from DNP calculated according to the Corden (1966) formula (i.e., from the tariff
on the final good a weighted average of tariffs on inputs is subtracted where the weights are taken from
an input-output matrix for Andean countries in 1982). Imports and exports at 3 and 4-digit levels were
obtained from Jorge Garcia. Coverage of domestic production by import licenses was obtained from the
World Bank. The correlations across measures are presented in Fernandes (2002).
10



regimes. For most industries, a sharp increase in tariffs from the first liberalization period
to the protection period is followed by a decline in the second liberalization period. Broad
trends in trade orientation across industries are presented in Table 2. For most industries,
import penetration decreases between 1980 and 1985 and increases between 1985 and 1991
(though less than expected given a strong depreciation in 1985-1988). Export orientation
declines between 1980 and 1984, under increased protection and RER appreciation, then
increases between 1984 and 1991.21
4. Production Function and Productivity Results
The results from the nonparametric/GMM production function estimation are presented
in Table 3 and Figure 1.22 Observe that our parameter estimates are reasonable relative to
previous studies. The variable inputs' coefficients are precisely estimated at the 1% con-
fidence level in most industries, even though bootstrapped standard errors are very high
when compared to those obtained using OLS. The use of estimated regressors at different
stages of the procedure increases the final coefficients' variability. The coefficient on capital
is significant in the major industries. Also, we cannot reject the hypothesis that nonpara-
metric/GMM estimates exhibit constant returns to scale in most industries.
If more labor is hired and more energy is consumed in periods of high productivity, OLS
estimates for variable inputs' coefficients are upwardly biased. Figure 1 depicts the relation-
ship between OLS and nonparametric/GMM coefficients for all inputs and industries and
the 45 degree line. Figures la, b and c indicate that in most industries, the OLS coefficients
on labor, energy and raw materials are upwardly biased relative to those from nonparamet-
ric/GMM estimation. If capital is correlated with lagged or expected productivity, then an
2lLicenses limiting the imports of items across tariff lines are another important measure of trade protection
in Colombia. They would ideally be measured by tariff or price equivalents. Unfortunately, only data for
coverage ratios of domestic production is available for a single year, indicating the percentage of domestic
production for which competing imports are subject to licensing restrictions. But these ratios do not show
which licenses are truly binding and which are issued automatically. In 1989, coverage ratios, tariff and ERP
levels are highly correlated. Tariffs place a minimum bound on the protection of items for which licenses are
the binding constraint. Also, Colombian tariffs are higher for the items also subject to import licenses (World
Bank, 1989, 1991). Given this positive correlation, it is possible that the impact of tariffs on productivity in
Section 5 is overestimated, as it picks up also the impact of the omitted licenses.
22The production functions are estimated at a slightly modified 3-digit industry level. In the original sample,
less than 1% of plants belong to different 3-digit industries across years. All of a plant's observations enter
estimation for one industry. So, we reclassify plants into the industry to which they belong in a majority of
years and eliminate the few plants for which no majority industry is found. Food plus food-miscellaneous,
textiles plus apparel, wood products plus furniture are considered as three industries for estimation, since
many plants belong an equal number of years to the two industries in the pair. Moreover, the technologies
of the two industries in those pairs are sufficiently similar.



upward bias in its OLS coefficient is possible. However, if capital is uncorrelated with ex-
pected productivity but is positively correlated with variable inputs, then a negative bias in
the OLS capital coefficient could result. In fact, the correlations between capital and labor,
materials and energy are positive and significant for all industries. Figure ld shows that the
OLS coefficient on capital is higher than that from-nonparametric/GMM estimation in half
of the industries.
As a robustness check, we use Olley and Pakes' techniques with raw materials, rather than
investment, correcting for simultaneity to obtain production function coefficients that do not
differ much from those in Table 3 (see Fernandes, 2002). Also, the trade and productivity
literature has often relied upon Solow TFP residuals, assuming that the contribution of
an input to output is equal to its share in total revenue. So, it is interesting to compare
our coefficients to average and median input revenue shares in Table 3. Labor and energy
revenue shares are lower than the corresponding coefficients in most industries, whereas the
opposite is true for capital revenue shares. Raw materials' revenue shares are lower than the
coefficients in half of the industries. Overall, using Solow residual TFP measures could bias
the estimated link between trade policy and plant produictivity.21
For each plant, we obtain the relative TFP measure prit and privately known productivity
Sit= 4t (mit, kit) -  kkit -  ..mmit. The correlation between these measures is positive and
significant, with coefficients ranging from 0.61 for ceramics to 0.93 for petroleum derivatives.
A decomposition of the variance of TFP within industries indicates that the main source of
cross-plant variation in TFP is variation in privately known productivity.
Across industries, TFP levels (obtained as output-share weighted sums of plant-level
productivities) and growth rates are generally procyclical relative to manufacturing output
growth and exhibit significant heterogeneity. Also, there is evidence of intra-industry het-
erogeneity: at any point in time, some plants' TFP evolves differently from industry TFP,
which stresses the importance of using disaggregated data for an accurate analysis of trade
and productivity.
23We also compare our coefficients to those obtained from plant fixed effects estimation. Fixed effects
coefficients on labor, energy and raw materials are smaller than nonparametric/GMM coefficients in most
industries, whereas the coefficients on capital are larger than nonparametric/GMM coefficients in half the
industries. These results are expected, since downward biases due to measurement error in inputs are
exacerbated with estimates obtained from within-plant variation in output and inputs. As a final robustness
check, we use electricity to correct for simultaneity in the nonparametric/GMM estimation and obtain
coefficients relatively close in magnitude to those in Table 3.
12



5. Productivity and Trade Policy
5.1. Average Impact of Tariffs
In this section we examine the effect of tariffs on plant productivity. The results from
estimating Eq. (6) are presented in Table 4 for OLS estimation with robust standard er-
rors (White correction for heteroskedasticity) clustered by plant and for plant fixed effects
estimation.24 We include plant age in Xi,, allowing it to have a non-linear effect on pro-
ductivity (see Campbell, 1998; Jensen, McGuckin and Stiroh, 2001; Power, 1998). Through
year effects, we control for shocks (e.g., macroeconomic cycles) affecting equally plant pro-
ductivity in all industries. The coefficient of interest 61 is negative and precisely estimated
at the 1% confidence level in columns (2)-(3) and (7)-(8), where unobserved fixed industry
heterogeneity is controlled for.25 Note that these columns provide estimates of the impact
of tariffs on plant productivity within industries over time. We also find a negative and sig-
nificant impact of tariffs on productivity in columns (4)-(5) and (9)-(10), where we account
for unobserved persistent characteristics causing serial correlation in plants' error terms.
Nominal tariffs are measured in fractional terms, so their reduction by one percentage point
increases productivity by /1%. For example, the coefficient in column (7) implies that a
10% reduction in 4-digit tariffs would lead to an increase in plant productivity of almost 3%.
Overall, the results support the hypothesis that lowering protection from foreign competition
generates productivity gains, even after controlling for unobservable industry or plant-level
heterogeneity.
The coefficient on tariffs in Eq. (6) is unbiased, unless some omitted factor influencing
productivity is correlated with tariffs. A difficulty that may arise in our framework is the
possibility of endogenous trade policy with respect to plant productivity, especially when
considering current (rather than lagged) trade policy. A biased estimate of /1 would be
obtained if (i) the government increased trade protection in response to lobbying by firms
in industries with lower productivity, or (ii) it adjusted protection to reflect industries'
productivity relative to other domestic or foreign industries. Changes in tariff could be taken
241n Eqs. (6)-(9), when trade protection TP is measured by tariffs, time subscripts require a careful inter-
pretation. Suppose t-1, t, 7- 1, r are sample years in chronological order. pri, is affected by tariffs TPt-1,
where t and t -1 are consecutive sample years. But for the next pair considered pr,i, and TP41, r -1 may
be strictly larger than t if tariff data at t is unavailable (though the same one year lag separates t - 1 from t
and r - 1 from r): e.g., TFP in 1981 is affected by tariffs in 1980, but the following pair considered is TFP
in 1984 affected by tariffs in 1983, since tariff data is unavailable in 1981-1982.
25Sample sizes differ in columns (1)-(5) and (6)-(10) since tariff data for some 4-digit industries is not
disclosed.
13



as exogenous if they resulted from GATT negotiations. But although Colombia became a
GATT member in 1981, it did not participate in trade negotiations before the Uruguay Round
(1986-1994), so tariff changes during our sample period are unrelated to GATT regulations.
We argue that during our sample period, trade policy determination in Colombia was
such that its endogeneity with respect to productivity is not a serious concern. The changes
in tariffs were cyclical and a policy response by the government to macroeconomic shocks for
short-run stabilization purposes. Import barriers were alternatively loosened or tightened
to smooth out aggregate expenditure in response to coffee booms in the first liberalization
period or to external payment deficits in the protection period (Hallberg and Takacs, 1992).
While the first liberalization period was "stimulated by a desire to control money supply and
inflation without an export-destroying revaluation" (Urrutia, 1994, p. 297), the protection
period responded to escalating fiscal and current account deficits. Indeed, tariff revenue
represented an important fraction of government tax revenue in Colombia - more than 16%
in 1981-1986 (World Bank, 1989).
Furthermore, across any pair of years, tariffs moved uniformly in an upward or downward
direction. Although the magnitude of changes in tariffs differed significantly across industries,
they did not result from the government asymmetrically changing the tariffs protecting less
productive industries in response to pressures. Rather, the differential changes resulted from
the government's interest in changing more strongly the tariffs protecting goods "whose
demand was relatively more elastic to price movements" (Garay, 1991, p. 19), so that
imports increased or declined more rapidly.26
For most of the sample period, there is no evidence that trade policy was adjusted
by the government to reflect industries' relative productivities. Trade liberalization was
not considered as a channel to accelerate growth or to improve the allocation of resources
(Garcia, 1991). In fact, trade liberalization was presented by DNP as a change needed for the
economy to "start achieving greater productivity growth and efficiency" (Urrutia, 1994, pp.
304-305) only in 1990. Moreover, industrialists only began to realize that (i) the domestic
market was not a dynamic source of growth, (ii) trade protection had high costs and, (iii)
exporters became a strong pressure group only by the end of the 1980s.
26Our discussion does not imply the absence of political economy pressures during the sample period. Colom-
bian producers expected government protection from foreign competition since the 1950s' import substitution
industrialization, but the pressures were not specific to less productive industries, but rather were widespread
across industries. It is possible that these pressures operated through nontariff barriers and not through tar-
iffs, since the approval processes in import licensing regimes are more highly subject to discretion.
14



