1 TRADE STATUS, TRADE POLICY AND PRODUCTVITY: THE BRAZILIAN CASE X. Cirera 1 , D. Lederman 2 , J.A. Mañez 3 , M.E. Rochina 3 and J. A. Sanchis 3 1 IDS, University of Sussex 2 Economics Research World Bank Group 3 University of Valencia and ERICES 4-september-2012 Abstract. The literature on firm level productivity recognizes the important role played by firm trade status and trade policy on the evolution of firm productivity. There are many recent studies that have highlighted the importance of considering trading status (either exporting or importing) as well as the effects of trade policy in the analysis of total factor productivity. The aim of this paper is to integrate both firms’ trade status and trade policy in the analysis of productivity in Brazil. We use a two-step strategy: first, we estimate TFP following De Loecker (2010) approach and Wooldridge (2009) estimation procedure; and, second, we use this estimated TFP as the dependent variable in a model with trade policy and firm trade status as covariates, in order to disentangle the effects of those variables on TFP. From our results we can conclude that trade liberalisation (lower input and/or output tariffs) increases productivity. We also find that decreasing input tariffs has a larger effect increasing productivity for high import intensity firms and that decreasing output tariffs increases productivity for exporters more than for non- exporters. Finally, even after controlling for the effects of tariffs, there is still evidence of both learning-by-exporting and learning-by-importing effects on productivity. Key words: trade status, trade policy, Total Factor Productivity, GMM
42
Embed
TRADE STATUS, TRADE POLICY AND … 1 TRADE STATUS, TRADE POLICY AND PRODUCTVITY: THE BRAZILIAN CASE X. Cirera1, D. Lederman2, J.A. Mañez3, M.E. Rochina3 and J. A. Sanchis3 1IDS, University
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
! 1
TRADE STATUS, TRADE POLICY AND PRODUCTVITY: THE BRAZILIAN CASE X. Cirera1, D. Lederman2, J.A. Mañez3, M.E. Rochina3 and J. A. Sanchis3
1 IDS, University of Sussex
2 Economics Research World Bank Group
3 University of Valencia and ERICES
4-september-2012
Abstract. The literature on firm level productivity recognizes the important role played by firm trade
status and trade policy on the evolution of firm productivity. There are many recent studies that have highlighted the importance of considering trading status (either exporting or importing) as well as the effects of trade policy in the analysis of total factor productivity. The aim of this paper is to integrate both firms’ trade status and trade policy in the analysis of productivity in Brazil. We use a two-step strategy: first, we estimate TFP following De Loecker (2010) approach and Wooldridge (2009) estimation procedure; and, second, we use this estimated TFP as the dependent variable in a model with trade policy and firm trade status as covariates, in order to disentangle the effects of those variables on TFP. From our results we can conclude that trade liberalisation (lower input and/or output tariffs) increases productivity. We also find that decreasing input tariffs has a larger effect increasing productivity for high import intensity firms and that decreasing output tariffs increases productivity for exporters more than for non-exporters. Finally, even after controlling for the effects of tariffs, there is still evidence of both learning-by-exporting and learning-by-importing effects on productivity. Key words: trade status, trade policy, Total Factor Productivity, GMM
! 2
1. Introduction.
The literature on firm productivity recognizes the important role played by firm trade status and
trade policy on the evolution of firm productivity. We find recent studies such as De Loecker
(2007, 2010), Van Biesebroeck (2005) or Kasahara and Rodrigue (2008) that have highlighted
the importance of considering trading status in the analysis of total factor productivity (TFP,
hereafter). However, whereas De Loecker (2007, 2010), De Loecker and Warzyniski (2011) or
Van Biesebroeck (2005) only consider the role of exporting, the study by Kasahara and Rodrigue
(2008) only analyses the role of importing. In relation to the effects of trade policy on firm TFP, in
recent years, Fernandes (2007) or Amiti and Konings (2007) have carried sound studies on the
impact of trade policy (proxied by tariffs) on productivity for two developing countries, such as
Colombia and Indonesia, respectively. Both studies find that the impact of tariffs reductions on
productivity is large. And the extent trade policy is expected to affect firm’s level productivity
critically depends on the size of trade policy changes.
The aim of this paper is to integrate both firms’ trade status and trade policy in the
analysis of productivity in Brazil. Further, we aim to consider jointly the role of exporting and
importing to check if there is learning-by-exporting and learning-by-importing.
Our empirical strategy consists of two steps. In the first step, we estimate TFP following
De Loecker (2010) approach and Wooldridge (2009) estimation procedure. In the second step,
we use this estimated TFP as dependent variable of a model with trade policy and trade status as
covariates in order to disentangle the effects of those variables on TFP.
The TFP estimation procedure used in this study presents various novelties with respect
to the typical control-function based estimation methods (Olley and Pakes, 1996; Levinshon and
Petrin, 2003) used to analyse the effects of trade policy. First, we allow for different demands of
intermediate materials according to firm trade status (non-traders, only exporters, only importers
! 3
and two-way traders). Second, we move from an exogenous law of motion of productivity to an
endogenous law of motion in which we allow past trading experience to affect productivity
(following De Loecker, 2007, 2010).
In the second step of our estimation strategy, similarly to Amiti and Konings (2007), we
regress our TFP estimate against trade policy measures (input and output tariffs, trade status
variables and their interactions). Our aim is to analyse the impact of input and output tariffs on
firm productivity and whether it depends on the firm’s trading status.
In order to analyse the relationship between firm productivity and both trade status and
trade policy, we use a Brazilian dataset that links firm’s characteristics, production and export
data for Brazilian firms for the period 2000 to 2008. It is important to note that while Brazil had
undergone an intense period of trade liberalization during the 1980s and 1990s, this process has
slowed down during the 2000s, where trade policy has been quite stable. In general, tariffs were
reduced very slowly until 2007, and increased in 2008.
To anticipate our results, we find that both higher output tariffs (tariffs on imports of final
goods) and higher input tariffs (tariffs on imports of intermediate inputs) decrease productivity.
Higher output tariffs decrease productivity by lowering import competition, as firms are less forced
to improve efficiency. Higher input tariffs decrease productivity by decreasing, for instance,
access to a wider range of foreign inputs, to higher quality inputs, or to foreign technology
incorporated in imported inputs (Bustos, 2011). We also find that trade liberalization, by
decreasing input tariffs, has a larger effect increasing productivity for high import intensity firms;
and, that trade liberalization, by decreasing output tariffs, increases productivity for exporters
more than for non-exporters. Finally, even after controlling for the effects of tariffs, there is still
evidence of both learning-by-exporting and learning-by-importing effects on productivity.
! 4
The rest of the paper is organized as follows. Section 2 summarises the related literature.
Section 3 explains the main features of the two-step method pursued and the production function
estimation method. Section 4 describes the data. In section 5 we discuss the results and some
Most of the relevant literature that analyses the relationship between productivity and trade status
and trade policy focuses separately either on the impact of trade status on productivity or the
effects of trade policy on productivity. We start reviewing first the most relevant papers that either
analyse the effects of trade status on productivity or the effects of trade policy on productivity, and
then we review those studies that jointly study the effects of both trade policy and trade status on
productivity.
2.1. Trade status and productivity.
Whereas there is a large amount of papers that have analysed whether exporting improves firm
productivity (learning-by-exporting hypothesis, LBE hereafter), the evidence on the analysis of the
impact of importing on productivity (learning-by-importing hypothesis, LBI) is much more scarce.
