The World Economy

The World Economy (2014)

doi: 10.1111/twec.12141

Protection and Performance

Joseph Francois1 and Miriam Manchin2
1
University of Bern and CEPR, London, UK and 2University College London, UK

1. INTRODUCTION

I N a classic paper, Lerner (1936) demonstrated that under perfect competition, full employ-
ment, balanced trade and in the absence of transport costs, the imposition of import tariffs has
the same effect as an export tax. The intuition behind Lerner’s proposition is that higher protection
shifts home demand towards home goods, which makes less supply available for sale to export
partners. The general equilibrium correspondence (or even an exact symmetric correspondence
under appropriate conditions) between import policy (i.e. tariffs) and export policy now stands as
a classic result of general equilibrium trade theory.1 Despite this correspondence, the recent litera-
ture on policy and patterns of trade focuses on the policy of importers. This is especially true in
the now voluminous literature on the gravity equation. In this literature, bilateral import protec-
tion affects trade directly, while the general level of protection of individual countries also plays a
role. Indeed, general levels of import protection in third countries can also deflect trade to alterna-
tive markets. The critical point is the emphasis on policy in importing markets.
In contrast, in this paper, we emphasise import protection by exporters. To do this, we bridge
both the concepts of multilateral resistance from recent gravity literature to the macroconcept of
Lerner symmetry. Although largely ignored by the recent literature, we show that the role of pol-
icy in exporting markets in explaining general levels of trade should be comparable to policy in
importing markets. In particular, we extend the classical, analytical mapping of aggregate, mac-
roeconomic trade volume effects that follow from Lerner-type mechanisms to more recent con-
cepts from the empirical trade literature linked to bilateral gravity models of trade. This is
particularly relevant for the recent empirical literature on trade, policy, and trade cost estimation
as specified using this class of models. Working with a panel of global and bilateral trade data
spanning over 15 years, we find evidence at both the aggregate level, and also at the bilateral
level, that the import policies implemented by exporters are a significant macroeconomic factor
in explaining overall – and hence also bilateral – export performance. These results reinforce the
growing body of recent evidence on the importance of economic environmental (policy and infra-
structure) conditions in developing countries in explaining their relative export performance.

Miriam Manchin would like to acknowledge the financial support received from the European Commis-
sion under the Marie Curie Intra-European Fellowship Programme. Particular thanks are due to Jeff
Bergstrand, Peter Neary, Doug Nelson, and Alan Woodland for constructive comments and encourage-
ment, as well as Will Martin, Peter Egger, Eddy Bekkers, and participants in assorted sessions at the
Econometric Society meetings, the European Trade Study Group, UCL, the Tinbergen Institute, and
CEPR. We would also like to thank Paulo Guimaraes for helpful suggestions on methodological issues.
1
At the macroeconomic level, this relationship has been well explored in the literature with alternative
assumption sets. McKinnon (1966) extended the Lerner symmetry theorem to the three-commodity case
with two import and one export good. The theorem was further extended to non-tradables (McDougall,
1970; Kaempfer and Tower, 1982; Canto et al., 1992), to a three-sector model with non-tradables (Chen
and Devereux, 1994; Milner, 1995). Other extensions have involved imperfect competition (Ray, 1975),
bilateral tariffs (Gardner and Kimbrough, 1990), quantitative restrictions (Lopez and Panagariya, 1995)
and the role of the trade balance (Blanchard, 2009). At the macroeconomic level, Lerner effects relate to
the links between the real and financial sides of the economy.
© 2014 John Wiley & Sons Ltd 567
568 J. FRANCOIS AND M. MANCHIN

We have organised the paper as follows. In Section 2, we outline a basic analytical repre-
sentation of the gravity model that includes both aggregate and bilateral Lerner effects. Sec-
tion 3 provides our empirical analysis, highlighting the effects of import policy on export
performance both at aggregate and bilateral level. We conclude in Section 4.