While we believe that the endogeneity of trade policy with respect to productivity is not
pronounced for Colombian industries, we address it as follows. We estimate the effect of
lagged rather than contemporaneous tariffs on plant productivity.27 In addition, we control
for unobserved fixed industry characteristics that influence simultaneously productivity and
lagged tariffs with industry fixed effects. In fact, the positive sign of the tariff coefficients
in columns (1) and (6) is reversed once those fixed effects are controlled for. Goldberg
and Pavcnik (2001) argue that industry fixed effects account for time-invariant political
economy factors underlying higher or lower trade protection across industries. Finally, since
it is plausible that unobserved time-varying industry characteristics influence simultaneously
productivity and tariffs, we instrument for tariffs with the same industry characteristics used
by Trefler (1993) in explaining U.S. nontariff barriers.28 These characteristics, such as the
share of unskilled labor in total employment, reflect industries' propensity to get organized
and political economy determinants of protection. The instrumented coefficients on lagged
3 and 4-digit tariffs remain negative and significant and increase four times in magnitude,
relative to OLS coefficients.29 Nevertheless, these IV estimates should be viewed with caution
since they suffer from serious caveats.30
As discussed earlier, OLS production function estimates are biased due to the correlation
between inputs and productivity. Such bias is transmitted to the corresponding residual
productivity measures. Our main concern is that this would be reflected in the estimated
impact of tariffs on productivity. In fact, although the results are qualitatively unchanged,
27This also addresses Tybout's (1992) concern that uncertainty about the sustainability of changes in trade
policy delays the ensuing changes in productivity, a relevant concern for Colombia, given the frequent changes
in trade regime. For GATT member countries, another source of trade policy uncertainty is the freedom of
authorities to vary tariff levels above or below bound levels. This is a minor concern for Colombia since only
36 items were bound upon GATT membership representing a small percentage of imports (GATT, 1990).
However, the dynamics may be more complex than one-period lagged tariffs affecting plant productivity.
28Ideally, a dynamic model with simultaneous determination of protection and productivity should be used
to derive instruments for time-varying cross-sectional patterns of protection, since most political economy
models such as Grossman and Helpman (1994) predict only cross-sectional patterns of protection. Developing
such a model, however, is beyond the scope of this paper.
29WV estimation covers the period 1981-1989 only, given some restrictions in the availability of data for the
instruments. The corresponding OLS estimates of Eq. (6) are higher than those in Table 4. The first-stage
regressions for 3-digit tariffs indicate that these depend negatively on Herfindahl indexes and on capital and
positively on total employment, on a proxy for minimum efficient scale, on the share of unskilled labor in
total employment and on output growth. The first-stage regressions for 4-digit tariffs indicate that these
depend negatively on capital and on output growth and positively on total employment and on the share of
unskilled labor in total employment. The results for the first-stage are available upon request and include
further instruments not mentioned here due to their lack of significance in explaining tariffs.
3OSome instruments exhibit little variation over time and hence cannot account for the variability in trade
policy. More crucially, some instruments are correlated with productivity, as it will be shown in Section 7.
15



the actual impact of tariffs on OLS productivity is overestimated relative to the unbiased
impacts in Table 4.
5.3. Robustness of the Impact of Tariffs
We now investigate the robustness of our main findings along several dimensions. First,
the productivity gains associated with tariff declines could reflect the exit of less productive
plants with no change in the remaining plants' productivity levels. Reduced trade protection
and the consequent decline in output prices may push previously profitable low productivity
producers out of business, if exit barriers are not too high. We use four approaches to
examine whether plant exit affects our results. The first approach consists of decomposing
changes in industry productivity into: (i) changes in continuing plants' productivity, (ii)
reallocations of output among continuing plants with different productivity levels, and (iii)
a term representing differences in productivity between cohorts of entrant and exiting plants
(see the Appendix). The major sources of change in industry productivity are (i) and (ii)
while (iii) contributes little to the change, particularly the lower productivity of exiting
plants. The second approach involves estimating Eq. (6) only for plants that remain in
operation until 1989. or after. The negative impact of tariffs is maintained. The third
approach uses probit regressions to assess how exit probabilities vary with trade policy.
Exit probabilities increase with tariffs, controlling for plant productivity and year effects.
However, when a control for time-invariant differences in exit barriers across industries is
added, exit probabilities decrease with tariffs. As an alternative to tariffs, we use trade
regimes. The results show that exit probabilities decrease in the protection and the second
liberalization periods relative to the first liberalization period, with or without industry
effects. The fourth approach considers average exit rates out of industries with relatively
stronger tariff reductions and shows that they are not systematically higher than exit rates
out of industries with weaker tariff reductions. In sum, there is evidence that less productive
plants exit, but higher exit is not linked to liberalization. So, the gains associated with lower
tariffs in Table 4 reflect largely within-plant changes in productivity.
Second, RER variation could confound the impact of trade protection on productivity.
Year effects account for macroeconomic changes, but the RER may affect plants differently
depending on their industry's trade orientation (Levinsohn, 1999). In Colombia, during
the first liberalization period the RER appreciates, whereas it depreciates during the sec-
16



ond liberalization period." A RER depreciation increases the demand for and profitability
of traded industries' output.32 In the short-run, an increase in measured productivity oc-
curs when plants respond by exploiting unobserved unused capacity, before adjusting input
choices. If such a depreciation accompanies trade liberalization, then the estimated pro-
ductivity gains (in traded industries) could reflect this capacity adjustment. Using a broad
concept of productivity, an increase in capacity utilization is indistinguishable from an in-
crease in productivity, so this RER effect, if verified, does not affect our results. However, we
estimate a specification where productivity depends on the RER individually and interacted
with an indicator for traded industries, that indicator individually, a time trend and the
trend interacted with the traded industries' indicator.3 The evidence in Table 5, columns
(1)-(5), suggests that the RER effect is not verified: RER depreciations are significantly
associated with a productivity decline in traded industries. To examine directly whether the
productivity gains due to lower tariffs are robust to variation in the RER, we include the
RER in Eq. (6). The results in columns (6)-(9) show a negative and significant impact of
tariffs on productivity. In the long run, a RER depreciation accompanying trade liberaliza-
tion may protect producers by increasing the' relative price of imports, partly counteracting
the pressure for productivity improvement and survival brought by tariff reductions. How-
ever, estimating Eq. (6) for a period where both liberalization and RER depreciation occur
(1985-1989), the negative effect of tariffs is maintained. The same is verified if the RER is
included directly. So, the effect of a depreciation on producers' incentives does not overcome
the effect of tariff liberalization. Overall, the negative impact of tariffs on productivity is
robust to variation in the RER.
Third, several other robustness checks are performed. Eq. (6) is reestimated (i) using
3'RER data is taken from the IMF International Financial Statistics (based on relative consumer prices).
32We classify industries according to their degree of trade orientation as in Nishimizu and Robinson (1984).
During 1980-1991, import-competing industries have an average import penetration ratio above 10%, while
export-oriented industries have an average export orientation ratio above 10%, all remaining industries are
nontraded. If an industry has both ratios above 10%, it is classified as traded. All results are robust to a
change in the cutoff point defining traded industries from 10 to 8%.
33Also, we attempt to find evidence of plants' changes in capacity utilization by considering correlations of
plant productivity growth and output growth as in Pavcnik (2002). If in response to a RER depreciation
output expands (contracts) in traded (nontraded) industries without a change in inputs, and correspondingly
measured productivity expands (contracts), the correlation between changes in plant output and changes in
plant productivity should be strong and positive. This correlation is positive but weak, ranging from 0.044
in glass to 0.335 in furniture. Also, in traded industries in years of RER depreciation, average inventories of
finished goods (in levels or as a fraction of output) do not systematically decline and the percentage of plants
running down output and raw materials inventories (i.e., reporting lower inventories in December than in
January of that year) does not systematically increase.
17



data for the protection and second liberalization periods only (which amounts to a framework
directly comparable to that in before-after studies), (ii) measuring trade protection by two-
period lagged tariffs, and (iii) adding industry-level output as a regressor to account for
procyclicality and address the possibility that the negative effect of tariffs is not a real
productivity effect but rather it results from the procyclical nature of aggregate productivity
combined with lower output levels under high trade protection. In all three cases, the
negative and significant effect of tariffs on productivity is maintained.34 Finally, to address
the concern that our findings could result from the very disaggregated level of analysis, we
examine whether reduced trade protection generates productivity gains at the industry-level.
Tariffs have a negative effect on industry-level productivity.
5.4. Skill Intensity, Imports of Intermediate Inputs and Investment
The impact of tariff protection on plant productivity reflects to a large extent within-
plant changes in productivity. In this section, we provide evidence on the possible sources
of productivity gains at the plant level related to trade liberalization.
Focusing on the ratio of skilled employment to total employment, we find that a large
majority of plants with productivity gains under liberalization experience an increase in their
skilled labor intensity of production.35 This finding can be interpreted as an improvement in
the underlying product-mix and/or an increase in plants' technological sophistication (Hunt
and Tybout, 1998) due to increased foreign competition.
Increased imports of intermediate inputs can be a crucial mechanism for productivity
gains under trade liberalization. Imported inputs are utilized by about 25% of the plants
in our sample. Almost half of the plants do not change their imports of inputs across years
or trade regimes, most remain null. However, for plants that change their imports, most of
those having productivity gains under trade liberalization increase their imports of inputs.
Industries differ in the degree to which production. relies on imported inputs, therefore in
the degree to which this mechanism may underlie productivity gains. We find that a much
larger fraction of plants with productivity gains under trade liberalization increase their
imports of inputs in the industries whose production depends more heavily on imported
34We also reestimate Eq. (6) using contemporaneous tariffs (despite the endogeneity problems). Their
impact on productivity is negative and smaller than that of one-period lagged tariffs.
35This finding is robust to focusing on each year separately, on 3-digit industries with the largest relative
changes in tariffs and to measuring skill intensity by the ratio of plant skilled to unskilled employment.



inputs, compared to the fraction of plants increasing their imports in other industries.36
Some models of investment in productivity improvement through technology acquisi-
tion show that protection favors that acquisition. Nevertheless, it is appealing to argue the
contrary, i.e., for productivity to increase under liberalization, plants need to invest in tech-
nology, in particular imported. We use plant-level machinery investments and industry-level
machinery imports to examine this issue.37 Colombian plants' investment is lumpy and in
any year or trade regime, about 25% of plants do not invest in machinery. Under trade liber-
alization, however, most plants with productivity gains increase their machinery investments
in real terms or as a fraction of output. Interestingly, under trade protection, a large fraction
of plants with productivity losses reduce their investments in machinery. These conclusions
are tentative given the imperfections of the data, e.g., machinery investments do not nec-
essarily represent the acquisition of productivity-enhancing technology. Industry machinery
imports provide some insight on whether liberalization-related technology diffusion is crucial
for productivity gains. Imports of machinery for use in manufacturing (as a percentage of
GDP) increase in 1980-1981, sharply decline with increased protection and recover after 1986
(declining in 1991). Similar paths are verified for machinery used by the 3-digit industries
textiles, leather, paper, printing, food and metals.
A growing micro-level literature on exporting and productivity tests the hypotheses of
self-selection of more productive plants into exports markets versus learning-by-exporting
(Bernard and Jensen, 1999a, 1999b for the U.S.; Clerides, Lach and Tybout, 1998 for Colom-
bia, Mexico and Morocco). We examine whether a plant's export status influences its pro-
ductivity gains due to trade liberalization. As in previous studies, our findings suggest that
exporting in year t is not associated with productivity gains from year t to year t + 1.
36Nominal imports of inputs are normalized by a deflator that weights the costs of domestic and imported
inputs. If imported input prices increase, e.g., by less than domestic input prices, the deflated values are
measured with error, being lower than true imported inputs. We believe that this measurement problem is
not serious since the results are qualitatively similar for deflated values of imported inputs and for imported
inputs as a fraction of total intermediate inputs. We consider two measures for the dependence of an
industry's production on imported inputs: the average ratio of industry-level imported inputs to industry-
level output calculated across years and the average ratio of plant-level imported inputs to plant-level output
calculated within industries across years.
37Note that plant-level machinery investments are not disaggregated into domestic and foreign (as is the
case in the theoretical models). Industry-level machinery imports are taken from World Trade Analyzer
1980-1991 (Statistics Canada CD-Rom), commodity class 7 (Machinery and transport equipment) in the
SITC revision 2 classification. Aggregating the subclasses in class 7, one obtains machinery imports for
use in manufacturing industries. For some subclasses, it is possible to identify unambiguously which 3-digit
ISIC industry uses the imports: e.g., subclass 726 printing and bookbinding machinery and parts is used by
industry 342 printing. But these data do not indicate which plants use the imported machinery.
19