According to the LBE mechanism firms improve their productivity after entering a foreign
market (Clerides et al., 1998). These potential productivity gains for firms, from participating into
export markets arise from (among others): growth in sales that allows firms to profit from
economies of scale, knowledge flows from international customers that provide information about
innovations reducing costs and improving quality, or from increased competition in export markets
that force firms to behave more efficiently. In spite of the amount of studies analysing this
hypothesis, evidence on LBE is far from conclusive, whereas there are papers that do not find
! 5
any evidence on LBE, those that find it differ both on the intensity and the duration of the LBE
effect.1 However, as De Loecker (2010) has recently shown, most previous tests on the existence
of the LBE mechanism could be flawed. The usual empirical strategy is to look at whether a
productivity estimate, typically obtained as the residual of a production function estimation,
increases after firms enter in the export market. But for such an estimate to make sense, past
export experience should be allowed to impact future productivity. Yet some previous studies
(implicitly) assume that the productivity term in the production function specification is just an
idiosyncratic shock (Wagner, 2002; Hansson and Lundin, 2004; Greenaway and Kneller, 2004,
2007b, 2008; Girma et al., 2004; Máñez et al., 2010), while others assume that this term is
governed by an exogenous Markov process (Arnold and Hussinger, 2005; Serti and Tomassi,
2008). It is this sort of assumptions, often critical to obtain consistent estimates (Ackerberg et al.,
2006), what make these tests of the existence of LBE to lack internal consistency. To the best of
our knowledge, only recent papers by De Loecker (2007, 2010), De Loecker and Warzyniski
(2011) and Manjón et al. (2013) allow past export experience to impact future productivity.
Similarly, the papers testing for LBI hypothesize that the diffusion and adoption of new
technologies by importing intermediates can be an important source of productivity
improvements, especially in developing countries.2 Among them, Kasahara and Rodrigue (2008)
test for LBI allowing past import experience to affect productivity for Chilean manufacturing
plants.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 Silva et al. (2010) provide a detailed survey of the learning by exporting literature. Further, Martins and Yang (2009) provide a meta-analysis of 33 empirical studies. Singh (2010) concludes that studies supporting self-selection overwhelm studies supporting learning-by-exporting. 2 Previous empirical studies using aggregate country or industry-level data found that importing intermediate goods that embody R&D from an industrial country can boost a country’s productivity, see for example Coe and Helpman, 1995, and Coe et al., 1997.
! 6
2.2. Trade policy and productivity.
Most of the traditional studies analysing the effect of trade liberalization on productivity have
focused on output tariffs, and most of them found that a reduction in output tariffs increases
productivity due to the increase in import competition. Thus, Treffler (2004) for the US and
Canada, using highly disaggregated tariff data, found that labour productivity gains amounted up
to 14% for those industries with the largest tariff cuts. In the same line, Pavnick (2002) for Chile
encounters that trade liberalization induced up to 10% higher gains for import competing
industries than for industries non exposed to foreign imports.3
However, these studies do not account for the possible role of input tariffs. Furthermore,
the theoretical underpinnings of these papers are the traditional theoretical models based on
economies of scale by Krugman (1979) and Helpman and Krugman (1985). In these models,
exposure to foreign competition increases the elasticity of demand faced by domestic producers,
reducing market power and forcing firms to move down along their average cost curve. But gains
from trade liberalization could also accrue from reallocation effects in favour of the more efficient
firms that would gain market share with the consequent increase in average productivity (Roberts
and Tybout, 1991). Additionally, according to Amiti and Konings other potential gains could arise
from externalities from: technical innovation, managerial efforts or domestic spillovers and
learning-by-exporting.
Among the few theoretical papers analysing the relationship between the reduction of
tariff inputs and productivity, one can find models that support a positive impact of tariff inputs on
productivity and models that suggest a negative relationship. According to Corden (1971) lower
input tariffs results in lower effective protection what softens import competition and could lead to
lower productivity. However, models by Ethier (1982), Markusen (1989) and Grossman and
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!3!Other relevant works on output tariffs and productivity with a lower level of disaggregation are Tybout, Melo and Corbo (1991), Levinsohn (1993), Harrison (1994), Tybout ad Westwrook (1995), Gaston and Treffler (1997), Krishna and Mitra (1998), Head and Ries (1999) or Topalova (2004).
! 7
Helpman (1991) show that tariff reductions on inputs could result in higher productivity through al
least three channels: i) availability of a larger variety of imported inputs; ii) access to higher
quality inputs; and, iii) learning effects from the foreign technology embodied in the imported
inputs. In the same vein, lower input tariffs reduce the price of international outsourcing of
material inputs, and Amiti and Konings (2007) results suggest that international outsourcing is
associated with higher TFP.
Two papers closely related to ours that analyse the effects of trade policy on firms’
productivity accounting both for output and input tariffs are Fernandes (2007) and Schor (2004),
For Fernandes (2007), the general argument linking the reduction of tariffs to productivity is that
trade liberalization results in wider exposure to foreign competition what forces domestic firms to
behave more efficiently. Fernandes (2007) widens the production function to include trade policy
as and additional input (which therefore implies allowing trade policy to shift the mean of the
production function). Further, the demand for intermediate inputs in her production function also
depends on trade policy, and so it remains in the inversion rule for productivity. However, she
maintains the assumption of an exogenous Markov process for the law of motion of productivity.
The results from her study strongly support the presence of competitive pressure-induced TFP
gains due to trade liberalization for Colombian plants. However, it is important to note that
Fernandes (2003) uses alternatively 3-digit industry output tariffs or the effective rate of
protection, and thus she is unable to separately identify the effect of output and input tariffs.
Schor (2004) estimates the effects of nominal output and input tariffs on productivity
using Brazilian firm data for the period 1986-1998 (which corresponds to the period previous to
the one we use in this study), and she finds that both types of tariffs have a negative effect on
productivity. She uses a two-step methodology: in the first step, she obtains an estimate for TFP
using a standard Levinshon and Petrin (2003, LP hereafter) methodology that does not include
! 8
trade policy measures neither as additional inputs in the production function, nor in the demand
for intermediate materials, nor in the Markov process; and, in a second step, the estimated TFP is
regressed on tariffs. Schor (2004) finds that the estimated coefficient for output tariffs in the
productivity equation turns out to be negative. Further, when a measure of tariffs on inputs is
added in the productivity equation, the coefficient associated with this measure is also negative,
and the inclusion of this new variable reduces the size of the estimated coefficient of nominal
(output) tariffs. Thus, her results seem to indicate that, along with the increased competition, the
new access to inputs that embody better foreign technology also contributes to productivity gains
after trade liberalization.
2.3. Trade policy, trade status and productivity.
Finally, among the papers that analyse jointly the effects of trade policy and trade status on
productivity are worth mentioning Muendler (2004) and Amiti and Konings (2007). Muendler
(2004) uses also data from Brazil but for the period 1986-1998 (as Schor, 2004). The empirical
strategy used consists of two steps: in the first step, he introduces as additional inputs in the
production function the foreign shares of capital and intermediate inputs to measure the impact of
differences in quality between national and foreign inputs; in the second step, the growth of TFP
estimated in the first step is regressed on import penetration4 (as a proxy to control for non-tariff
barriers), output tariffs and the share of capital and intermediate inputs. His empirical results
suggest that the use of foreign inputs plays a negligible role to explain productivity changes,
whereas foreign competition (measured by larger import penetration and lower output tariffs)
pressures firms to raise productivity.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4 Import penetration seemed to be very important in Brazil during the period analysed by Schor (2004).