2. IMPORT POLICY AND EXPORTS

We start with a relatively general representation of the aggregate links between import and
export values, linking import tariffs in exporting markets analytically to exports in the context of
modern gravity model specifications. This is closely linked to the concept of multilateral resis-
tance as developed by Anderson and van Wincoop (2003), where here we include the impact of
import policy on exports. The Anderson et al. (2003) representation of the gravity model, although
constant elasticity of substitution (CES) based, actually follows from a wide range of standard
models (Novy, 2013b). The derivations here provide us with motivation for a set of estimating
equations, both in aggregate and for bilateral gravity modelling of trade, augmented to incorporate
not only standard gravity terms like trade policy in the destination market, but also aggregate trade
protection in the source or export markets as a source of ‘multilateral export resistance’ through
general equilibrium Lerner effects. We start with CES-based aggregate demand in equation (1),
including domestic absorption.
!1=q
X r
q
Qd ¼ as gs;d : (1)
s¼1

In equation (1), g denotes both domestic absorption and imports, and r includes all coun-
tries (and so indexes domestic purchases as well as imports). We adopt the indexing conven-
tion that s = 1…r denotes source and d = 1…r denotes destination. We normalise f.o.b.
prices to unity. We can then define GDP as the quantity of the national good gs summed
across destination markets. This implies the supply constraint in equation (2).
X
r
Gs ¼ gs;d : (2)
d¼1

We now introduce a natural bilateral distance cost cs, j and a policy cost ss, j. Both operate
as iceberg costs. From the properties of constrained optimisation of consumption given
equation (1), treating GDP as the income constraint, we then have
 r
gs;d ¼ Gd Pr1 r
d as ss;d cs;d ; (3)
where in equation (3) the term Pd denotes the CES price index for destination country d associated
with equation (1). Combining equations (2) and (3) and re-arranging, we arrive at the following:
8 " 9
< X X  1r #1 =
  r
r
s
G1 r1 r
 ars s~r1 arh
h;s
gs;d ¼ Gs 1  s G j Pj aj ss;j cs;j ch;s : (4)
: j6¼d;s
s
h¼1
~
s s ;

Here, we have also made a substitution of the CES price index for country s associated
with equation (1). In equation (4), the term s~s denotes average trade costs imposed by
exporter s. We can see that viewed from the supply side, bilateral exports hinge on total avail-
able supply, third-country demand, but also the level of import protection applied by the
exporter s. This is because, to close Lerner’s classic argument, higher protection shifts home

© 2014 John Wiley & Sons Ltd

 PROTECTION AND PERFORMANCE 569

demand towards home goods, which makes less supply available for sale to export partners.
This is reflected, in equation (4), in the last term of the equation.2
A standard estimating equation follows from equation (3), where taking logs we have a
simple log-liner equation. Similarly, we can also modify equation (4). First define V such that
we can rewrite (4) as follows:
gs;d ¼ Gs Vs;d : (5)
Next, we define proportional changes as x^ ¼ dx=x  d ln x. Differentiating equation (5)
and using our definition of x^, we can rewrite (5) and hence (4) as follows:
X
d ln ðgs;d Þ  d ln ðGs Þ þ ev;k d ln ðkÞ: (6)
Basically, log deviations in exports depend on log deviations in supplier GDP, and in the
arguments that go into V, which is the second term in brackets in equation (4). This also depends
on the elasticities of V with respect to the arguments k that go into V. These elasticities are ev,k.
We can also start from equations (2) and (3) to derive a similar condition for aggregate
exports. We note first that domestic absorption is also defined by equation (3).
gs;s ¼ Gs Pr1
s ars : (7)
Combining this with equation (2) and re-arranging, we arrive at aggregate exports zs. Note
that in equations (7) and (8) we have normalised home trade and distance costs to 1.
 
zs ¼ Gs 1  Pr1
s ars ¼ Gs Hs : (8)
The term Hs represents the last right-hand expression in extended form of the equation. It
depends on rates of protection, and on factors like distance from supplier countries. Applying
derivations to equation (8) similar to those for equation (6), we arrive at the following.
X
d ln ðzs Þ  d ln ðGs Þ þ eh;k d ln ðkÞ: (9)
As in bilateral exports, from equation (9), the log deviation in aggregate exports depends on log
deviations in supplier GDP, and in the arguments that go into H, such as average distance and the rate
of protection. It also depends on the elasticities of V with respect to the arguments k that go into H.