5.5. Average Impact of ERP and Import Penetration
To check the robustness of the findings with tariffs, we use alternative measures of trade
protection. ERP provide information on the protective structure resulting from tariffs on
output and on imported inputs.38 The results from estimating Eq. (6) with lagged 3-digit
ERP are presented in Table 6, columns (1)-(4). The impact of ERP on plant productivity is
positive, not significant with industry effects and significant with plant fixed effects.39 The
contrast relative to the negative impact of tariffs could stem from a difference in the samples
used. However, using the ERP sample, a negative impact of tariffs on productivity is still
obtained. There appears to be a real difference between the impact of nominal protection and
that of effective protection to final output on productivity that is curious given the positive
correlation between tariffs and ERP during the period. Ultimately, ERP coefficients are
not robust and become negative eliminating the most influential observations or the outliers,
whereas the coefficients on tariffs are always negative. Also, ERP coefficients are expected to
be insignificant since data requirements (e.g., coefficients from input-output tables) introduce
serious noise in ERP calculations, which is confirmed in OLS specifications.
Import penetration ratios are an outcome of trade policy and measure the exposure to
foreign competition. We expect these ratios to affect productivity positively if plants lower
costs and become more efficient when import competition increases. However, if imports are
endogenous with respect to domestic industries' productivity (e.g., in a Ricardian framework,
imports are attracted to relatively less productive industries), a negative correlation between
import penetration and productivity could be found. Eq. (6) is estimated using lagged 3
and 4-digit import penetration ratios and the results are presented in Table 6, columns (5)-
(12). In general, import penetration has a large, positive and significant impact on plant
productivity. For example, column (7) indicates that a 10% increase in import penetration
into 3-digit industries increases plant productivity by 6%. These findings are not driven by
the period used in the import penetration specifications. In fact, using an alternative period,
38However, the use of ERP could be conceptually problematic. Our production function estimation allows for
changes in the mix of primary factors and intermediate inputs, but ERP represent protection to value-added
which assumes that intermediate inputs are used in fixed proportion to output.
39The ERP on a final good declines if either tariffs on the final good decline or if tariffs on intermediate
inputs increase, in relative terms. So, an interpretation of the significant positive impact in fixed effects
specifications could be that lowering ERP leads to productivity losses if that occurs via increased tariffs
on intermediate inputs. With no information on the tariffs on inputs imported and used in each industry,
we cannot tell whether this interpretation is correct. But a requirement for its validity is verified: in a
fixed effects specification relating productivity to tariffs and ERP, a negative tariff coefficient (-1.013) and a
positive ERP coefficient (0.465) are obtained and both are significant.
20



for which data on tariffs -and import penetration is available (1981, 1984-1989), the same
robust conclusion is drawn: lagged import penetration affects positively plant productivity
and lagged tariffs affect it negatively.
6. Productivity, Trade Policy and Plant Size
6.1. Tariffs, Plant Size and Exports
Of particular interest is the question of whether tariffs affect productivity differentially
according to plant size.40 We estimate the following specification:
prit =3` 0S +I3OL + At X  I + At x I + f ls(TPt-1 x IS) +I31L(TPtj I x IL) + I i + U j, (7)
where /%s, 80L, Is, IL are intercepts and indicator variables for small and large plants,
respectively, and TPtj1, At, Ii are defined as in Eq. (6). Size is defined as employment at a
plant in its first year in the sample to mitigate the endogeneity problem that would result if
Is and IL were indexed by time, given the potential impact of trade liberalization on plant
size.41 Panel A of Table 7 presents the results from estimating Eq. (7) with small plants
having less than 50 employees. F-tests indicate that large plants are significantly more
productive than small plants. F-tests also show that the effect of tariffs on productivity
differs significantly across plant size. This effect is much more negative for large plants. For
example, the coefficients in column (5) imply that reducing tariffs by 10% would result in
a productivity gain of 4% for large plants, twice as much as the gain for small plants. The
results are robust to restricting year effects to be equal across all plants and to changing the
cutoff defining small plants to 20 or 100 employees.
An alternative definition of plant size is the plant's market share in total industry output
in its first year in the sample, msh1. The following specification is estimated, allowing for
4OSome theoretical and empirical evidence is available regarding the differential effect of trade policy on
output and price-cost margins across plant size. Roberts and Tybout (1996) find that, within Colombian
industries facing increased import penetration, larger plants experience stronger declines in price-cost mar-
gins. In contrast, Dutz (1996) shows how incumbents adjust output to liberalized import quotas in an
oligopoly model, concluding that smaller plants, with lower market shares and higher marginal costs, expe-
rience relatively stronger output contractions in response to increased imports. Also, developing countries'
manufacturing industries are often dualistic, accommodating a few oligopolistic producers and a large num-
ber of small firms that operate under stronger competition, are more sensitive to the economic environment
and more flexible to change.
4llncreased exposure to foreign competition may increase plant size by increasing the elasticity of demand
(reinforced by entry and exit). But, it may reduce demand, causing industry contraction and decreasing
plant size. Most empirical studies find that trade liberalization is associated with reduced plant size (Dutz,
1996; Roberts and Tybout, 1991; Tybout and Westbrook, 1995).
21



a non-linear relationship between plant size and productivity:
Pri=    + At +   limsh         shXi)2 + 33TPtS1 + I 4(TPt11 x mshii) + Ii + u,j.    (8)
Panel B of Table 7 presents the results for plant market shares relative to 3-digit industry
output. Plant productivity increases with market share at a diminishing rate. Controlling
for industry effects, tariffs affect negatively and significantly productivity, as in Table 4.
Tariff liberalization has a more positive impact on the productivity of plants with higher
market shares. In column (2), for example, the marginal effect of tariffs on productivity for
a plant with the average market share is -0.11.42 The results are qualitatively similar for
market shares relative to 4-digit industry output.
The finding that productivity gains are higher for large plants suggests that trade liber-
alization brings a decline in 'inefficiency rents' benefiting large producers and/or that large
producers' output competes more directly with imports. Also, large plants use imported
inputs in production more often than small plants. Therefore, the stronger effect of lib-
eralization on large plants' productivity could be operating through the imported inputs
mechanism. In fact, we find a much higher proportion of large plants with productivity
gains under liberalization that increase their imports of inputs than the proportion of small
plants doing so.
We also examine whether the impact of trade policy on productivity is stronger for
plants engaging in export activities. Our specification parallels that of Eq. (7) and includes
lagged tariffs, year and industry effects, an indicator variable for exporters defined as plants
exporting in their first year in the sample and the interaction of this variable with tariffs.43
The estimates indicate that the productivity of exporters is more positively affected by trade
liberalization than that of non-exporters. A similar result is obtained when exporters are
defined as plants exporting in every sample year. These results are consistent with the finding
of a stronger positive effect of trade liberalization on large plants, since Colombian exporters
are significantly larger than non-exporters. Size and export status are highly correlated
across plants, so it is difficult to sort out their individual influence on the impact of trade
policy on plant productivity.
42The marginal effect of tariffs in Eq. (8) is obtained as ,3 + 34msh (msh is the average market share in the
sample). The averages of the regressors used to calculate the corresponding marginal effects are reported in
Tables 7-10. The marginal effects of the regressors in Eqs. (7)-(9) are shown in Fernandes (2002).
43Allowing a plant's export status to be time-varying would result in an endogeneity problem since trade
policy affects a plant's decision to export.
22



6.2. ERP, Import Penetration and Plant Size
Table 8 assesses the robustness of the finding that the effect of tariffs on productivity
depends on plant size. Panel A shows the results from estimating Eq. (7) with lagged ERP
(columns (1)-(2)) and lagged import penetration ratios (columns (3)-(6)). Panel B presents
the results from estimating Eq. (8) with lagged ERP (columns (1)-(2)) and lagged import
penetration ratios (columns (3)-(6)). Both panels are consistent with our findings when
using tariffs as a measure of trade protection."4
7. Productivity, Trade Policy and Domestic Competition
7. 1. Tariffs and Domestic Competition
In this section, we examine whether trade liberalization affects more strongly the produc-
tivity of plants in industries with less domestic competition.45 We use two measures from
the industrial organization literature that capture different dimensions of domestic competi-
tion. Herfindahl indexes summarize the degree of market share inequality across plants in an
industry. Turnover rates reflect, at least imperfectly, the market power of large plants and
their ability to inhibit entry into an industry, as well as sunk costs preventing exit. Given
the potential impact of trade liberalization on concentration, entry and exit, the degree of
domestic competition is measured by the value of Herfindahl indexes and turnover rates in
the first sample year and taken as a fixed industry characteristic. We estimate the following
specification:
,k=                  ++X                        + Ik +Ui,k
pr7i' =0+A +  A l3TPtjj 13+2DC  +  3(TPtl DC')         k +3,k        (9)
where DC' represents the degree of domestic competition in industry j and TPt-ji, At, I i
are defined as in Eq. (6). Since DCO is a fixed characteristic of industries indexed by j,
only industry effects at a higher level of aggregation, k, are identified in Eq. (9). The results
from estimation with lagged tariffs and Herfindahl indexes are presented in Table 9, columns
(1)-(6). Overall, we find that plants in less competitive domestic industries according to 3 or
4-digit Herfindahl indexes have lower productivity. Columns (1)-(6) show that the effect of
tariff liberalization on plant productivity depends on whether 3 or 4-digit Herfindahl indexes
44The findings in Panels A and B are qualitatively unchanged when restricting year effects to be equal across
all plants, changing the cutoff defining small plants from 50 to 20 or 100 employees or using plant market
shares relative to 4-digit industry output, respectively.
45Roberts and Tybout (1996) find that increased import penetration leads to larger reductions in the price-
cost margins of the more concentrated Colombian industries.
23