! 9
Also Amiti and Konings (2007) use a two-step procedure to analyse the effect of trade
policy on Indonesian firms productivity for the period 1991-2001. In the first step, following the
approach proposed by Kasahara and Rodrigue (2005), Van Biesebroeck (2005) or De Loecker
(2007), they modify the Olley and Pakes (1996, OP hereafter) two-stage estimation setup to
account for different firms trade status in the demand of investment. This implies that the
investment demand function becomes a function of four variables, the standard capital and
productivity variables plus the export and import decisions. However, they do not incorporate firm
past trading experience into the law of motion of productivity (they do not modify the assumption
of an exogenous Markov process for the law of motion of productivity), and therefore they do not
allow past trading experience to affect current productivity. In the second step, they regress the
estimated TFP on trade policy variables (output and input tariffs) and make some robustness
checks about the possibility of input tariffs affecting more to input importers. However, they do not
check whether output tariffs affect more intensely to exporting than non-exporting firms. The
results by Amiti and Konings (2007) show that the effect of reducing input tariffs significantly
increases productivity, and that this effect is much higher than reducing output tariffs.
!
3. Methodology.
3.1. Some features of our two-step method.
As in Amiti and Konings (2007) and Muendler (2004), our empirical strategy consists of two steps.
In the first step, we estimate TFP using a procedure that introduces some novelties with respect
to the papers reviewed above. In the second step, we use this estimated TFP as dependent
variable in a regression model with trade policy and trade status as covariates.
In this two-step analysis it is crucial to decide whether to include the trade status and
trade policy variables in the TFP estimation and/or as covariates in the equation explaining the
! 10
estimated TFP. Different authors opt for different solutions. Therefore, we devote the following
paragraphs to carry out a detailed reasoning of our choices.
First, following De Loecker (2007) to capture differences in market structure among firms
with different trading status (related to the mode of competition, demand conditions, and exit
barriers), we allow for different demands of materials according to firm trading status (non-
traders, only importers, only exporters and two-way traders). Thus, we do not include firm trade
status as other inputs in the production function but only in the demand for intermediate inputs.5
According to De Loecker (2007, 2010), there are at least three reasons that advice to do so: i)
including a vector of trading status dummies as inputs (only exporters, only importers, two-way
traders dummies) implies to assume, for example, that the impact of importing in productivity is
deterministic, i.e. the productivity of all importers will increase by the estimate of the import
dummy (the same applies to only exporters or two-way traders dummies); ii) including trading
status variables in the demand for intermediate inputs allows trading status to have a different
impact for firms with different characteristics (capital, intermediate materials, etc.); and, iii) the
Cobb-Douglas production function implies that if we include trading status variables as additional
inputs, a firm can substitute any input with being an exporter/importer with a unit elasticity of
substitution.6
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!5 According to Van Biesebroeck (2005) and Kasahara and Rodrigue (2005) the export and import dummies should be treated also as state variables, in particular, as non-deterministic state variables. However, De Loecker (2007, 2010) and De Loecker and Warzyinski (2011), argue that these variables affect the characteristics of the markets where firms operate. Notwithstanding, from an empirical point of view both alternatives conduct to the same end: these type of variables should be included in the investment/inputs demand function. In fact, De Loecker and Warzyinski (2011) explain that regardless of considering export status as a state variable of the underlying dynamic problem of the firm or not, it has to be controlled for in the input demand function if exporters face different demand conditions. However, they recognize that it is reasonable to think that export status is a state variable, given the empirical evidence about export entry decisions facing significant sunk costs (Roberts and Tyboy, 1997, and Mañez et al., 2008). For us, the same types of arguments also work for the import status. 6 However, Van Biesebroeck (2005) introduces the export status as an additional input into the production function. This treatment can be methodologically justified by modelling log productivity, in a previously linearized production function, as with two components: one observable depending on the export status with its corresponding coefficient and another unobserved component of productivity. Under this methodological approach if the estimated coefficient on the variable export status in the production function is greater than zero it can be interpreted like a shifter out of the production frontier.
! 11
Second, we move from an exogenous law of motion of productivity to an endogenous law
of motion in which we allow past trading experience to affect productivity (following De Loecker,
2007, 2010, for export status or Kasahara and Rodrigue, 2008, for import status).7 An exogenous
Markov process is only appropriate when productivity shocks are exogenous to the firm but not if
future productivity is determined endogenously by firm choices, such as firm export and import
decisions. Therefore, those methods that do not use an endogenous Markov process suffer from
an internal inconsistency as do not accommodate endogenous productivity processes like LBE
and LBI. The same arguments are put forward in De Loecker and Warzynski (2011).
Finally, in the same vein than Amiti and Konings (2007), we do not include trade policy
variables (tariffs) either as additional inputs in the production function or in the demand for
intermediate materials, or in the law of motion of productivity. The reasons explaining this
decision are that: i) including them as additional inputs would imply the same problems explained
above for trade status variables; ii) the demand of intermediate inputs should include only firm’s
state variables (as capital) and other factors related to demand conditions, market conditions or
affecting firms’ input choices (as far as they have firm variation); and, iii) trade policy variables are
not firm level decisions that endogenously determine the evolution of productivity and, therefore,
they should not be included in the law of motion of productivity.
With respect to the technique used to estimate the TFP, we follow Wooldridge (2009) that
argues that both Olley and Pakes (1996) and Levinshon and Petrin (2003) two step estimation
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!7 Kasahara and Rodrigue (2008), analogously to Van Biesebroeck (2005) for exports, introduce a dummy variable for importers in the production function. They introduce this variable not because they consider it as an additional input, but because they consider that productivity depends both on a productivity shock (unobserved) and on the range of intermediate inputs available to firms. They expect that having access to a larger range of intermediate inputs will probably increase productivity. Thus they make the assumption that importing firms have access to more variety of intermediate inputs than domestic firms (assuming that domestic firms have only access to domestic intermediate inputs). But they also include the import dummy in the intermediate inputs demand function and in the law of motion for productivity, the latter because it is also considered a firm level decision. They interpret the coefficient associated to the import dummy in the production function as a static/immediate effect on productivity and, differently, the coefficient linked to the import dummy in the law of motion for productivity as measuring something dynamic (such as learning-by-importing).
! 12
procedures can be reconsidered as consisting of two equations that can be jointly estimated by
GMM in a one step estimation procedure. This joint estimation strategy has the advantages of
increasing efficiency with respect to two-step procedures, of making unnecessary bootstrapping
for the calculus of the standard errors, and solving the labour coefficient identification problem
posed by Ackelberg et al. (2006). Therefore, our estimation technique represents a step further
with respect to both De Loecker (2010) and Amiti and Konings (2007).
In the second step of our estimation strategy, similarly to Amiti and Konings (2007), we
regress our TFP estimate against trade policy measures (input and output tariffs), trade status
and their interactions. Our aim is to analyse the impact of inputs and output tariffs on firm
productivity and to analyse whether they depend on the firm’s trading status.
3.2. Productivity estimation.
We assume that firms produce a homogeneous good using a Cobb-Douglas technology:
y it = β0+β l l it +βk kit +βmmit + µt +ω it +ηit (1)
where yit is the natural log of production of firm i at time t, lit is the natural log of labour, kit is the
natural log of capital, mit is the natural log of intermediate materials, and µt are time effects. As for
the unobservables, ωit is the productivity (not observed by the econometrician but observable or
predictable by firms) and ηit is a standard i.i.d. error term that is neither observed nor predictable
by the firm.