3. EMPIRICS

We next turn to an empirical analysis of the impact of import policy on export perfor-
mance both at aggregate level and bilateral trade level.3

2
Technically in equation (4), we have also included the tariff applied to domestic absorption. This can
be cleaned up by imposing a normalisation condition on ck,k.
3
The recent empirical literature on the impact of tariffs on exports includes a mix of econometrics and
computable general equilibrium (CGE) models. Tokarick (2006) uses a CGE model to quantify the
extent to which import tariffs act as an export tax. OECD (2012) find empirical support for the Lerner
symmetry using an aggregate export equation. Other papers have looked at the effects of import protec-
tion on particular export sectors in particular countries. This includes Schiff and Valdes (1992), and
Manzur et al. (1995). More recently, in their empirical work on the role of the WTO in promoting trade,
Subramanian and Wei (2007) invoke own-liberalisation in their econometric model of the evolution of
bilateral trade. Our use of selection modelling is a break from the general approach followed in the liter-
ature.

© 2014 John Wiley & Sons Ltd

570 J. FRANCOIS AND M. MANCHIN

a. Empirical Methodology
(i) Aggregate trade
In the sub section that follows, we start with versions of equation (9). Our estimating equa-
tion for total trade is defined below in equation (10):
ln vx it ¼b0 þ b1 ln ðownsit Þ þ b2 ln ðworldsit Þ þ b3 ln ðsizeit Þ
(10)
þ b4 fit þ b5 ln ðW distanceit Þ þ Fi þ Ft þ eit :
Motivated by natural trade costs in P in equation (9), we have included the GDP-weighted
distance from the world Wdistance. To represent the size of the economy (the term G in equa-
tion 9), we use both population and GDP. In addition to the exporter’s import tariff, own s,
we also include third-country policy (another aspect of the trade cost environment) as the
trade-weighted average tariff faced in export markets, represented by the variable world s.
The term f measures the role of the current account balance.4 Finally, we have also included
exporter and time fixed effects in the regressions. We estimate equation (10) using OLS.

(ii) Bilateral trade

In specifying the underlying structure of equation (11) for the bilateral regressions, or iden-
tically the right-hand side variables that make up vxi,j, we rely on equation (6). There are
many paths that lead to the now standard functional relationship we use here. The first to pro-
pose a gravity equation for trade flows as an empirical specification for trade without theory
was Tinbergen (1962). Anderson (1979) was the first to provide microfoundations based on
the Armington assumption. Among the more recent literature, Anderson and van Wincoop
(2003) elaborate on Anderson (1979) adding a practical way to estimate the gravity equation
structurally (Deardorff, 1988; Evenett and Keller, 2002; Arvis et al., 2013; Novy, 2013a,
2013b). A basic point of Anderson and van Wincoop (2003) is multilateral resistance. Not
accounting for multilateral resistance terms in a gravity model can lead to biased parameter
estimates. This can be addressed with country-level fixed effects, but one then loses scope for
analysis of country-level factors. To get around this, a recent strategy involves Taylor approx-
imations of the multilateral resistance terms to solve for the multilateral resistance terms
(Baier and Bergstrand, 2009). This allows for estimation of the gravity equation, inclusive of
country-level variables. In this paper, we follow Baier and Bergstrand (2009) extended to
accommodate our Lerner variable and time variation in the data.
Following the gravity literature, we expect trade flows to be a function of importer and
exporter income, as well as of determinants of bilateral trade costs, namely distance, tariffs,
and whether countries speak the same language. Finally, pulling all this together yields the
following estimating equation.
ln v x i;j;t ¼a0 þ a1 ln importer GDPj;t þ a2 ln exporterGDPi;t þ a3 ln distancei; j
þ a4 commonlanguagei;j þ a5 ln importersi;j;t þ a6 ln exportersi;t þ a7 fi;t þ ui;j;t :
(11)
x
Equation (11) assesses the determinants of the value of bilateral trade. v i,j,t is the value of
country i exports to country j at time t. ln importerGDPj,t and ln exporterGDPi,t measure the
market size of importers and exporters using GDP. Distance is well established in the gravity