are used. However, a more careful analysis with marginal effects of tariffs at average 3 and
4-digit Herfindahl indexes indicates that tariff liberalization affects more positively plant
productivity in less competitive domestic industries.
The results from estimating Eq. (9) with 3-digit turnover rates are presented in Table
9, columns (7)-(8). The marginal effects of turnover rates at average tariffs indicate that
plants in less competitive domestic industries have lower productivity. Both the coefficients
on tariffs and the marginal effects of tariffs at average turnover rates indicate that tariff lib-
eralization affects more positively plant productivity in less competitive domestic industries.
7.2. ERP, Import Penetration and Domestic Competition
Table 10 examines the robustness of the findings in section 7.1. Columns (1)-(4) show
the results from estimating Eq. (9) with lagged ER.P, Herfindahl indexes and turnover
rates and columns (5)-(12) show the results from estimating Eq. (9) with lagged import
penetration ratios, Herfindahl indexes and turnover rates. These results are consistent with
those obtained when using tariffs as a measure of trade protection except when ERP and 3-
digit Herfindahl indexes are used as measures of trade protection and domestic competition,
respectively. In this case, the marginal effects suggest that ERP affect positively plant
productivity in less competitive industries.
8. Conclusion
This paper provides new evidence on the link between trade policy and plant-level pro-
ductivity using Colombian data. The consistent production function estimates reveal, by
comparison to OLS, the endogeneity of plants' input choices with respect to productivity.
Relying on productivity estimates that correct for such endogeneity is crucial to estimate
accurately the effect of trade policy on productivity. Our analysis relies on cross-industry
and time variation in measures of trade policy and thereby circumvents the shortfalls of
previous studies that focus on a single episode of liberalization.
Our findings suggest that trade liberalization has a strong positive impact on plant pro-
ductivity, even after controlling for unobserved plant and industry heterogeneity and for
the real exchange rate. The use of lagged tariffs, the control for unobserved industry fixed
characteristics, and the evidence that changes in Colombian tariffs were generally a policy
response to fiscal and external imbalances suggest that the negative impact of tariffs does
24



not reflect the endogeneity of protection. Plant exit is not a major source for increased pro-
ductivity under trade liberalization. Within-plant productivity gains are the crucial source
and appear to be associated with increased skilled labor intensity of production, imports of
intermediate inputs and machinery investments. We find that the impact of trade liberaliza-
tion on productivity is stronger for large plants and for plants in industries with less domestic
competition. Further research on the plant-level characteristics associated with productivity
gains due to trade liberalization would be fruitful.
From a policy perspective, our research suggests that the liberalization of policies pro-
tecting manufacturing industries from foreign competition brings benefits in terms of pro-
ductivity. Of course, the generalization of our results to other countries' micro-level datasets
is not immediate. The effect of trade liberalization on plant productivity depends on the
details of the liberalization and on its interaction with institutions and other policies.
25



Appendix
Two decompositions of changes in industry productivity are considered in Section 5.3.
The first is given by:
Pt -PtJit=( E      11r)             pr_1:) E  priL. (s - s   1) +  E spri  s  pri 
iECont                iECont             iEEnt        iEEx
where Pt' is industry productivity (weighted average of plants' productivities), prit and sit
are plant i's productivity and market share in industry j output for continuing plants Cont,
entrant plants Ent and exiting plants Ex. The first term in the decomposition represents
the change in continuing plants' productivity. The second term represents the output share
reallocation among continuing plants with different productivity levels. Finally, the third
and fourth terms represent average productivity of entrant and exiting plants, respectively.
The second follows Aw, Chen and Roberts (2001) and is given by:
t-PA   =    E   (si+t-  st (m'd -Prit-1) +  E  2(pri1L +Priti)(s. -s ¾)
iECont                        iECont
+2 (sExt_1 I+ Sint t)(PrE'xt-1 -PrEnt t) + 2 (PrE-'xt-1 + PrE-nt t)(SEx t-1 SEnt t
22
where pri and st are defined as above, sE tl and S  t are the market share of entrant and
exiting plants in industry j, and wE t-land wEnt t are the output share-weighted average
productivity of entrant and exiting plants in industry j. The first term represents the change
in continuing plants' productivity. The second term represents the output share reallocation
among continuing plants with different productivity levels. The third term represents the
difference between the productivity of entrant plants in year t and of exiting plants in year
t - 1. Finally, the fourth term represents the output share reallocation between entrant and
exiting plants.
26



References
Aw, B., Chen, X. and M. Roberts, 2001. "Firm-Level Evidence on Productivity Differentials
and Turnover in Taiwanese Manufacturing," Journal of Development Economics 66(1), pp.
51-86.
Basu, S. 1996. "Procyclical Productivity: Increasing Returns or Cyclical Utilization?,"
Quarterly Journal of Economics 111(3), pp. 719-51.
Bernard, A. and B. Jensen, 1999a. "Exceptional Exporter Performance: Cause, Effects or
Both?," Journal of International Economics 47(1), pp. 1-25.
Bernard, A. and B. Jensen, 1999b. "Exporting and Productivity: The Importance of Real-
location," Tuck School of Business, Dartmouth College, Manuscript.
Campbell, J., 1998. "Entry, Exit, Embodied Technology and Business Cycles," Review of
Economic Dynamics 1(2), pp. 371-408.
Clerides, S., Lach, S. and J. Tybout, 1998. "Is Learning by Exporting Important? Micro-
dynamic Evidence from Colombia, Mexico, and Morocco," Quarterly Journal of Economics
113(3), pp. 903-947.
Corden, W., 1966. "The Structure of a Tariff System and the Effective Protective Rate,"
Journal of Political Economy 74(3), pp. 221-237.
Dollar, D., 1992. "Outward-Oriented Developing Economies Really Do Grow More Rapidly:
Evidence from 95 LDCs, 1976-85," Economic Development and Cultural Change 40(3), pp.
523-544.
Dutz, M., 1996. "Oligopolistic firms' Adjustment to Quota Liberalization: Theory and
Evidence," in Roberts, M. and J. Tybout, eds. Industrial Evolution in Developing Countries:
Micro Patterns of Turnover, Productivity and Market Structure. NY: Oxford University
Press.
Edwards, S., 1993. "Openness, Trade Liberalization, and Growth in Developing Countries,"
Journal of Economic Literature 31(3), pp. 1358-1393.
Edwards, S., 1998. "Openness, Productivity and Growth: What Do We Really Know?,"
Economic Journal 108(447), pp. 383-398.
Fernandes, A., 2002. "Trade Policy, Trade Volumes and Plant-Level Productivity in Colom-
bian Manufacturing Industries," Yale University Economic Growth Center Discussion Paper
No. 847.
Garay, L., 1991. Apertura y Proteccion: Evaluacion de la Politica de Importaciones. Tercer
Mundo Editores Universidad Nacional de Colombia.
Garcia, J., 1991. "Colombia," in Papageorgiou, D., Michaely, M. and A. Choski eds. Liber-
alizing Foreign Irade, Vol. 4.
GATT, 1990. Trade Policy Review Colombia.
Goh, A.-T., 2000. "Opportunity Cost, TRade Policies and the Efficiency of firms," Journal
of Development Economics 62(2), pp. 363-383.
27



Goldberg, P. and N. Pavcnik, 2001. "Trade Protection and Wages: Evidence from the
Colombian Trade Reforms," NBER Working Paper 8575.
Griliches, Z. and J. Mairesse, 1995. "Production Functions: the Search for Identification,"
NBER Working Paper 5067.
Grossman, G. and E. Helpman, 1991a. Innovation and Growth in the World Economy. MIT
Press. Cambridge, MA.
Grossman, G. and E. Helpman, 1991b. "Quality Ladders in the Theory of Growth," Review
of Economic Studies 58(1), pp. 43-61.
Grossman, G. and E. Helpman, 1994. "Protection for Sale," American Economic Review
84(4), pp. 833-850.
Hallberg, K. and W. Takacs, 1992. "Trade Reform in Colombia: 1990-94," in Cohen, A.
and F. Gunter, eds. The Colombian Economy: Issues of Trade and Development. Oxford:
Westview Press.
Harrison, A. 1994. "Productivity, Imperfect Competition and Trade Reform: Theory and
Evidence," Journal of International Economics 36(1-2), pp. 53-73.
Harrison, A. 1996. "Openness and Growth: a Time-Series Cross-Country Analysis for De-
veloping Countries," Journal of Development Economics 48(2), pp. 419-447.
Holmes, T. and J. Schmitz, 2001. "A Gain from Trade: from Unproductive to Productive
Activities," Journal of Monetary Economics 47(2), pp. 417-446.
Hopenhayn, H. and R. Rogerson, 1993. "Job Turnover and Policy Evaluation: A General
Equilibrium Analysis," Journal of Political Economy 101(5), pp. 915-938.
Hunt, J. and J. Tybout, 1998. "Does Promoting High-Tech Products Spurs Development?,"
FEEM Fondazione Eni Enrico Matteo Working Paper No. REG 42.98.
Jensen, J., McGuckin R. and K. Stiroh, 2001. "The Impact of Vintage and Age on Pro-
ductivity: Evidence from U.S. Manufacturing Plants," Review of Economics and Statistics
83(2), pp. 323-332.
Kim, E., 2000. "Trade Liberalization and Productivity Growth in Korean Manufacturing
Industries: Price Protection, Market Power and Scale Efficiency," Journal of Development
Economics 62(1), pp. 55-83.
Klette, J. and Z. Griliches, 1996. "The Inconsistency of Common Scale Estimators When
Output Prices are Unobserved and Endogenous," Journal of Applied Econometrics 11(4),
pp. 343-361.
Krishna, P. and D., Mitra, 1998. "Trade Liberalization, Market Discipline and Productivity
Growth: New Evidence from India," Journal of Development Economics 56(2), pp. 447-462.
Lee, J.-W., 1995. "Goverinment Interventions and Productivity Growth in Korean Manufac-
turing Industries," NBER Working Paper 5060.
Levine, R., and D. Renelt, 1992. "A Sensitivity Analysis of Cross-Country Growth Regres-
sions," American Economic Review 82(4), pp. 942-963.
28