It is also assumed that capital evolves following a certain law of motion that is not directly
related to current productivity shocks (i.e. it is a state variable), whereas labour and intermediate
! 13
materials are inputs that can be adjusted whenever the firm faces a productivity shock (i.e. they
are variable factors).8
Under these assumptions, OP show how to obtain consistent estimates of the production
function coefficients using a semiparametric procedure; see also LP for a closely related
estimation strategy. However, here we follow Wooldridge (2009), who argues that both OP and
LP estimation methods can be reconsidered as consisting of two equations which can be jointly
estimated by GMM: the first equation tackles the problem of endogeneity of the non-dynamic
inputs (that is, the variable factors); and, the second equation deals with the issue of the law of
motion of productivity. Next we consider each in detail.
Let us start considering first the problem of endogeneity of the non-dynamic inputs.
Correlation between labour and intermediate inputs with productivity complicates the estimation
of equation (1), because it makes the OLS estimator biased and the fixed-effects and
instrumental variables methods generally unreliable (Ackerberg et al., 2006). Both OP and LP
methods use a control function approach to solve this problem, by using investment in capital and
materials, respectively, to proxy for “unobserved” firm productivity.
In particular, the OP method assumes that the demand for investment in capital,
i it = i kit ,ω it( ) , is a function of firms’ capital and productivity. To circumvent the problem of firms
with zero investment in capital, the LP method uses the demand for materials, mit = m kit ,ω it( ) ,
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!8 The law of motion for capital follows a deterministic dynamic process according to which
kit = (1−δ )kit−1+ Iit−1 .
Where Iit-1 is investment in capital for firm i in period t-1. Thus, it is assumed that the capital the firm uses in period t was actually decided in period t-1 (it takes a full production period for the capital to be ordered, received and installed by the firm before it becomes operative). Labour and materials (unlike capital) are chosen in period t, the period they actually get used (and, therefore, they can be a function of ). These timing assumptions make them non-dynamic
inputs, in the sense that (and again unlike capital) current choices for them have no impact on future choices.
ωit
! 14
instead, as a proxy variable to recover “unobserved” firm productivity. Since we follow this last
approach, we concentrate on the demand of materials hereafter.9
Therefore, when estimating productivity using these general versions of OP and LP in a
sample where some firms do not participate in foreign markets and others participate either
exporting, importing or both, it is assumed that the demand of intermediate materials for the
different types of firms according to trade status is identical. However, as explained in the
previous section, heterogeneity in trade status may influence the demand function of intermediate
inputs. Therefore, analogously to De Loecker (2007, 2010) when analysing the effects of
exporting on firms’ productivity, and Kasahara and Rodrigue (2008) when analysing the effects of
importing on firms’ productivity, we consider different demands of intermediate materials for non-
traders, only exporters, only input importers and both exporters and input importers (two way
traders). In this sense, we extend De Loecker (2010) by introducing firms’ choices related to
imports of intermediate inputs (as in Kasahara and Rodrigue, 2008 and Amiti and Konings, 2007).
Thus, we write the demand of materials as:
mit = mTS kit ,ω it( ) (2)
where we include the subscript TS (trade status) to denote different demands of intermediate
inputs for only exporters, only input importers or both, and non-traders. Also, since the demand of
intermediate materials is assumed to be monotonic in productivity, it can be inverted to generate
the following inverse demand function for materials:
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!9 Both the investment of capital demand function and the demand for intermediate materials are assumed to be strictly increasing in ωit (in the case of the investment of capital this is assumed in the region in which iit>0). That is, conditional on kit, a firm with higher ωit optimally invests more (or demands more materials).
! 15
ω it = hTS kit ,mit( ) (3)
where hTS is an unknown function of kit and mit. Then, substituting the above expression (3) into
the production function (1) we get:
y it = β0+β l l it +βk kit +βmmit + µt + hTS kit ,mit( )+ηit (4)
Finally, by considering four different demand functions for intermediate materials (for non-
traders, only exporters, only input importers, two-way traders), our first estimation equation
where 1(NT), 1(E), 1(I) and 1(EI) are indicator functions that take value one for non-traders (NT) ,
only exporters (E), only input importers (I) and two way traders (EI), respectively. Further, the
unknown functions H in (5) are proxied by second degree polynomials in their respective
arguments.
With the specification in equation 5, the difference in the inverse demand function of firms
with different trading status arises not only from differences in the coefficients of kit and mit but
also by the fact that each inverse demand function includes a dummy variable capturing the
corresponding trading status. This is not equivalent to introduce the set of trade status dummies
as additional inputs in the production function, as each one of these dummies is interacted with all
! 16
the terms , it itk m of its corresponding polynomial. De Loecker (2007, 2010) underlines that
introducing an export dummy as an input in the production function will cause at least two
problems. First, an identification problem, as we will need another estimation step to identify the
parameter associated to that variable. Second, introducing an export dummy in the Cobb-
Douglas production function implies that a firm can substitute any input with being an exporter at
constant unit elasticity.
Notice, however, that we cannot identify βk and βm from (5). This is achieved by the
inclusion of a second estimation equation in the GMM-system that deals with the law of motion of
productivity.
The standard OP/LP approaches consider that productivity evolves according to an
exogenous Markov process:
ω it = E ω it ω it−1
⎡⎣ ⎤⎦+ξ it = f ω it−1( )+ξ it (6)
where f is an unknown function that relates productivity in t with productivity in t-1 and ξit is an
innovation term uncorrelated by definition with kit. However, this assumption neglects the
possibility of previous trading experience to affect productivity. Consequently, here we consider a
more general (endogenous Markov) process in which previous trading experience can influence
the dynamics of productivity:
ω it = E ω it ω it−1
,Eit−1,Iit−1
,EIit−1⎡⎣ ⎤⎦+ξ it = f ω it−1
,Eit−1,Iit−1
,EIit−1( )+ξ it (7)
! 17
where Eit-1, Iit-1 and EIit-1 indicate whether the firm, in period t-1, choses to only export, to only
import inputs, or both to export and import inputs, respectively. Obviously, the reference category
is to be a non-trader.
Let us now rewrite the production function (1) using (7) as:
y it = β0+β l l it +βk kit +βmmit + µt + f ω it−1
,Eit−1,Iit−1
,EIit−1( )+ξ it +ηit (8)
Further, since ω it = hTS kit ,mit( ) , we can rewrite
f ω it−1,Eit−1
,Iit−1,EIit−1( ) as:
f ω it−1,Eit−1
,Iit−1,EIit−1( ) = f hTS kit−1
,mit−1( ),Eit−1,Iit−1
,EIit−1⎡⎣ ⎤⎦ = FTS kit−1
,mit−1( )=1(NT )FNT kit−1
,mit−1( )+1(E)FE kit−1,mit−1
,Eit−1( )+1(I)FI kit−1,mit−1
,Iit−1( )+1(EI)FEI kit−1
,mit−1,EIit−1( )
(9)
with F being unknown functions to be proxied by second degree polynomials in their respective
arguments.
Lastly, substituting (9) into (8), our second estimation equation is given by:
y it = β0+β l l it +βk kit +βmmit + µt +
1(NT )FNT kit−1,mit−1( )+1(E)FE kit−1
,mit−1,Eit−1( )+1(I)FI kit−1
,mit−1,Iit−1( )+
1(EI)FEI kit−1,mit−1
,EIit−1( )+uit
(10)
where uit=ξit+ηit is a composed error term.