4
See Francois and Manchin (2013) for more discussion on this point.

© 2014 John Wiley & Sons Ltd

 PROTECTION AND PERFORMANCE 571

equation literature (Disdier et al., 2003; Anderson and van Wincoop, 2003). The variable
dummy commonlanguage captures if the traders of the two trading partners can speak the same
language or generally share the same linguistic
 heritage.
 For bilateral import protection, we
use applied tariffs, ln importersi;j;t ¼ ln 1 þ si;j;t , where si,j,t indicates the tariff applied
against exporter i by importer j in period t. The variable ln exportersi,t measures the exporter
country’s own average import tariff rate vis-a-vis the rest of the world. The term f, as in the
aggregate regressions, measures the role of the current account balance from equation (10).
To include multilateral resistance terms, equation (11) is extended following Baier and
Bergstrand (2009). Indexing importers by (j,k,h), and exporters by (i,m,z), equations (19) and
(20) from Baier and Bergstrand (2009, p. 80) are reproduced as equations (12) and (13)
below.
X GDPmt X X GDPht GDPzt
Pit ¼ ln Timt  ð0:5Þ ln Thzt : (12)
m62i
GDPwt h z
GDPwt GDPwt
X GDPkt X X GDPht GDPzt
Pjt ¼ ln Tkjt  ð0:5Þ ln Thzt : (13)
j62j
GDPwt h z
GDPwt GDPwt

Here, we have modified the basic Baier and Bergstrand specification to include time index-
ing. In the case of bilateral tariffs ln Tijt, we can specify multilateral resistance as in equation
(14) below.
X GDPmt X GDPkt X X GDPht GDPmt
MRTijt ¼ Pit þ Pjt ¼ ln Timt þ ln Tkjt  ln Thzt :
m62i
GDPwt k62j
GDPwt h z
GDPwt GDPzt
(14)
We can easily extend equation (14) to the more general case of bilateral time varying vari-
ables Gijt as in equation (15) and importer and exporter multilateral resistance term for the
average tariff of exporters Texport:jt as in equation (16).
X GDPmt X GDPkt X N X N
GDPht GDPmt
MRGijt ¼ ln Gimt þ lnGkjt  ln Ghzt : (15)
m62i
GDPwt k62j
GDP wt h¼1 z¼1
GDP wt GDPzt

X
N
GDPmt X
N
GDPkt X X GDPht GDPzt
MRexport:ijt ¼ ln Iit þ ln Ikt  ln Iht : (16)
m62i
GDPwt k62j
GDPwt h z
GDPwt GDPwt

Our estimating equation augmented by the controls for multilateral resistance for all the
variables proxying for transport costs:5
ln vx i;j;t ¼a0 þ a1 ln importerGDPj;t þ a2 ln exporterGDPi;t þ a3 ln distancei;j
þ a4 commonlanguagei;j þ a5 ln importersi;j;t þ a6 ln exportersi;t þ a7 fi;t
(17)
þ a8 MR ln distancei;j;t þ a9 MRcommonlanguagei;j;t þ a10 MR ln importersi;j;t
þ a11 MR ln exportersi;t þ ui;j;t ;

5
We also run the regression as a robustness check to smooth out imbalances with three years moving
averages. The results are almost identical to those presented here.

© 2014 John Wiley & Sons Ltd

572 J. FRANCOIS AND M. MANCHIN

where MR ln distancei,j,t, MRcommonlanguagei,j,t, and MR ln importersi,j,t have been

constructed following (15),MR ln exportersi,t has been constructed following (16). Also,
following Baier and Bergstrand (2009), we impose constraints linking direct terms to MR
terms in the estimating equation.6
To account for zero bilateral trade flows, we employ a poisson estimator.7 This implies that
we do not take the log of the dependent variable in equation (17). Santos Silva et al. (2006)
argue that gravity-type equations should be estimated in their multiplicative form and propose
to use a Poisson estimation. Using this methodology is consistent in the presence of heterosce-
dasticity and provides a way to deal with zero values of the dependent variable.

b. Data
Our trade and tariff data spanning from 1988 to 2002 were obtained from the UN/World
Bank WITS database system (World Integrated Trade Solution). The data in WITS come, pri-
marily, from the UNCTAD TRAINS and COMTRADE systems and the World Trade Organi-
zation’s integrated tariff database (IDB). The data on GDP were obtained from the World
Bank’s World Development Indicators Database. Geographic data, together with dummies for
same language and colonial links, are taken from Clair et al. (2004; http://www.cepii.fr/ang-
laisgraph/bdd/distances.htm). The distance data are calculated following the great circle for-
mula, which uses latitudes and longitudes of the relevant capital cities. The countries included
in the sample are listed in the appendix.8
There are several country combinations for which trade data are not reported. Following
the recent literature, we assume that these missing observations from the database represent
zero trade (Coe et al., 2002; Felbermayr and Kohler, 2004; Baldwin et al., 2007). However,
we replace zero observations with missing observations in case a country did not report trade
with any other country in a given year since in these cases the data are most probably missing
from the database. We use import data as it is likely to be more reliable than export data
since imports constitute a tax base and governments have an incentive to track import data.
Whenever import data were missing, we used mirrored export data if those were available
(this represented only one-half of one per cent of our observations).