Levinsohn, J., 1999. "Employment Responses to International Liberalization in Chile,"
Journal of International Economics 47(2), pp. 321-344.
Levinsohn, J. and A. Petrin, 2001. "Estimating Production Functions Using Inputs to
Control for Unobservables," Review of Economic Studies, forthcoming.
Marschak, J. and W. Andrews, 1944. "Random Simultaneous Equations and the Theory of
Production," Econometrica 12(3-4), pp. 143-205.
Nishimizu, M. and S. Robinson, 1984. "Trade Policies and Productivity Change in Semi-
industrialized Countries," Journal of Development Economics 16(1-2), pp. 177-206.
Olley, S. and A. Pakes, 1996. "The Dynamics of Productivity in the Telecommunications
Equipment Industry," Econometrica 64(6), pp. 1263-1297.
Pakes, A., 1994. "Dynamic Structural Models, Problems and Prospects Part II: Mixed
Continuous-Discrete Control Problems and Market Interactions," in Sims, C. ed. Advances
in Econometrics. Cambridge University Press. Cambridge, MA.
Pavcnik, N., 2002. "Trade Liberalization, Exit and Productivity Improvements: Evidence
from Chilean Plants," Review of Economic Studies 69(1), pp. 245-276.
Power, L., 1998. "The Missing Link: Technology, Productivity and Investment," Review of
Economics and Statistics 80(2), pp. 300-313.
Pritchett, L., 1996. "Measuring Outward Orientation in LDCs: Can It Be Done?," Journal
of Development Economics 49(2), pp. 307-335.
Roberts, M. and J. Tybout, 1991. "Size Rationalization and Trade Exposure in Developing
Countries," in Baldwin, R. ed. Empirical Studies of Commercial Policy. University of
Chicago Press for NBER.
Roberts, M. and J. Tybout, eds. 1996. Industrial Evolution in Developing Countries: Micro
Patterns of Thrnover, Productivity and Market Structure. NY: Oxford University Press.
Rodrik, D., 1991. "Closing the Productivity Gap: Does Trade Liberalization Really Help?,"
in Helleiner, G. ed. Trade policy, Industrialization and Development. Clarendon Press.
Oxford.
Rodrik, D., 1995. "Trade and Industrial Policy Reform," in Behrman, J. and T. N. Srinivasan
eds. Handbook of Development Economics, Vol. 3. North Holland, Amsterdam.
Rodrik, D., and F. Rodriguez, 2001. "Trade Policy and Economic Growth: A Skeptic's
Guide to the Cross-National Evidence," in Bernanke, B. and K. Rogoff eds. Macroeconomics
Annual 2000. MIT Press for NBER. Cambridge, MA.
Sachs, J. and A. Warner, 1995. "Economic Reform and the Process of Global Integration,"
Brookings Papers on Economic Activity 1, pp. 1-118.
Solow, R., 1957. "Technical Change and the Aggregate Production Function," Review of
Economics and Statistics 39(3), pp. 312-320.
Trac,a, D., 1997. "Import-Competition, Market Power and Productivity Change," Wis-
senschaftszentrum Berlin fur Sozialforschung Working Paper FS-IV 97-29.
29



Trefler, D., 1993. 'Trade Liberalization and the Theory of Endogenous Protection: An
Econometric Study of U.S. Import Policy," Journal of Political Economy 101(1), pp. 138-
160.
Tybout, J., 1992k "Linking Trad'e and Productivity: New Research Directions," The World
Bank Economic Review 6(2), pp. 189-211.
Tybout, J. and M. Westbrook 1995. 'rade Liberalization and the Dimensions of Efficiency
Change in Mexican Manufacturing Industries ," Journal of International Economics 39(1-2),
pp. 53-78.
Urrutia, M., 1994. "Colombia," in Williamson, J., ed. The Political Economy of Policy
Reform. Washington, DC: Institute for International Economics.
Vousden, N. and N. Campbell, 1994. "The Organizational Cost of Protection," Journal of
International Economics 37(3-4), pp. 219-238.
World Bank, 1984. Colombia: Economic Development and Policy under Changing Condi-
tions. World Bank. Washington, DC.
World Bank, 1989. "Colombia: Commercial Policy Survey," Report no. 7510-CO. World
Bank. Washington, DC.
World Bank, 1991. Colombia: Industrial Competition and Performance. World Bank. Wash-
ington, DC.
Zerda, A., 1992. "El Comportamiento de la Productividade y sus Determinantes: Un Debate
Abierto," in Garay, L. ed. Estrategia Industrial e Insercion Internacional. FESCOL.
30



Table 1
Nominal Tariffs across 3-digit Industriesa
Industry                      Average           Average           Average
Tariff (%)        Tariff (%)       Tariff (%)
First Liberalization  Protection  Second Liberalization
Period            Period            Period
311 Food                        30.5              40.6              43.4
312 Food-miscellaneous          28.4              35.8              37.7
313 Beverages                   54.5              73.9              58.1
314 Tobacco                     26.4              34.5              42.0
321 Textiles                    57.3              82.6              48.4
322 Apparel                     75.6             109;6              66.4
323 Leather products            40.8              53.1              39.8
324 Footwear                    56.4              84.0              71.8
331 Wood products               41.4              55.7              43.7
332 Furniture                   54.2              75.9              47.5
341 Paper                       28.7              38.9              35.3
342 Printing                    38.6              45.9              42.6
351 Industrial chemicals        20.2              26.7              24.0
352 Other chemicals             19.8              24.4              22.3
354 Petroleum derivatives       18.8              24.5              23.3
355 Rubber products             47.8              55.6              43.9
356 Plastics                    61.9              73.1              55.2
361 Ceramics                    47.4              61.6              47.8
362 Glass                       35.8              38.9              32.1
369 Nonmetallic minerals        29.4              36.2              30.8
371 Iron and steel              20.2              25.8              20.9
372 Nonferrous metals           20.1              26.6              18.9
381 Metal products              40.1              49.6              39.0
382 Nonelectrical machinery     23.6              30.1              20.9
383 Electrical machinery        34.4              43.5              31.7
384 Transport equipment         26.7              37.2              31.3
385 Professional equipment      25.1              30.4              24.4
390 Other manufacturing         37.1              49.2              37.3
a The periods used to compute the average tariffs are as follows: (i) 1976-1980 for the first
liberalization period, (ii) 1983-1984 for the protection period, and (iii) 1985-1988 for the second
liberalization period. Note that for industries 311 and 312 average tariffs in the protection period
are higher than in the first liberalization period but lower than in the second liberalization period.
The largest increase in tariffs for those industries only occurred in 1984 when the tariff levels were
45.1 for industry 311 and 39.8 for industry 312.
31



Table 2
Import Penetration Ratios and Export Orientation Ratios across 3-digit Industriesa
Industry                         Export Orientation Ratio (%)                      Import Penetration Ratio (%)
Year                                             Year
1980     1984    1985     1988    1991           1980    1984     1985     1988    1991
311 Food                      9.1      3.6     3.8      5.8    11.4           6.4      5.3      4.0      4.5     3.4
312 Food-miscellaneous        2.4      5.1     6.6      7.0     7.1           2.3      3.2      3.2     2.8      2.0
313 Beverages                 0.0      0.2     0.2      0.2     0.6           1.5      1.4      1.4      1.9     1.0
314 Tobacco                   0.6      0.9     0.7      3.1    13.4           9.5      4.4      3.1      1.8     4.1
321 Textiles                  7.7      5.3     5.7      7.6    16.8           3.1      1.9      2.1      2.8     4.8
322 Apparel                  17.6      7.3     9.8     27.7    47.3           1.9      3.1      2.7      3.3     4.5
323 Leather products         13.2     14.6    23.9     34.9    45.0            1.7     1.2      1.3      2.0     3.5
324 Footwear                 10.4      5.0     8.4     11.4    31.6           0.8      1.0      1.1      0.7     0.6
331 Wood products            10.5      5.1    12.6      6.5    15.8            7.3     6.6      3.2      5.2     5.4
332 Furniture                 4.6      2.4     6.2      4.1     8.1           1.5      0.9      0.3      0.9     1.0
341 Paper                     5.1      4.5     3.3      1.8     2.9          16.6     15.7     17.2     17.0    16.9
342 Printing                 10.1      7.0    11.9     16.9    26.5           11.5     7.9     11.2      6.2     5.3
351 Industrial chemicals      7.0      6.4     8.1     12.1     18.1          40.5    38.6     65.6     44.3    43.9
352 Otherchemicals            2.9      2.9     3.1      2.3     3.2           14.4    14.5     16.9     12.6    11.8
354 Petroleum derivatives     n a.    n.a.     n.a.     na.     n.a.           5.8     4.2      3.3      3.5     4.7
355 Rubber products           2.0      1.3     1.5      4.2     6.1           11.3     8.5      8.0      8.6    10.1
356 Plastics                  2.8      2.1     2.1     .1.3     2.5            2.0     2.2      1.9      2.1     3.6
361 Ceramics                  9.9      2.8     5.1      7.0    28.3            7.2     3.8      2.0      2.9     6.1
362 Glass                     9.7      5.2     4.3      4.2    10.6           9.5      6.1      5.7      4.7     8.2
369 Nonmetallic minerals      8.6      4.0     5.1      5.6     9.1           3.5      2.5      2.9.     3.5     2.9
371 Iron and steel            na.     na.      na.      na.    nra.          38.5     35.8     82.0     34.0    39.7
372 Nonferrous metals         2.6     10.3    13.9      7.4      4.8          51.2    51.8     92.2    .45.4    50.2
381 Metal products            7.1     .3.4     3.8      4.6    10.1           15.1    14.4     14.6     12.3    15.9
382 Nonelectrical machinery  12.1     4.1      6.6      5.7    17.6          71.8     70.5    178.8     68.3    70.2
383 Electrical machinery      3.1     1.4      3.3      4.6     9.6          38.6     37.8     42.4     36.4    43.5
384Transportequipment         3.0      0.9     1.2      0.7      5.6          40.7    31.8     46.6     28.3    29.7
385 Professional equipment   14.6      7.6     8.6      6.8    14.4           57.4    53.6    116.8     46.3    54.1
390 Other manufacturing      37.0     18.4    17.5     42.9    58.1           11.7     9.6      5.7      7.2    12.4
a The export orientation ratio is defined as the ratio of exports to total output (domestic output plus exports). The import penetration ratio is
defined as the ratio of imports to domestic demand (domestic output plus imports). n.a indicates that export data for petroleum derivatives and
iron and steel has irregularities.