Wooldridge (2009) proposes to estimate jointly equations (5) and (10) by GMM using the
appropriate instruments and moment conditions for each equation. This joint estimation strategy
has several advantages: i) it increases efficiency relatively to the two step traditional procedures;
ii) it makes unnecessary to do bootstrapping for the calculus of standard errors; and, iii) it solves
! 18
the problem, pointed out by Ackerberg et al. (2006), of identification of the labour coefficient in the
separate estimation of equation (5). This procedure allows us to obtain, per each one of the 22
industries considered, both the coefficient estimates of the production function and firms’
productivity estimates.
4. Data and descriptive analysis.
In order to analyse firm productivity and trade exposure we use a dataset that links firm
characteristics, production and export data for Brazilian firms for the period 2000 to 2008. For
production and firm characteristics, we use the PIA empresa (Pesquisa Industrial Anual). PIA is a
survey for manufacturing and mining sectors conducted annually by the IBGE (Instituto Brasileiro
de Geografia e Estatistica), which focus on firms characteristics. Firms with 30 or more
employees are included in the sample, while smaller firms of up to 29 workers are included
randomly in the sample. In total PIA covers more than 40,000 firms.
For exports we use a dataset created by SECEX (Secretaria Comercio Exterior). SECEX
provides the universe of registered trade flows at the firm level, by HS-8 product and market
destination for the period 2000-2008. The dataset uses aggregates export FOB values per year,
product and destination. We complement the dataset with tariffs included in the TRAINS
database (TRAINS is a database maintained by the UNCTAD with data on tariff, imports, para-
tariffs and non-tariff measures at national level).
Table A.1 in the Appendix shows the main variables used in the analysis. We proxy
capital with assets and also include electricity and energy as intermediate inputs. We use sector
deflators provided by the IBGE to deflate the variables in the production function, with the
exception of labour. In order to calculate tariffs for inputs we first calculate the average tariff for
each of the Brazilian Input-Output sectors and, then, for each sector we use the input-output
! 19
coefficients to weight the sector tariff for those sectors that provide inputs. These input tariffs are
then mapped from Input-Output sectors to CNAE 4 digits sectors using correspondence tables
supplied by the IBGE. For the period 2000 to 2004 we use the 2000 I-O table and for the 2005 to
2008 period we use the 2005 I-O table. Both tables are available from the IBGE national
accounts. Changes in yearly tariffs and the change in I-O tables create variation on tariffs in
inputs.
Regarding tariffs on outputs, each firm is associated to a 4 digits CNAE sector based on
its main sector of production. We first convert HS-8 trade codes with tariffs to the Prodlist code
equivalent (product extension of CNAE classification) using IBGE conversion table. Then we
average the tariff for Prodlist products for each CNAE 4 digits sector.10 Finally, since we do not
have information regarding value added, we calculate the effective rate of protection (ERP) as the
difference between tariffs on outputs and inputs. 11
It is important to note that while Brazil has undergone an intense period of trade
liberalization during the 1980s and 1990s, this process slowed down during the 2000s, where
protection levels remained quite stable. Final good tariffs fell from an average of 17% to an
average of 15.34%, and input tariffs slightly increased from an average of 8.38% to 9.25% (see
Figure 1). However, these figures hide two periods: average tariffs were reduced very slowly until
2007, and increased in 2008. Up to 2007, both input and output average tariffs decreased, but the
decrease in average output tariffs was much higher (3.36%) than the one of input tariffs (0.625%).
The 2008 tariffs upturn reversed the decreasing trend in average input tariffs; as a result they are
1.42% higher in 2008 than in 2000. It also smoothed the decrease in average output tariffs, thus
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!10 We have also calculated an alternative tariff for outputs based on the products produced by each firm. The problem with this alternative measure is the fact that this information is only available using the survey PIA produto, and prior to 2005 only large firms entered the sample. Therefore, there are many missing values for smaller firms, which are mainly firms not engaging in trade and competing with imports. 11 For the estimations we use log transformations of tariff variables, i.e. by taking the log of (1 + tariff) in order to keep zero values for both tariffs and ERPs.
! 20
in 2008 they are 1.66% lower than in 2000. It is also noteworthy to stand out that average output
tariffs all along the period are higher than average input tariffs. Further, this is true for every
industry of the sample (Table 1)
We observe more variation in average input and output tariffs between industries than
over time (see Table 1). Thus, the lowest average input tariff over the period of analysis (4.55%)
corresponds to industry 26 (Non-metallic Mineral Product Manufacturing) and the highest one
(13.37%) to industry 18 (Apparel Manufacturing). As for the output tariffs, the lowest one (5.34%)
corresponds to industry 10-14 (Extractive industries) and the highest one (22.55%) again to
industry 18. In general, final goods tariffs show for all industries a decreasing trend from 2000 to
2007. Between 2007 and 2008, we observe an upturn of output tariffs for 10 out of the 22 sectors
(from 17 to 23 and 25, 28 and 36).
Evidence on the evolution of input tariffs is more mixed, 14 out of the 22 industries show
a decreasing trend from 2000 to 2007, and as a consequence of the generalised increase in
tariffs from 2007 to 2008, we observe an increase in tariffs for 16 industries.
Table 2 reports the main features of our data set. As can be observed, two-way traders
(both exporters and importers) are larger in terms of output, labour, capital and materials and pay
higher wages as compared to one-way traders (either exporters or importers) and to non-traders.
One-way traders are, in general, more similar in all variables. If we compare these firms with no
traders we find that are larger in terms of output, labour, capital and materials and pay higher
wages.
As regards to trade variables, we find that export intensity is larger for only exporters (as
compared to two-way traders) and that import intensity is larger for only importers (also as
compared to two-way traders).
! 21
5. Results: the effects of trade policy and firm trade status on firm productivity.
In this section we first present the main results from our analysis and then we discuss some
robustness checks we have carried out.
5.1. Main results.
In the first step of our analysis, using the methodology explained above, we estimate the
production function (1) separately for each of the 22 industries, in order to obtain estimates of the
log of TFP.12 Using these estimates, we calculate the (log) of TFP for firm i at time t and industry
s, denoted , as
tfpits = y it −β0
−β l l it −βk kit −βmmit − µt (11)
It is important to note that including a vector of time dummies ( ) in the TFP estimation
makes the estimated TFP time effects free. Controlling for time effects in this setup is crucial as
we are interested in disentangling the effects of trade policy from other possible changes in
macroeconomic policy or macroeconomic instability, or even from any other uncontrolled events,
that occurred in Brazil during our sample period.
In the second step, we use our TFP estimates as the dependent variable of a series of
reduced form equations that include as covariates either trade policy variables or both trade
policy variables and trade status variables. There are two reasons that advise to include trade
status also in the second step estimation: on the one hand, we expect the dynamic evolution of
productivity to be affected by LBI and LBE; and, on the other hand, we believe that the effects of
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!12 The coefficients estimated at industry level are reported in Table 3.
tfpits
µt
! 22
input and output tariffs on the evolution of firms productivity may depend on whether the firm
imports inputs and/or exports, respectively.