c. Results
(i) Aggregate trade
Estimation results for the aggregate export flows are reported in Table 1 where the depen-
dent variable is export flows to the world. Results presented in Table 1 include time and

6
a1 = 1,a2 = 1,a3 = a8, a4 = a9, a5 = a10, a6 = a11.
7
When examining the global pattern of bilateral trade flows, one striking feature of the landscape is that
many country pairs do not trade. See Baldwin et al. (2006) and Baldwin and Harrigan (2007). In our ini-
tial sample, 42 per cent of importer–exporter pairings had zero bilateral trade (in our final sample
including observations only when tariffs were available the share of zeros was around 20 per cent).
Analysing the determinants of trade flows without taking into account potential trade which does not
take place between country pairs may bias results. At a minimum, unobserved trade may contain infor-
mation about the factors driving bilateral trade relationships.
8
While trade data are available for a wide range of country pairs, the available tariff data are more lim-
ited. For this reason, we utilise a standard WITS procedure of matching the nearest adjacent year to rep-
resent otherwise missing tariff data. Interpolation is then used for wider gaps.

© 2014 John Wiley & Sons Ltd

 PROTECTION AND PERFORMANCE 573

TABLE 1
Lerner Effects at a Macrolevel

ln Exports ln Exports

b1: ln (ownsit) 0.361* 0.515**

(0.210) (0.215)
b2: ln (worldsit) 2.957*** 3.239***
(0.415) (0.424)
b3: ln (gdp) 0.896***
(0.0946)
b3: ln (population) 0.499**
(0.222)
b4:f 0.173*** 0.170***
(0.0198) (0.0202)
b5: ln (Wdistanceit) 0.462*** 1.323***
(0.149) (0.146)
Constant 6.692*** 9.609***
(1.381) (1.379)
Observations 1,095 1,137
R2 0.912 0.581
F(Pr > 0) 5.95 (0.00) 5.98 (0.00)
Number of observations 1,095 1,137

Notes:
(i) Standard errors in parentheses.
(ii) Specification includes time and country fixed effects.
(iii) *** p < 0.01; **p < 0.05; *p < 0.1.

country fixed effects. To test the Lerner effect, the average import tariffs of the exporting
country were included in the regressions. Two different estimates are presented in Table 1.
The difference between the two specifications is that the first specification uses GDP of the
exporter country as a proxy for the size of the economy, while the specification presented in
column 2 includes population as a proxy for size.
The results of both specifications indicate that the exporting country’s own import tariffs
have a negative effect on its own exports. Thus, based on our aggregate regression results, we
cannot reject that the exporting country’s own import policy influences its export perfor-
mance. Trade costs, such as distance, measured as a GDP-weighted distance from the rest of
the world, and the average import tariffs that are applied on the country’s exports are both
negatively influencing the value of total exports. The variable measuring current account defi-
cit is also negative and significant as expected.
The second column presents results for a specification which uses population as a proxy
for size instead of GDP. We estimate this alternative specification as a robustness check of
the results presented in the first column as GDP of the exporting country is correlated with
distance and also tariffs. The sign and significance of the variables do not change; however,
the coefficient of distance and the tariff variables becomes somewhat higher with this specifi-
cation.9 Based on these results, the Lerner effect cannot be rejected. Thus, we find evidence
that the exporting country’s own import tariffs have a negative impact on its exports.

9
The difference in the coefficients between the two specifications is due to the correlation of GDP of
the exporting country with distance and tariffs.