Table 3
Production Function Estimates with Materials Controlling for Simultaneity Bias'
Industry            Input      OLS       Revenue   Nonparam/    R.                   Industry           Input       OLS       Revenue   Nonparam/    R
shares     GMM       scale                                                           shares     GMM       scale
321+322Textiles   Labor       0.316***    0.275     0.242'**                         369 Nonmetallic  Labor       04050**      0.330    0.381**
and Apparel                   (0.004)               (0.014)                          minerals                      (0.01)                 (0.02)
Energy      0.1460*     0.018     0.115***   1 .028'                                 Energy      0.212*0*    0.090     0.186***   0.992'
(0.003)               (0.007)                                                       (0.005)                (0.012)
Nobs              Materials   0.471 **    0.500     0.66*0*                          Nobs             Materials   0.406***     0.344     0.31 ***
20379                         (0.003)               (0.053)                          4502                          (0.004)               (0.084)
Capital     0.049***     0.207     0.011                                             Capital     0.047***    0.237     0.116***
(0 .003)              (0.03)                                                         (0 .005)              (0.038)
311+312 Food      Labor   .   022w        0.149     0.1540**                         3520ther         Labor       0.287***     0.194     0.269*-*
products                      (0.005)               (0.008)                          Chemnicals                    (0.007)               (0.014)
Energy       0.160**     0.034    0.095***   1.058'                                  Energy      0.018***    0.012      0.008     0.996'
(0.004)               (0.007)                                                       (0.004)                (0.009)
N. obs.           Materials   0.588***    0.647     0.731 ***                        Nobs             Materials   0.707***     0.521     0.658*-
17651                         (0.002)               (0.043)                         4296                           (0.005)               (0.045)
Capital     0.088***     0.170     0.077*                                            Capital     0.071***    0.273     0.061**
(0.003)               (0.042)                                                       (0.004)                (0.027)
381 Metal         Labor       0.329**0    0.266     0.288***                         356 Plastics     Labor -     0.3250**     0.196     0.303**
products                      (0.006)               (0.012)                                                        (0.008)               (0.015)
Energy      0.095*      .0.024    0.053***   0.962'                                  Energy      0.01 40**   0.032      -0.015*   1.032'
(0.004)               (0.007)                                                       (0.005)                (0.008)
Nobs              Materials   0.587*0*    0.509     0.523**                         NMobs             Materials   0.596***     0.548     0.642***
8581                          (0.004)               (0.046)                          4059                          (0.005)               (0.034)
Capital     0.048***     0.201    0.098***                                           Capital     0.112***    0.224     0.103***
(0.003)               (0.031)                                                       (0.005)                (0.03)
331 +332 Wood     Labor       0.234***    0.302     0.21 *"                          324 Footwear     Labor       0.259*0*     0.255     0.228*0*
products and                  (0.007)               (0.015)                                                        (0.008)               (0.021)
furniture         Energy      0.115***    0.021     0.096***   0.886'                                 Energy       0.109**-    0.010     0.0670*    1.018'
(0.005)               (0.008)                                                       (0.006)                (0.029)
Nobs              Materials,  0.6350*0    0.496     0.480**                          Nobs             Materials   0.6080**     0.531     0.674***
5652                          (0.005)               (0.069)                          3594                          (0.006)               (0.049)
Capital     0.033***     0.181     0.1***                                            Capital     0.027*0*    0.204      0.049
(0.003)               (0.026)                                                       (0.005)                (0.032)
342 Printing      Labor       0.586*"*    0.295     0.516"*                          384 Transport    Labor       0.372**-     0.274     0.353**
(0.011)               (0.025)                         equipment                     (0.012)                (0.025)
Energy      -0.026*0*   0.017    -0.055*"    1.096                                   Energy      0.025**'    0.024      0.009     0.922'
(0.007)               (0.013)                                                       (0.007)                (0.012)
Nobs              Materials   0.484*"*    0.443     0.523***                         Nobs             Materials   0.574"'0     0.491     0.47*"*
5224                          (0.007)                (0.04)                          3310                          (0.006)               (0.083)
Capital     0.077*"*     0.244    0.112*0*                                           Capital     0.097*0*    0.211       0.09
(0.005)               (0.03)                                                         (0,007)               (0.058)
382 Nonelectrical  Labor      0.302"'*    0.280     0.284*0*                         383 Electrical   Labor        0.291 "*-   0.264     0.2860**
machinery                      (0.01)               (0.016)                          mnachinery                    (0.01)                (0.022)
Energy      0.056'**    0.023     0.046***   0.842'                ,                 Energy      0.04"       0.018     0.031***   0.907'
(0.006)              (0.009)                                                 -       (0.006)               (0.012)
Nobs              Materials   0.612"*'    0.468     0.381*"                          Nobs             Materials    0.669*"*    0.518     0.526**
4585                          (0.005)               (0.162)                          2824                          (0.007)               (0.121)
Capital     0.0840**     0.229     0.131                                             Capial     .0068**.     0.200    - 0.064..
(0.005)              (0.084)                                                        (0.006)                (0.055)



Table 3 (continued)
Industry            Input       OLS      Revenue   Nonparan./   R.                  Industry            Input       OLS      Revenue   Nonparam./    R.
shares     GMM       scale                                                          shares     GMM        srale
341 Paper         Labor        O.2***     0.181     0.204***                        371 Isnnandsteel  Labor       0.238***    0.227     0.201***
(0.01)                (0.03)                                                       (0.019)                (0.029)
Energy      0.065***     0.032    0.043***   0.917'                                 Energy       O.O7***     0.049    0.O51***    1.041'
(0.005)               (0.007)                                                       (0.012)               (0.013)
Robs              Materials   0.723***    0.581     0.57*s*                         R~obs             Materials   0.674***    0.528     0.78***
2017                          (0.005)                (0.09)                         823                           (0.009)                (0.183)
Capital     0.067***     0.205      0.1**                                           Capital     0.0780**     0.196      0.009
(0.005)               (0.044)                                                       (0.01)                (0.131)
31 3 Beverages    Labor       0.265*0*    0.197     0.233***                        362 Glass         Labor       0.35***     0.272     0.342***
(0.016)                (0.03)                                                       (0.015)               (0.031)
Energy      0.1 93***    0.027    0.119***   0.909'                                 Energy      0.119***     0.075    0.102***    1. 124'
(0.009)               (0.021)                                                       (0.008)               (0.016)
N nbs.           Materials    0.55S**$    0.438     0.543*0*                       NRobs              Matenals    0.581*5*    0.435     0.67***
1975                           (0.01)               (0.079)                         815                           (0.011)                (0.071)
Capital      0.05***     0.338      0.014                                           Capital     0.025***     0.218      0.01
(0.009)               (0.056)                                                        (0.01)               (0.056)
351 Industrial    Labor       0.095***    0.148     0.1 16**                        361 Ceranmics     Labor        0.**       0.455     0.506***
chenmicals                    (0.017)                (0.03)                                                       (0.026)                (0.067)
Energy      0.084***     0.062     0.080**   0.913'                                 Energy       0.1 2***    0.092    0.081 **    1.05'
(0.009)               (0.022)                                                       ( 0.014)              (0.022)
Nobs              Materials   0.555***    0.497     0.278**                         NMobs             Materials   0.45***     0.259     0.386***
1713                          (0.008)               (0.121)                         503                           (0.01 8)               (0.133)
Capital     0.238***     0.294    0.439***                                          Capital     0.026**      0.194      0.077
(0.011)               (0.132)                                                     --(0.012)               (0.094)
323 Leather       Labor       0.198***    0.233     0.243*'                         372 Nonferrous    Labor       0.355***    0.230     0.315**
products                       (0.01)               (0.01 9)                        metals                        (0.035)                (0.047)
Energy      O.O3S**$     0.014      0.009    0.845'                                 Energy       0.1 7***    0.039    .0.08**    0.978'
(0.007)               (0.013)                                                       (0.023)               (0.035)
R.obs            Materials    0.684***    0.571     0.53*w                           Robs             Materials   0.416 *     0.489     0.549 **
1462                          (0.009)                (0.1 )                         435                           (0. 01 9)              (0.17)
Capital     O.O35**      0.182      0.063                                           Capital      0.1 38***   0.242      0.034
(0.006)               (0.048)                                                       (0.018)               (0.176)
355 Rubber        Labor       0.294***    0.234     0.261 ***                       354 Petmieum      Labor       0.213***    0.104     0.283***
products                      (0.01 3)               (0.025)                        derivatives                   (0.041)                (0.079)
Energy      0.046*0*     0.035    0.051 **   0.862                                  Energy      0.052***     0.042      0.027     1.001'
(0.009)               (0.014)    .(0.017)                                                                 (0.029)
Nobs             Materials    0.673***    0.511     0.53**                          NRobs             Materials   0.821 *'*'  0.634     0.520*
1196                           (0.01 )              (0.054)                         306                           (0.01 9)               (0.159)
Capital     0.05l **     0.220      0.02                                            Capital       0.007      0.220    0.1 71 **
(0.009)               (0.057)                                                       (0.022)               (0.087)
385 Professional  Labor       0.39***     0.321     0.396***                        314 Tobacco       Labor       0.266***    0.202     0.322***
equipment                      (0.02)                (0.04)                                                       (0.053)                (0.08)
Energy      0.042***     0.019      0.017    0.898'                                 Energy      0.1 87*      0.009      0.018     0. 91 7'
(0.013)               (0.016)                                                       (0.035)               (0.039)
N. obs.           Materials   0.539**     0.423     0.41 8**                        Mobs              Materials   0.535***    0.516     0.389***
935                           (0.012)                (0.135)                        269                           (0.034)                (0.224)
Capital     0.091 ***    0.238      0.075                                           Capital     0.077***     0.274      0.188
(0.01 )               (0.09)                                                       (0.029)                (0.193)
Bootstmpped standaftd errors in parentheses. ~,~and  indicate significance at 1%I/, 5% and 10% levels, respectively. 'indicates that the hypothesis of constant retumns to scale cannot be
rejected. Average revenue shares are conLsidered and the average capital revenue share is obtained as one minus the average revenue shares of labor,energy andmrw materials.



Table 4
Impact of Lagged Tariffs on Plant Productivitya
Regressors                  OLS         OLS         OLS         Plant       Plant
Fixed Effects Fixed Effects
(1)         (2)         (3)         (4)         (5)
Nominal tariff 3-digit     0.177***   -0.095***   -0.092***   -0.051***   .-0.07***
(0.018)     (0.025)     (0.025)     (0.015)     (0.016)
Year effects                 Yes         Yes         Yes         Yes         Yes
Industry effects 3-digit                 Yes
Industry effects 4-digit                             Yes                     Yes
N. observations             57861       57861       57861       57861       57861
R-squared                   0.01        0.11        0.15        0.03         0.06
Regressors                  OLS         OLS         OLS         Plant       Plant
Fixed Effects Fixed Effects
(6)         (7)         (8)         (9)         (10)
Nominal tariff 4-digit     0.08***    -0.268***   -0.096 ***  -0.077***    -0.076***
(0.017)     (0.027)     (0.023)     (0.015)     (0.016)
Year effects                 Yes         Yes         Yes         Yes         Yes
Industry effects 3-digit                 Yes
Industry effects 4-digit                             Yes                     Yes
N. observations             54501       54501       54501       54501       54501
R-squared                   0.01         0.12        0.15        0.04        0.05
a The dependent variable is plant productivity. Robust standard errors are in parentheses. ***, ** and
* indicate significance at 1%, 5% and 10% levels, respectively. All regressions include a constant, plant
age and age squared. The coefficients on age are significant at the 1% level and range from 0.002 to
0.006. The coefficients on age squared range from -0.00001 to -0.0008. Sample years included are
1977, 1979, 1981, 1984-1989. One-period lagged tariff measures are used.
35



Table 5
Impact of Real Exchange Rates and Lagged Tariffs on Plant Productivitya
Regressor                 OLS         Plant        OLS         OLS        Plant        OLS         Plant       OLS     -   Plant
Fixed Effects                       Fixed Effects            Fixed Effects           Fixed Effects
(1)         (2)         (3)         (4)         (5)         (6)         (7)          (8)         (9)
Traded 3-digit Industries  -0.234***  -0.149***
(0.039)     (0.025)
Traded 4-digit Industries                       -0.225***   -0.179***    -0.184***
(0.04)     (0.042)      (0.025)
RER              .      -0.002***   -0.002***   -0.002***   -0.002***    -0.002***    -0.0002   -0.0004***   0.00006    -0.0003***
(0.0002)    (0.0001)    (0.0002)    (0.0002)    (0.0001)    (0.0002)     (0.0001)    (0.0002)    (0.0001)
RERxTradedInd.          0.001***     0.001***    0.001***    0.001***    0.001***
(0.0002)    (0.0001)    (0.0002)    (0.0002)    (0.0001)
Trend                   -0.019***   -0.024***   -0.019***    -0.018***   -0.025***    0.001      -0.009***     0.003     -0.01***
(0.002)     (0.001)     (0.002)     (0.002)     (0.001)     (0.003)      (0.002)     (0.003)     (0.002)
Trend x Traded Ind.     0.015***     0.014***    0.013***    0.01***     0.013***
(0.002)     (0.001)     (0.002)     (0.002)     (0.001)
Nominal Tariff 3-digit                                                              -0.134***    -0.12***
-  .   . (0.02)  (0.011)
Nominal Tariff 4-digit                                                                                       -0.234***   -0.116***
(0.015)     (0.011)
Industry effects 3-digit                                       Yes                     Yes                     Yes
N. observations  .        77423       77423       72651       72651       72651       45304       45304       42630        42630
R-squared                 0.002       0.012       0.003       0.361       0.012       0.157        0.005       0.166       0.005
' The dependent variable -is plant productivity. Robust standard errors are in parentheses. **, ** and * indicate significance at 1%, 5% and 10% levels, respectively.
All regressions include a constant. Plant age and age squared are included in the regressions in columns (6)-(9). An increase in RER represents a real appreciation of
the Colombian peso (IMF definition). In columns (1)-(2), interactions refer to 3-digit traded industries, in columns (3)-(5) they refer to 4-digit traded industries. In
columns (1)-(5), petroleum derivatives and iron and steel are excluded due to irregularities in the export data and the omitted category are nontraded industries.
Sample years included are 1980-1991 in columns (l)-(5) and 1981, 1984-1989 in columns (6)-(9). One-period lagged tariff measures are used.