In this second step regression analysis we pool TFP estimates for all industries and use
panel data fixed effects estimation to simultaneously control for individual firm and industry fixed
effects.13,14 Using firm level fixed effects allows us to control for the existence of a self-selection
mechanism, that would arise if only the (a priori) more efficient firms were the ones getting
involved in international markets either as buyers, sellers or both buyers and sellers. This self-
selection process is based on the existence of higher sunk entry costs in international markets
that can only be overcome by the more productive firms (see for example Bernard and Jensen,
1999, and Melitz, 2003). The results for these firm fixed effects estimations are reported in Table
4.15
We start our analysis of the effects of trade policy and trade status by using the simplest
possible specification (see equation 12 below), where the only covariate that we include to
explain productivity is output tariffs (TO). This specification (specification 1) has been widely used
in the literature on trade liberalization and productivity.
tfpit =α +α i +γ 1TO +uit (12)
where α is a constant term and αi is and individual fixed effect.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!13 We report robust standard errors by clustering at the firm level. Clustering at the industry level gives similar results. 14 Controlling for industry fixed effects, among other things, allows to account for time-invariant characteristics coming from trade policy that could make the country policy related to tariffs endogenous with respect to productivity (due to possible policy pressure from particular industries). 15 We have estimated the same set of reduced form equations linking TFP to trade policy and trade status, using a random effects approach (these results are reported in Table A.2 in the Appendix). The fact that the random effects estimates for the export and import status variables are higher than the fixed effects ones suggests that the random effects estimates suffer from an endogeneity bias problem associated to the existence of self-selection by the more productive firms, that could introduce and upwards bias in the estimation of both LBI and LBE. Further, this bias problem is larger for the import dummy than for the export dummy.
! 23
In this specification we expect to be negative. Trade liberalization policies, implying a
reduction of output tariff on imports (of final goods), may increase competitive pressure from
competing imported products and so force firms to use inputs more efficiently, and, consequently,
this would increase productivity. As the dependent variable is the log of TFP, the effect of a unit
increase in the output tariffs on TFP is computed from the estimated coefficient as
100 exp γ
1( )−1( ) . This measure shows the percentage change on the TFP when the tariff on
output increases by one unit. The estimate of γ1 (see Table 4) shows that, as expected, a
decrease in output tariff increases productivity. More specifically, a unit decrease in output tariff
increases TFP by 0.84%.
In our second specification (specification 2), we add as additional covariates both a
dummy that takes value one if the firm exports and zero otherwise (DE), and an interaction that
results from multiplying DE by the output tariff ( TO ⋅DE ). The aim of the first of these variables is to
capture whether there is a direct effect of exporting on productivity. The role of the second one is
to test whether the effect of output tariffs on productivity is different for exporters and non-
exporters.
tfpit =α +α i +γ 1TO +γ 2
TO ⋅DE +γ 3DE +uit (13)
Our results for specification 2 (see second column of Table 4) suggest that a unit
decrease in output tariffs increases productivity by 0.72% for non-exporters and by 0.94% for
exporters (we get that both γ1 and γ2 are negative). These results mean that trade liberalization
(in the form of reducing tariffs on imports of equivalent competing products) will have a larger
effect in the productivity of exporters than that of non-exporters. This can be the result of two
γ 1
γ 1
! 24
effects that work in opposite direction: on the one hand, the positive effect of a reduction in output
tariffs on productivity operates tightening competition and forcing both exporting and non-
exporting firms to behave more efficiently; on the other hand, if trade liberalization reduces
market shares of domestic firms, its impact could be larger in the market shares of the less
productive non-exporting firms (Cirera et al., 2012 show that the self-selection mechanism fully
works for Brazilian manufacturing firms) and this could lessen their incentives to increase
productivity. Additionally, the transformed estimate 100 exp γ
3( )−1( ) shows that the direct effect of
exporting (average difference in TFP between exporters and non-exporters) is of 9.33%. Once we
control by firm level fixed effects, this can be interpreted as evidence of LBE.
Traditional studies on the effects of exports have only considered the effect of firm’s
export status on productivity. The natural evolution of this literature has been to incorporate also
the role of imported inputs. Including an import variable in our analysis could allow disentangling
the effect of exporting on productivity from the effect of importing. Importing inputs may produce
efficiency improvements through the availability both of a wider range of inputs and of inputs of
superior quality for importing firms. Further, closely related to the decision of importing we aim to
analyse the role of import tariffs on productivity. We expect an increase in input tariffs to have a
negative impact on productivity as the increase in the price of the inputs could reduce the range
and quality available for domestic producers.
Thus, to start the analysis of imported inputs, in specification 3 we widen our baseline
specification 1 to include as covariates both output (TO) and input tariffs (TI):
tfpit =α +α i +γ 1TO +γ 2
TI +uit (14)
! 25
The negative sign of the estimate of γ2 in specification 3 confirms that, as expected, a
decrease in inputs tariffs has a positive effect on productivity. More specifically, a unit reduction in
input tariff increases TFP by about 0.50%. As for the effects of output tariffs on productivity, its
correspondent estimate maintains its negative sign. However, once we introduce in the analysis
input tariffs the effect of a unit reduction in output tariffs is lower: whereas in the specification
without input tariffs a unit reduction in the output tariffs increases TFP by 0.84%; when we
consider simultaneously both input and output tariffs this increase is 0.73%.
Finally, in specification 4 we widen specification 2 to take into account both the direct
effect of importing inputs on productivity and whether or not the effect of input tariffs differs
depending on whether the firm import inputs. Thus, we expect a lower impact of changes of input
tariffs for firms that do not import inputs. Therefore, in addition to the covariates already included
in specification 2, we include a dummy that takes value one if the firm imports and zero otherwise
(DI), input tariffs and an interaction that results from multiplying DI by the input tariffs variable.
Therefore, this specification allows us to analyse whether the effects of trading policy (proxied by
inputs and output tariffs) depend on the trade status of the firm,
tfpit =α +α i +γ 1TO +γ 2
TO ⋅DE +γ 3DE +γ 4
TI +γ 5TI ⋅DI +γ 6
DI +uit (15)
As for the new covariates included in specification 4 (with respect to specification 2), our
estimates for TI and TIDI suggest that a unit decrease in input tariffs increases productivity by
0.45% both for importers and non-importers. Additionally, the direct effect of importing measured
by the average difference in productivity between importers and non-importers (given by
100 exp γ
6( )−1( ) ) is 8.87%; i.e. importing inputs increases firm productivity by 8.87%, and so
provides evidence in favour of LBI. As for the exporting and output tariff related covariates (DE, TO
! 26
and TO ⋅DE ), the estimates of a unit decrease in output tariffs are lower in specification 4 than in
specification 2, both for exporters and non-exporters. When we account for input tariffs and
whether the firm imports inputs (specification 4), a unit decrease in output tariff increases TFP by
0.60% for non-exporters and by 0.82% for exporters. However, these figures are higher when we
do not account for them (specification 2), as they are 0.72% and 0.94%. The direct effect of
exporting (that can be interpreted as a measure of LBE) also gets reduced in specification 4 in
comparison to specification 2 (8.97% and 9.33%, respectively).
The larger size of the estimates obtained for output tariffs and export status in
specification 2 could be due to an omitted variable bias, produced by omitting other relevant
factors affecting TFP such as inputs tariffs and import status. The fact that when we introduce
these variables in specification 4 their coefficients are sizeable and significant confirms the
suspects of an omitted variable bias in specification 2.