© 2014 John Wiley & Sons Ltd

574 J. FRANCOIS AND M. MANCHIN

(ii) Bilateral trade

Next, we turn to bilateral trade flows. The first column in Table 2 presents results
using poisson estimation including multilateral resistance terms and yearly fixed effects.
The results of the bilateral regressions are similar to those found at the aggregate level
although the coefficients are somewhat different. Following Baier et al. (2009), the coeffi-
cients of the exporter and importer country’s GDP are constrained to be one. All the vari-
ables have the expected sign and significance. The variable measuring the effects of
current account deficit on exports and the coefficient of distance is negative and signifi-
cant. The coefficient of the exporter country’s own import tariff is close to what we found
at the aggregate regression and also to the coefficient of the bilateral import tariffs. Both
the bilateral tariff elasticity and the exporter country’s own tariff coefficient are around
0.6. Thus, these results support the existence of the Lerner symmetry also for bilateral
trade flows.
Lerner effects can also be magnified by intermediate linkages, as stressed for example by
McKinnon (1966). Given the importance of imported inputs for exports, Lerner effects are
likely to be magnified with increasing prevalence of global value chains and production frag-
mentation. This is also in line with the findings of Johnson and Noguera (2012) who conclude
that the content of trade has fallen most in the past two decades among trade partners that
have lower trade costs. This also means that the recent work on detailed cross-border value
chains (Baldwin and Taglioni, 2011) also has implications at the macroeconomic level for
trade balance.

TABLE 2
Poisson Estimates for Bilateral Exports

Non-preferential Trade
Full Sample with Non-imputed Tariffs

a1: ln importerGDPj,t 1 1
(0.000) (0.000)
a2: ln exporterGDPi,t 1 1
(0.000) (0.000)
a3: ln (distanceij) 0.609*** 0.500***
(0.0187) (0.0208)
a4: ln (commonlanguageij) 1.041*** 0.717***
(0.0519) (0.0646)
a5: ln importersi,j,t 0.627*** 0.607***
(0.0398) (0.0479)
a6: ln exportersi,t 0.582*** 0.408***
(0.0740) (0.0435)
a7:fi,t 0.169*** 0.208***
(0.0144) (0.0187)
Constant 11.66*** 11.43***
(0.117) (0.136)
Observations 106,561 82,625

Notes
(i) Robust standard errors in parentheses.
(ii) Regressions include annual fixed effects and multilateral resistance terms for all trade cost variables.
(iii) *** p < 0.01.

© 2014 John Wiley & Sons Ltd

 PROTECTION AND PERFORMANCE 575

(iii) Robustness
A potential endogeneity problem can be present in our bilateral regression. There is a possi-
bility of reverse causality in case bilateral exports would influence import policy and thus
bilateral import tariffs.10 To address this potential reverse causality, we restrict our sample to
non-preferential trade flows. The bilateral tariffs applicable in the case of non-preferential trade
flows are the most favoured nation tariffs which are not determined by country-pair trade rela-
tions but set equally for all partner countries; thus, reverse causality is unlikely. We also omit
from the sample imputed missing values as a further robustness check.
The results for this reduced sample are presented in the second column in Table 2. This
sample is smaller and includes only non-preferential trade. The results are similar to those
using the full sample with the coefficient on the exporter’s own tariffs being a slightly lower
(0.408 instead of 0.582). Nevertheless, our results hold. Lerner effects are confirmed based
on our results also at bilateral level.

4. SUMMARY
In this paper, we examine linkages between the trade policy of exporting countries and
their export performance, both at the aggregate and bilateral level. This involves analytical
extension of the classic definition of aggregate effects (linked to Lerner symmetry) to bilateral
effects allowing us to use a bilateral gravity model. We test the importance of the exporting
country’s import policy for its own export performance (the ‘Lerner effect’ leading to export
resistance) both with aggregate and bilateral trade flows. This is based on the theoretical
framework developed in the paper. We find at both the aggregate level, and also at the bilat-
eral level, that the trade policy of the exporting countries is a significant factor in explaining
export performance. Indeed, given approximate symmetry as suggested by theory, the policy
of exporters is as important, econometrically, as policy in import or destination markets. The
results imply that, when exploring multilateral or country specific determinants of trade in a
gravity context, trade policies in exporting markets deserve place of importance next to trade
policies in importing markets. Given the importance of imported inputs for exports, Lerner
effects are likely to be magnified with increasing importance of global value chains and pro-
duction fragmentation.