Table 6
Irnpact of Lagged Effective Rates of Protection and Import Penetration Ratios on Plant Produc-
tiVitya
Regressors                       OLS         OLS         Plant       Plant
Fixed Effects Fixed Effects
(1)         (2)         (3)         (4)
ERP 34igit                       0.004       0.004      0.023***    0.03***
(0.014)     (0.013)    (0.009)     (0.009)
Year effects                      Yes         Yes         Yes         Yes
Industry effects 3-digit          Yes
Industry effects 4-digit                      Yes                     Yes
N. observations                  32456       32456       32456       32456
R-squared                         0;15        0.19       0.02        0.06
Regressors                        OLS         OLS        Plant       Plant
Fixed Effects Fixed Effects
(5)         (6)         (7)         (8) 
Import penetration ratio 34igit  1.797***   1.811***    0.601***    1.59***
(0.077)     (0.075)     (0.024)    (0.049)
Year effects                      Yes         Yes         Yes         Yes
Industry effects 3-digit          Yes
Industry effects 44igit                       Yes                     Yes
N. observations                  71928       71928       71928       71928
R-squared                         0.15        0.19        0.02        0.05
Regressors                        OLS         OLS        Plant       Plant
Fixed Effects Fixed Effects
(9)        (10)        (11)        (12)
Import penetration ratio 4-digit  -0.034*   0.668***    0.335***    0.634***
(0.018)     (0.052)     (0.019)     (0.033)
Year effects                      Yes         Yes         Yes         Yes
Industry effects 3-digit          Yes
Industry effects 4-digit                      Yes                     Yes
N. observations                  67686       67686       67686       67686
R-squared                         0.15        0.19        0.01        0.04
a The dependent variable is plant productivity. Robust standard errors are in parentheses. ***,
** and * indicate significance at 1%, 5% and 10% levels, respectively. All regressions include
a constant, plant age and age squared. Sample years included are 1980, 1984, 1985, 1990,
1991 in columns (1)-(4) and 1981-1991 in columns (5)-(12). One-period lagged ERP measures
and import penetration ratios are used.
37



Table 7
Impact of Lagged Tariffs on Plant Productivity Differentiated by Size'
Panel A Size Measured by Plant Employment
Regressor                    OLS         OLS         OLS         OLS         OLS         OLS
(1)         (2)         (3)         (4)         (5)          (6)
Small                     -0.163***     -0.02       -0.012    -0.122***    0.098***     -0.009
(0.007)     (0.017)     (0.024)     (0.007)      (0.011)     (0.023)
Large                      0.172***    0.331***    0.318***    0.189***    0.399***     0.292***
(0.014)     (0.019)     (0.027)     (0.014)      (0.016)     (0.026)
Nominal tariff 3-d. x Small  0.253***   -0.017      -0.013
(0.011)     (0.026)     (0.026)
Nominal tariff 3-d. x Large  -0.104***  -0.336***  -0.33***
(0.022)     (0.033)     (0.033)
Nominal tariff 4-d. x Small                                    0.154***    -0.171***     -0.025
(0.01)     (0.018)      (0.025)
Nominal tariff 4-d. x Large                                    -0.143***   -0.412***   -0.268***
(0.02)     (0.024)      (0.031)
Year effects Small Large     Yes         Yes         Yes         Yes          Yes         Yes
Industry effects 3-digit                  Yes                                 Yes
Industry effects 4-digit                             Yes                                  Yes
N. observations             57861       57861       57861        54501       54501       54501
R-squared                    0.04        0.15        0.18        0.04         0.15        0.18
Panel B Size Measured by Plant Market Share
Regressor                    OLS         OLS         OLS         OLS         OLS          OLS
(1)         (2)         (3)         (4)          (5)         (6)
Market share 3-digit       7.969***    9.968***    9.417***    7.333***      8.7***     7.84***
(0.54)     (0.503)      (0.47)      (0.439)     (0.392)      (0.404)
M. share squared          -11.06***    -12.32***   -11.13***   -10.41***   -11.14***    -9.92***
(0.834)     (0.959)     (0.865)      (0.789)     (0.911)     (0.855)
Nominal tariff 3-digit     0.193***   -0.084***    -0.08***
(0.01)     (0.026)      (0.026)
Nom.tariff3-d. x M.share  -3.625***    -5.51***    -5.622***
(0.97)     (0.901)      (0.831)
Nominal tariff 4-digit                                         0.106***    -0.218***    -0.091***
(0.01)     (0.017)      (0.024)
Nom.tariff4-d. x M.share                                       -2.636***   -3.172***    -2.378***
(0.841)     (0.792)     (0.802)
Year effects                 Yes         Yes         Yes          Yes         Yes         Yes
Industry effects 3-digit                 Yes                                  Yes
Industry effects 4-digit                             Yes                                  Yes
N. observations             57861       57861        57861       54501       54501       54501
R-squared                    0.04        0.15        0.18         0.04        0.16        0.18
Average M. share 3-digit    0.004        0.004       0.004       0.004       0.004        0.004
Average M. share squared    0.0004      0.0004      0.0004       0.0004      0.0004      0.0004
Average Nom. tariff 3-digit  0.464       0.464       0.464
Average Nom. tariff 4-digit                                      0.477       0.477        0.477
a The dependent variable is plant productivity. Robust standard errors are in parentheses. ***, ** and * indicate
significance at 1%, 5% and 10% levels, respectively. In Panel A, small plants have less than 50 employees in their first
year in the sample. In Panel B, market shares are relative to 3-digit industry output in the plant's first year in the sample.
At the bottom of each column, the regressors' sample averages used to calculate the corresponding marginal effects are
reported. Sample years included are 1977, 1979, 1981, 1984-1989. One-period lagged tariff measures are used.
38



Table 8
Impact of Lagged Effective Rates of Protection and Import Penetration Ratios on Plant Productivity Differentiated b3
Size'
Panel A Size Measured by Plant Ernployment
Regressor                  OLS         OLS        OLS        OLS         OLS        OLS
(1)        (2)        (3)         (4)        (5)         (6)
Small                     -0.003    -0.065***   -0.096***  -0.164***   -0.006     -0.126***
(0.017)    (0.026)     (0.009)    (0.018)     (0.008)    (0.018)
Large                    0.321***    0.232***     0.01     -0.056***   0.137***     0.006
(0.022)    (0.029)     (0.013)    (0.02)     (0.013)     (0.02)
ERP 3 digit x Small      0.058***    0.052***
(0.014)    (0.014)
ERP 3-digit xLarge        -0.14*    -0.121***
(0.017)    (0.017)
Import Penet.3-digit x Small                    1.516***    1.646***
(0.073)    (0.072)
Import Penet.3-digit x Large                    2.241 ***   2.207***
(0.074)     (0.072)
Import Penet.4-digit x Small                                           -0.149***   0.585***
(0.019)     (0.05)
Import Penet.4-digit x Large                                            0.14***    0.876***
(0.023)     (0.052)
Year effects Small Large    Yes        Yes        Yes         Yes        Yes         Yes
Industry effects 3-digit    Yes                   Yes                    Yes
Industry effects 4-digit               Yes                    Yes                    Yes
N. observations            32456      32456       71928      71928       67686      67686
R-squared                  0.19        0.23       0.19        0.23       0.18        0.22
Panel B Size Measured by Plant Market Share
Regressor                  OLS         OLS        OLS         OLS        OLS         OLS
(1)        (2)        (3)         (4)         (5)        (6)
Market share 3-digit      11.22***   10.27***   8.738***    8.021***   9.039***    8.194***
(0.527)    (0.515)     (0.219).   (0.216)     (0.222)    (0.223)
M. share squared         -23.29***   -20.72***  -20.8***   -18.460**   -20.5***    -18.13***
(1.376)    (1.297)     (0.904)    (0.847)     (0.873)    (0.846)
ERP 3-digit                0.009      0.009
(0.014)    (0.013)
ERP 3-d. x M.share       -1.757***   -1,529***
(0.501)    (0.481)
Import penet. 3-digit                           1.774***    1.789***
(0.072)     (0.071)
Imp.penet.3-d. x M.share                         4.49***    4.238***
(0.596)     (0.54)
Import penet. 4-digit                                                   0.011      0.672***
(0.018)     (0.049)
Imp.penet.4-d. x M.share                                                3.211***   3.144***
(0.594)     (0.563)
Yeareffects                 Yes        Yes         Yes        Yes         Yes        Yes
Industry effects 3-digit    Yes                    Yes                    Yes
Industry effects 4-digit               Yes                    Yes                    Yes
N. observations            32456      32456       71928      71928       67686      67686
R-squared                  0.19        0.23       0.19        0.23        0.19       0.22
Average M. share 3-digit   0.004      0.004       0.004      0.004       0.004       0.004
Average M. share squared  0.00036    0.00036     0.00037    0.00037     0.00037     0.00037
Average ERP 3-digit        0.871      0.871
Average Imp. penet. 3-digit                       0.123      0.123
Average Imp. penet. 4-digit                                              0.115       0.115
The dependent variable is plant productivity. Robust standard errors are in parentheses. *, *  and * indicate
significance at 1%, 5% and 10% levels, respectively. In Panel A, smaOl plants have less than 50 employees in their
first year in the sample. In Panel B, market shares are relative to 3-digit industry output in their first year in the
sample. At the bottom of each column in Panel B, the regressors' sample averages used to calculate the
corresponding marginal effects are reported. Sample years included are 1980, 1984, 1985, 1990, 1991 in Panels A
and B, columns (l)-(2) and 1981-1991 in Panels A and B, columns (3)-(6). One-period lagged ERP measures and
import penetration ratios are used.
39