5.2. Some further robustness specifications.
In this section we present some robustness tests for the previous specifications in which we have
analysed the relationship between trade policy, trade status and productivity. The first of these
robustness specifications (specification 5) is based on specification 4 and simply substitutes the
export and import dummies by export and import intensity variables, respectively.
tfpit =α +α i +γ 1TO +γ 2
TO ⋅EI +γ3EI +γ
4TI +γ 5
TI ⋅II +γ 6⋅II +uit (16)
where EE and II stand for export intensity and import intensity, respectively. The most relevant
differences between the estimates of specifications 4 and 5 are as follows: first, the impact of
! 27
output tariffs on productivity is independent of the export intensity of the firm16 (whereas in
specification 4, where we only distinguished between exporters and non-exporters, the effect was
higher for exporting firms); and, second, the negative and significant estimate of the variable
suggests that the higher the import intensity of a firm the higher the impact of changes in input
tariffs. We also find that non-importers are also positively affected by a reduction in input tariffs.
This result suggests that reducing input tariffs affect the overall quality of domestic inputs though
greater competition for inputs in the domestic market.
In our second robustness specification (specification 6), we proxy trade policy by the
effective rate of protection (ERP, hereafter) instead of proxying it using input and output tariffs.
The aim of this second robustness specification is to test which are the drivers of the relationship
between trade policy and productivity in those papers that only include the ERP as a measure of
trade policy and do not distinguish between input and output tariffs.
The traditional literature linking trade liberalization and productivity has used the ERP as
the unique measure of trade policy. In this literature, a decrease in input tariffs increases the ERP
and would result in a decrease in productivity via a reduction in the intensity of competition
among national firms. However, the most recent literature on trade liberalization and productivity
suggests using both input and output tariffs to measure trade policy. Within this approach the
opposite argument arises relating input tariffs and productivity. According to this argument, a
decrease in input tariffs eases productivity increases by domestic firms as it allows them to profit
from: the learning effect derived from the use of the incorporated technology in imported inputs,
and from the wider range and quality of the inputs available to domestic firms.
In specification 6, we include as covariates not only the ERP but also the exporter and
importer dummies, and the crossed products between these and the ERP:
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!16 The estimate of TO•EI is not significant at any reasonable significance level.
TI ⋅II
! 28
tfpit =α +α i +γ 1ERP +γ
2ERP ⋅DE +γ 3
DE +γ 4ERP ⋅DI +γ 5
DI +uit (17)
The results obtained in the estimation of specification 6 (equation 17) can be summarized as
follows. In comparison with specification 4 (that includes both input and output tariffs instead of
ERP) the direct effect of exporting is lower (7.09% vs. 8.97%) and the direct effect of importing is
higher (10.12% vs. 8.87%). Second, a unit increase in the ERP decreases productivity by 33.10%
for firms that neither export nor import, by 47.53% for exporters and by 56.00% for importers.
Therefore, an increase in the ERP, which softens competition in the domestic market, results in a
reduction of firms’ productivity, independently of its trading status. However, from the estimates
associated to the ERP we cannot disentangle whether the effects come from an increase in
output tariffs, from a decrease in input tariffs or from changes in the share of intermediate inputs
in the value of the final good.
Our third robustness exercise (specification 7) checks the effects of the large currency
appreciation that Brazil experienced during the period of analysis, as these can affect productivity
without implying changes in efficiency. To interpret the results in this specification we have to take
into account that an increase in the real effective exchange rate (ER, hereafter) means a
depreciation of the national currency. In specification 7 (see expression 18), we extend
specification 4 to include as covariates the cross products of the ER with the export and import
dummies. The aim of including these cross products is to check whether the ER evolution has
different effects on the productivity of importers and exporters:
tfpit =α +α i +γ 1TO +γ 2
TO ⋅DE +γ 3DE +γ 4
TI +γ 5TI ⋅DI +γ 6
DI +
γ7ER ⋅DE +γ 8
ER ⋅DI +uit
(18)
! 29
Changes in the estimates corresponding to the output and/or input tariffs would suggest
that some of the productivity improvement attributed to a reduction in output and/or input tariffs in
specification 4 could be due to exchange rate changes. Further, the coefficients of the importer
and exporter status dummies could also be affected.
As expected (see column 7 of Table 4) the inclusion of the ER and its interactions with
the export and input dummies reduces the size (in absolute terms) of the estimates
corresponding to the output and input tariffs. Thus, whereas in the specification without ER
(specification 4) a unit reduction of output tariffs increases the productivity of non-exporters and
exporters by 0.60% and 0.82%, respectively, in the specification with the ER variables
(specification 7), the increase in productivity gets reduced to 0.45% both for exporters and non-
exporters. However, there is almost no difference between the effects on productivity of a unit
reduction of input tariffs both for importers and non-importers (0.46% in specification 7 vs. 0.45 in
specification 4).
Notice that the extra increase in productivity enjoyed by exporters (in comparison with
non-exporters) in specification 4 when the output tariffs decrease vanishes with the inclusion, in
specification 7, of the variable interacting ER with the export dummy. This finding suggests that
the evolution of the ER has special incidence in the evolution of the productivity of exporters since
firms need to become more productive due to the competitiveness loss in international markets
due to exchange rate appreciation. Therefore, omitting this variable can lead to overestimating
the effect of export status in specification 4.
However, both the direct effects of exporting and importing in productivity are higher in
the specification including the ER variables than in the specification that does not include them,
confirming the existence of both LBE and LBI processes. Thus, the export premium is 10.20% in
! 30
specification 7 in comparison to 8.97% in specification 4. In the same vein, the import premium is
12.52% and 8.87% in specifications 7 and 4, respectively. Finally, the two interaction variables of
the ER with the importer and exporter dummies are negative and significant (a unit decrease in
ER increases productivity by 6.47% and 5.48 for importers and exporters, respectivey). This
could be signalling that a real appreciation increases firm productivity. In addition, to increasing
competitive pressureon exporters, a real appreciation lowers imported input prices what
increases competition in the inputs markets and so has a positive effect on productivity.
6. Conclusions.
The results from all specifications led us to conclude the following. First, higher output tariffs
(tariffs on imports of final goods) decrease productivity by lowering import competition as firms
are less forced to improve efficiency.
Second, higher input tariffs (tariffs on imports of intermediate inputs) decrease
productivity by reducing access to a wider range of foreign inputs, to higher quality inputs, or to
foreign technology incorporated in imported inputs. Therefore, we do not find for input tariffs the
link with productivity predicted by the literature linking a trade policy measure such as the ERP
with productivity, but just the opposite. According to this literature a decrease in input tariffs
increases the ERP and decreases productivity, through the reduction in industry competition.
Third, we do not generally find that trade liberalization (in the form of reducing input
tariffs) has a larger effect increasing productivity for importing firms, except in that specification in
which we interact tariffs with import intensity.
Fourth, for the effects of output tariffs on productivity, for exporters and non-exporters we
find only statistically significant different results coming from the export status, but not from the
export intensity.
! 31
Fifth, our results indicate that the effects of tariffs in the economy do spread among all
firms in the economy, and do not only affect exporting or importing firms.
Sixth, we still find evidence of both learning-by-exporting and learning-by-importing
effects on productivity. This evidence comes by the fact that we get significant effects from the
firm importing and exporting status even after controlling for the effects of tariffs.
Seven, according to a trade policy measure such as the ERP we also confirm that an
increase of it, interpreted as a decrease in competition, produces a reduction on productivity.
However, we prefer specifications including separately output and input tariffs to be able to isolate
the effect of competition on productivity from the effect of better access to inputs on productivity.