REFERENCES

Anderson, J. (1979), ‘A Theoretical Foundation for the Gravity Equation’, American Economic Review,
69, 1, 106–16.
Anderson, J. and E. van Wincoop (2003), ‘Gravity with Gravitas: A Solution to the Border Puzzle’,
American Economic Review, 93, 1, 170–92.
Arvis, J.-F., Y. Duval, B. Shepherd and C. Utoktham (2013), ‘Trade Costs in the Developing World:
1995–2010’, World Bank Policy Research Working Paper 6309 (Washington, DC: World Bank).
Baier, S. L. and J. H. Bergstrand (2009), ‘Bonus Vetus OLS: A Simple Method for Approximating Inter-
national Trade-cost Effects Using the Gravity Equation’, Journal of International Economics, 77, 1,
77–85.
Baldwin, R. and J. Harrigan (2007), ‘Zeros, Quality and Space: Trade Theory and Trade Evidence’,
CEPR Discussion Paperr 6368 (London: CEPR).

10
Reverse causality is unlikely to be a problem for the exporting country’s own import tariffs as this
variable is an aggregate tariff over all products, and all import partners thus cannot be influenced by sec-
toral lobbies or other factors influencing trade policy.

© 2014 John Wiley & Sons Ltd

576 J. FRANCOIS AND M. MANCHIN

Baldwin, R. and D. Taglioni (2006), ‘Gravity for Dummies and Dummies for Gravity Equations’, CEPR
Discussion Paper 5850 (London: CEPR).
Baldwin, R. and D. Taglioni (2011), ‘Chains: Estimating Bilateral Trade Flows When Parts and Compo-
nents Trade Is Important’, NBER Working Paper 16672 (Cambridge, MA: NBER).
Blanchard, E. J. (2009), ‘Trade Taxes and International Investment’, Canadian Journal of Economics,
42, 3, 882–99.
Canto, V., W. H. Kaempfer and E. Tower (1992), ‘On the Validity of the Lerner Neutrality and Symme-
try Theorems in The Presence of Non-Traded Goods’, Contemporary Studies in Economic and Finan-
cial Analysis, 62, 87, 125–28.
Chen, L. L. and J. Devereux (1994), ‘Protection, Export Promotion and Development’, Journal of Devel-
opment Economics, 40, 387–95.
Clair, G., G. Gaulier, T. Mayer, and S. Zignago (2004), Notes on CEPII’s Distances Measures (Paris:
CEPII).
Coe, D. T., A. Subramanian and N. T. Tamirisa (2002). ‘The Missing Globalisation Puzzle’, IMF Work-
ing Paper WP/02/171. (Washington, DC: IMF).
Deardorff, A. (1988), ‘Determinants of Bilateral Trade: Does Gravity Work in a Neoclassical World?’,
in J. Frankel (ed.), The Regionalization of the World Economy (Chicago, IL: University of Chicago
Press), 7–22.
Disdier, A. C. and K. Head (2003), ‘Exaggerated Reports on the Death of Distance: Lessons from a
Meta-analysis’, mimeo (TEAM, Universitede Paris I Pantheon Sorbonne).
Evenett, S. J. and W. Keller (2002), ‘On Theories Explaining the Success of the Gravity Equation’,
Journal of Political Economy, 110, 2, 281–316.
Felbermayr, G. J. and W. Kohler (2004), ‘Exploring the Intensive and Extensive Margins of World
Trade’, Review of World Economics, 142, 4, 642–74.
Francois, J. F. and M. Manchin (2013), ‘Lerner Meets Gravity’, UCL, Economics and Business Working
Paper 123 (London: UCL).
Francois, J. F. and J. Woerz (2009), ‘Non-linear Panel Estimation of Import Quotas: The Evolution of
Quota Premiums under the ATC’, Journal of International Economics, 78, 2, 181–91.
Gardner, G. W. and K. P. Kimbrough (1990), ‘The Economics of Country-specific Tariffs’, International
Economic Review, 31, 3, 575–88.
Johnson, R. C. and G. Noguera (2012), ‘Fragmentation and Trade in Value Added Over Four Decades’,
NBER Working Paper 18186 (Cambridge, MA: NBER).
Kaempfer, W. E. and E. Tower (1982), ‘The Balance of Payments Approach to Trade Tax Symmetry
Theorems’, Weltwirtschaftliches Archiv, 118, 1, 148–65.
Lerner, A. (1936), ‘The Symmetry Between Import and Export Taxes’, Economica, 3, 11, 306–13.
Lopez, R. and A. Panagariya (1995), ‘The Lerner Symmetry and Other Results in the Presence of Quotas
and Nontraded Goods’ University of Maryland at College Park Working Paper 11 (College Park,
MD: University of Maryland at College Park).
Manzur, M. and A. Subramaniam (1995), ‘Who Pays For Protection in Malaysia?: A New General Equi-
librium Approach’, Journal of Economic Integration, 10, 3, 372–85.
McDougall, I. A. (1970), Non-traded Commodities and the Pure Theory of International Trade, in I. A.
McDougall and R. H. Snape (eds.), Studies in International Economics (Amsterdam: North Holland),
157–92.
McKinnon, R. I. (1966), ‘Intermediate Products and Differential Tariffs: A Generalization of Lerner’s
Symmetry Theorem’, Quarterly Journal of Economics, 80, 4, 584–615.
Milner, C. R. (1995), ‘Relative Incentives and Trade Strategies: Typologies and Possibilities’, Economic
Record, 71, 3, 217–26.
Novy, D. (2013a), ‘Gravity Redux: Measuring International Trade Costs with Panel Data’, Economic
Inquiry, 51, 1, 101–21.
Novy, D. (2013b), ‘International Trade without CES: Estimating Translog Gravity’, Journal of Interna-
tional Economics, 89, 2, 271–82.
OECD (2012), ‘Succeeding with Trade Reforms The Role of Aid for Trade’. Available at: http://www.
oecd.org/dac/aft/SucceedingwithTradeReforms.pdf (accessed 7 February 2014)
Ray, E. J. (1975), ‘The Impact of Monopoly Pricing on the Lerner Symmetry Theorem’, Quarterly Jour-
nal of Economics, 89, 4, 591–602.

© 2014 John Wiley & Sons Ltd

 PROTECTION AND PERFORMANCE 577

Santos Silva, J. M. C. and S. Tenreyro (2006), ‘The Log of Gravity’, Review of Economics and Statis-
tics, 88, 4, 641–58.
Schiff, M. and A. Valdes (1992), The Political Economy of Agricultural Pricing Policy. Volume 4, A
Synthesis of the Economics in Developing Countries (Baltimore, MD: Johns Hopkins University
Press).
Subramanian, A. and S. Wei (2007), ‘The WTO Promotes Trade, Strongly but Unevenly’, Journal of
International Economics, 72, 1, 151–75.
Tinbergen, J. (1962), ‘An Analysis of World Trade Flows’, in J. Tinbergen (ed.), Shaping the World
Economy. (New York: The Twentieth Century Fund).
Tokarick, S. (2006), ‘Does Import Protection Discourage Exports?’, IMF Working Paper WP/06/20
(Washington, DC: IMF).

© 2014 John Wiley & Sons Ltd

578 J. FRANCOIS AND M. MANCHIN

APPENDIX

TABLE A1
Sample Countries

Reporter and Partner

Albania Guyana Nepal
Argentina Hong Kong, China New Zealand
Australia Honduras Oman
Austria Croatia Pakistan
Belgium Hungary Panama
Benin Indonesia Peru
Bangladesh India Philippines
Bulgaria Ireland Papua New Guinea
Bahamas, The Iran, Islamic Republic Poland
Bolivia Iceland Portugal
Brazil Israel Paraguay
Barbados Italy Romania
Botswana Jamaica Russian Federation
Central African Republic Jordan Rwanda
Chile Japan Senegal
Cote d’Ivoire Kenya Singapore
Cameroon Korea, Republic El Salvador
Congo, Republic Kuwait Slovakia
Colombia Sri Lanka Slovenia
Costa Rica Lithuania South Africa
Cyprus Latvia Sweden
Czech Republic Luxembourg Syrian Arab Republic
Germany Morocco Chad
Dominican Republic Madagascar Togo
Algeria Mexico Thailand
Ecuador Mali Trinidad and Tobago
Egypt, Arab Republic Malta Tunisia
Spain Mauritius Turkey
Estonia Malawi Tanzania
Finland Malaysia Uganda
Gabon Namibia Ukraine
Ghana Nicaragua Venezuela
Guatemala Norway Zambia
Zimbabwe
Partner Only
Fiji Sierra Leone United Arab Emirates
Haiti

Source: Own calculations.

© 2014 John Wiley & Sons Ltd