Table 9
Impact of Lagged Tariffi on Plant Productivity Differentiated by Degree of Domestic Competitiona
Regressor                        OLS         OLS        OLS         OLS         OLS         OLS         OLS        OLS
(1)         (2)         (3)        (4)         (5)         (6)         (7)         (8)
Herfindahl Index 3-digit      -0.786***   -0.598***
(0.102)     (0.101)
Nominal tariff 3-digit         0.114***                0.105***   -0.115***                          -0.799***
(0.013)                (0.012)     (0.027)                             (0.042)
Nom.tariff3-d. x Herfindahl 3-d.  -0.781***
(0.259)
Nominal tariff 4-digit                     0.06***                             0.012     -0.313***               -0.939***
(0.013)                            (0.013)      (0.02)                 (0.038)
Nom.tariff4-d. x Herfindahl 3-d.          -1.441***
(0.242)
Herfindahl Index 4-d.                                 -0.282***   -0.218***   -0.339***   -0.227***
(0.054)    (0.063)     (0.065)     (0.075)
Nom.tariff3-d. x Herfindahl 4-d.                        0.161     0.532***
(0.119)    (0.144)
Nom.tariff4-d. x Herfindahl 4-d.                                               -0.131     0.406***
(0.129)    (0.142)
Turnover rate 3-digit                                                                                -1.115***   -1.127***
(0.067)     (0.068)
Nominal tariff 3-d. x Turnover 3-d.                                                                   3.122***
(0.143)
Nominal tariff 4-d. x Turnover 3-d.                                                                               3.363***
(0.135)
Year effects                     Yes         Yes         Yes        Yes         Yes         Yes         Yes         Yes
Industry effects 3-digit                                            Yes                     Yes
N. observations                 57861       54501       57861      57861       54501       54501       57861       54501
R-squared                        0.01        0.02       0.01        0.34        0.01        0.12        0.02        0.02
Average Herf. Index 3-digit     0.045       0.045
Average Herf. Index 4-digit                             0.088       0.088       0.082      0.082
Average Turn. rate 3-digit                                                                             25.66       25.78
Average Nom. tariff 3-digit     0.464                   0.464       0.464                              0.464
Average Nom. tariff 4-digit        I        0.477                               0.477      0.477                   0.477
a The dependent variable is plant productivity. Robust standard errors are in parentheses. *  ** and * indicate significance at 1%, 5% and 10% levels,
respectively. Herfindahl indexes for 1977 are used, defined as the sum of squared market shares of plants relative to industry output. Turnover rates for
1977-1978 are used, defined as the sum of entry and exit rates into 3-digit industries. At the bottom of each column, the regressors' sample averages used
to calculate the corresponding marginal effects are reported. Sample years included are 1977, 1979, 1981, 1984-1989. One-period lagged tariff measures
are used.



Table 10 Impact of Lagged Effective Rates of Protection and Import Penetration on Plant Productivity Differentiated by Degree of Domestic Competition
Regre§sor                   OLS       OLS        OLS       OLS       OLS        OLS       OLS       OLS        OLS       OLS       OLS        OLS
(1)       (2)       (3)       (4)        (5)       (6)        (7)       (8)       (9)       (10)      (11)       (12)
Herfindahl Index 3-digit  -1.346***                                -1.532***  -2.003**
(0.147)                                  (0.071)   (0.064)
ERP3-digit                0.066***   0.138***   0.011   -0.095***
(0.009)   (0.008)   (0.014)    (0.018)
ERP3d x Herf.3-d.         0.748***
(0.187)
Import penet. 3-digit                                              -0.036**             -0.05**   1.633***                       0.949*0*
(0.015)              (0.016)   (0.081)                        (0.033)
Imp.penet.3-d. x Herf.3-d.                                         2.239***
(0.2 17)
Import penet. 4-digit                                                        -0.165***                       -0.15***  -0.094***            0.6720*
(0.012)                        (0.014)   (0.025)              (0.029)
Imp.penet 4-d. x Herf.3-d.                                                    3.538***
(0.178)
Herfindahl lndex 4-digit              0.09     0.14***                                  -0.449***  -0.115***  -0.693+**  -0.135***
(0.055)   (0.06)                                   (0.022)    (0.024)   (0.031)    (0.036)
ERP3d xHerf.4-d.                    -0.348***  -0.137***
(0.06)   (0.063)
Imp.penet.3-d. x Herf.4-d.                                                              1.014***   1.117e*
(0.122)   (0.178)
Imp.penet.44 x Herf.4-d.                                                                                     1.32***   0.465***
(0.109)   (0.122)
Turnover 3-digit                                        -0.239***                                                                 1.077***  0.916***
(0.077)                                                                 (0.025)    (0.024)
ERP3-d. x Tun.3-digit                                    0.841**
(0.079)
Imp.penet.3-d. x Turn.3-d.                                                                                                       -5.156 **
(0.211)
Imp.penet.4-d. x Turn.3-d.                                                                                                                  4.226***
(0.186)
Year effects                Yes       Yes        Yes       Yes       Yes        Yes       Yes        Yes       Yes       Yes        Yes       Yes
Industry effects 3-digit                         Yes                                                 Yes                  Yes
N. observations             32456     32456     32456      32456     71928     67686      71928     71928     67686      67686     71928     67686
R-squared                   0.02       0.02      0.15      0.02      0.02       0.02      0.01      0.15       0.01      0.15       0.03      0.02
Average Herf. Index 3-digit  0.044                                   0.043     0.043
Average Herf. Index 4-digit           0.084     0.084                                     0.085     0.085     0.077      0.077
Average Turn, rate 3 -digit                                0.198                                                                   0.224     0.227
Average ERP 3-digit         0.871     0.871     0.871      0.871
Average Imp. penet. 3-digit                                          0.123                0.123     0.123                          0.123
Average Imp. penet. 4-digit                                                    0.115                          0.115      0.115                0.115
*The dependent variable is plant productivity. Robust standard errors are inparentheses. *  , ** and * indicate significance at 1%, 5% and 10% levels, respectively. Herfindahl indexes for
1980 are used in columns (1)-(4) and for 1981 in columns (5)-(12). Turnover rates for 1980-1981 are used in columns (1)-(4) and for 1981-1982 in columns (5)-(12). At the bottom of each
column, the regressors' sample averages used to calculate the corresponding marginal effects are reported. Sample years included are 1980, 1984, 1985, 1990, 1991 in column (1)-(4) and
1981-1991 in columns (5)-(12). One-period lagged ERP measures and import penetrationratios are used.



.19.
2                                           2~~~~~~~~~~~ .13-
S                                                         T-.07-
o*  .3-
E                                           0~~~~~~~~~~~~~~~~~
z                                          ~~~~~~~~~~~~~z
OLS coeff.                                                OLS coeff.
a. Labor                                                 b. Energy
.3-
.6  -D
*                                            ~~~~~~~~~~~~~~~~~~~~~~~~~~~.2
E
CL 4
5 .35                                                        05 -o
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _   .~~~ ~ ~~~~~01  *
A~ .4               I     ~         .b                      i   .6    . 
C.RWOLS coeff.                                                OLS coeff.
c.RwMaterials                                             d. Capital
Fig. 1 Production Function Coefficients (each dot is an industry)



Policy Research Working Paper Series
Contact
Title                           Author                  Date             for paper
WPS3043 The Incentive-Compatible Design  Thorsten Beck           May 2003          K. Labrie
of Deposit Insurance and Bank                                             31001
Failure Resolution: Concepts and
Country Studies
WPS3044 Impregnated Nets Cannot Fully    Mead over               May 2003          H. Sladovich
Substitute for DDT: Field       Bernard Bakote'e                          37698
Effectiveness of Malaria Prevention  Raman Velayudhan
in Solomon Islands              Peter Wilikai
Patricia M. Graves
WPS3045 Causes and Consequences of Civil  Klaus Deininger        May 2003          M. Fernandez
Strife: Micro-Level Evidence from                                         33766
Uganda
WPS3046 Migration, Spillovers, and Trade  Ross Levine            May 2003          E. Khine
Diversion: The Impact of        Sergio L. Schmukler                       37471
Internationalization on Stock Market
Liquidity
WPS3047 Comparing Mortgage Credit Risk    Robert Buckley         May 2003          0. Himid
Policies: An Options-Based Approach Gulmira Karaguishiyeva                80225
Robert Van Order
Laura Vecvagare
WPS3048 Targeted Transfers in Poor Countries: Martin Ravallion   May 2003          P. Sader
Revisiting the Tradeoffs and Policy                                       33902
Options
WPS3049 Hidden Impact? Ex-Post Evaluation  Shaohua Chen          May 2003          P. Sader
of an Anti-Poverty Program      Martin Ravallion                          33902
WPS3050 The Changing Financial Landscape:  Wafik Grais           May 2003          R. Vo
Opportunities and Challenges for  Zeynep Kantur                           33722
the Middle East and North Africa
WPS3051 Regional, Multilateral, and Unilateral  Glenn W. Harrison  May 2003        P. Flewitt
Trade Policies of MERCOSUR for  Thomas F. Rutherford                      32724
Growth and Poverty Reduction in  David G. Tarr
Brazil                          Angelo Gurgel
WPS3052 Long-Run Impacts of China's WTO   Kym Anderson           May 2003          S. Lipscomb
Accession on Farm-Nonfarm Income  Jikun Huang                             87266
Inequality and Rural Poverty    Elena lanchovichina
WPS3053 Economic Impacts of China's       Elena lanchovichina    May 2003          S. Lipscomb
Accession to the World Trade    William Martin                            87266
Organization
WPS3054 The New Comparative Economics     Simeon Djankov         May 2003          G. Sorensen
Edward Glaeser                            37088
Rafael La Porta
Florencio Lopez-de-Silanes
Andrei Shleifer
WPS3055 The Global Growth of Mutual Funds  Deepthi Fernando      May 2003          A. Yaptenco
Leora Klapper                             31823
Viktor Sulla
Dimitri Vittas



Policy Research Working Paper Series
Contact
Title                           Author                  Date             for paper
WPS3056 Avoiding the Piffalls in Taxing   Patrick Honohan        May 2003          A. Yaptenco
Financial Intermediation                                                  31823
WPS3057 Teaching Adults to Read Better and  Helen Abadzi         May 2003          H. Abadzi
Faster: Results from an Experiment                                        80375
in Burkina Faso
WPS3058 Working for God? Evaluating Service Ritva Reinikka       May 2003          H. Sladovich
Delivery of Religious Not-for-Profit  Jakob Svensson                      37698
Health Care Providers in Uganda
WPS3059 Do Capital Flows Respond to Risk  Cesar Calder6n         May 2003          T. Tourougui
and Return?                     Norman Loayza                             87431
Luis Serven
WPS3060 World Market Integration through the  Rui Albuquerque    May 2003          T. Tourougui
Lens of Foreign Direct Investors  Norman Loayza                           87431
Luis Serven
WPS3061 Reciprocity in Free Trade         Caroline Freund        May 2003           P. Flewitt
Agreements                                                                32724
WPS3062 Labor Effects of Adult Mortality in  Kathleen Beegle     May 2003          E. de Castro
Tanzanian Households                                                      89121
WPS3063 Trade Liberalization, Firm        Paolo Epifani          May 2003          P. Flewitt
Performance, and Labor Market                                             32724
Outcomes in the Developing World:
What Can We Learn from Micro-Level
Data?