Finally, from the more complete specification, specification 7, where results over
productivity for other variables are cleaned from the effect of the evolution of exchange rates over
the analysed period, we obtain that the effects of increasing output tariffs on decreasing
productivity are quite similar to the ones coming from increasing input tariffs, but that learning-by-
importing (as captured by the import status dummy) is larger than learning-by-exporting (as
captured by the export status dummy). Further, we also obtain that real appreciations of the
currency produce an increase in productivity, being importers more affected than exporters.
! 32
REFERENCES.
Ackerberg, D. A., K. Caves and G. Frazer (2006), Structural identification of production
functions, Working Paper, Department of Economics, UCLA.
Amiti, M. and J. Konings (2007), Trade Liberalization, Intermediate Inputs, and
Productivity: Evidence from Indonesia, American Economic Review, 97, 5, 1611-1638.
Arnold, J. and K. Hussinger (2005), Export Behavior and Firm Productivity in German
Manufacturing: A Firm-level Analysis, Review of World Economics ⁄ Weltwirtschaftliches Archiv,
141, 2, 219–43.
Bernard, A. B. and J. B. Jensen (1999), Exceptional Exporter Performance: Cause,
Effect, or Both? Journal of International Economics, 47(1), 1–25.
Bustos, P. (2011), Trade liberalizations, exports, and the technology upgrading:
evidence on the impact of MERCOSUR on Argentinian firms, American Economic Review, 101,
304-340.
Cirera, X., D. Lederman, J.A. Mañez, M.E. Rochina and J.A. Sanchis (2012), Self-
selection and learning-by-exporting: the Brazilian case. University of Valencia, mimeo.
Corden, Max W (1971). The Theory of Protection. Oxford: Oxford University Press.
Clerides, S. K., S. Lach and J.R. Tybout (1998), Is Learning by Exporting Important?
Micro-Dynamic Evidence from Colombia, Mexico, and Morocco, Quarterly journal of Economics,
113(2), 903–947.
Coe, D.T. and E. Helpman (1995), International R&D spillovers. European Economic
Review, 39, 859–887.
Coe, D.T., E. Helpman and A. Hoffmaister (1997), North–South R&D spillovers.
Economic Journal, 107, 134–149.
De Loecker, J. (2007), Do Exports Generate Higher Productivity? Evidence from
Slovenia. Journal of International Economics, 73, 1, 69–98.
De Loecker, J. (2010), A Note on Detecting Learning by Exporting, NBER Working
Papers 16548, National Bureau of Economic Research, Inc.
De Loecker, J. and F. Warzyniski (2011), Markups and firm-level status, NBER Working
Papers 15198, National Bureau of Economic Research, Inc.
Ethier, W. (1982), National and International Returns to Scale in the Modern Theory of
International Trade.” American Economic Review, 72(3), 389–405.
! 33
Fernandes, A.M. (2007), Trade policy, trade volumes and plant-level productivity in
Colombian manufacturing industries, Journal of International Economics 71, 52–71
Girma, S., D. Greenaway and R. Kneller (2004), Does Exporting Increase Productivity?
A Microeconometric Analysis of Matched Firms. Review of International Economics, 12, 5, 855–
66.
Gaston, N. and D. Trefler (1997), “The Labour Market Consequences of the Canada-
U.S. Free Trade Agreement.” Canadian Journal of Economics, 30(1) 18–41.
Greenaway, D. and R. Kneller (2004), Exporting and Productivity in the UK. Oxford
Review of Economic Policy, 20, 3, 358–71.
Greenaway, D. and R. Kneller (2007b), Industry Differences in the Effect of Export
Market Entry: Learning by Exporting? Review of World Economics ⁄ Weltwirtschaftliches Archiv,
143, 3, 416–32.
Greenaway, D. and R. Kneller (2008), Exporting, Productivity and Agglomeration.
European Economic Review, 52, 5, 919–39.
Grossman, G. M. and E. Helpman (1991). Innovation and Growth in the Global
Economy. Cambridge, MA: MIT Press.
Hansson, P. and N. Lundin (2004), Exports as Indicator on or a Promoter of Successful
Swedish Manufacturing Firms in the 1990s. Review of World Economics ⁄ Weltwirtschaftliches
Archiv, 140, 3, 415–45.
Harrison, A. E. (1994), Productivity, Imperfect Competition and Trade Reform: Theory
and Evidence. Journal of International Economics, 36(1-2), 53–73.
Head, C. K. and J. Ries. 1999. Rationalization Effects of Tariff Reductions. Journal of
International Economics, 47(2), 295–320.
Helpman, Elhanan, and Paul R. Krugman (1985), Market Structure and Foreign Trade:
Increas-ing Returns, Imperfect Competition, and the International Economy. Cambridge, MA: MIT
Press.
Kasahara, H. and J. Rodrigue (2005). Does the Use of Imported Intermediates Increase
Productivity? Plant-Level Evidence," University of Western Ontario, Economic Policy Research
! 34
Institute Working Papers 20057, University of Western Ontario, Economic Policy Research
Institute.
Kasahara, H. and J. Rodrigue (2008), Does the use of imported intermediates increase
productivity? Plant-level evidence, Journal of Development Economics 87, 106–118.
Krishna, P. and D. Mitra (1998), “Trade Liberalization, Market Discipline and
Productivity Growth: New Evidence from India.” Journal of Development Economics, 56(2), 447–
62.
Krugman, P. R. (1979), Increasing Returns, Monopolistic Competition, and International
Trade. Journal of International Economics, 9(4), 469–79.
Levinsohn, J. (1993), “Testing the Imports- as-Market-Discipline Hypothesis.” Journal of
International Economics, 35(1–2): 1–22.
Levinsohn, J. and A. Petrin (2003), Estimating production functions using inputs to
control for unobservables. Review of Economic Studies 70, 317–342.
Manjón, M., J.A. Máñez, M.E. Rochina-Barrachina and J.A. Sanchis-Llopis (2013),
Reconsidering learning by exporting. Review of World Economics, forthcoming.
Mánez-Castillejo, J.A., M.E. Rochina-Barrachina and J.A. Sanchis-Llopis (2010), Does
firm size affect self-selction and learning-by-exporting? The World Economy, 33 (3), 315-346.
Markusen, J. R. (1989), Trade in Producer Services and in Other Specialized Interme-
diate Inputs, American Economic Review, 79(1), 85–95.
Melitz, M. (2003), The impact of trade on intra-industry reallocations and aggregate
industry productivity. Econometrica 71 (4), 1695–1725.
Muendler, M. (2004), Trade, technology, and productivity: a study of Brazilian
manufacturers, 1986-1998. UCSD, mimeo.
Olley, G. S. and A. Pakes (1996), The dynamics of productivity in the
Note: Standard errors are in parentheses; ***, ** and * mean significance at the 1, 5 and 10% level, respectively.
! 41
APPENDIX
Table A.1. Variables description Production function variables
Output Gross output deflated 2.20E+07 2.31E+07
3.53E+06
Labour Number of employees
Capital Value of assets deflated
Materials Intermediate inputs including electricity and energy deflated
Wages Total wages deflated Trade variables
Export intensity Share of sales outside domestic market - -
Import intensity Share of inputs that are imported
Trade policy variables Effective rate of protection Difference between tariffs on outputs and inputs
Inputs tariffs Average tariff at CNAE 4 digits sector using Input-Output tables Output tariffs Average tariff at CNAE 4 digits sector Real effective exchange rate Average real effective exchange rate at CNAE 4 digits sector
! 42
Table A.2. TFP non-fixed effects regressions on trade policy and trade exposure variables. RE: