SSRN 3167749

6920
2018
February 2018
Institutions, Trade and

Development: A Quantitative
Analysis
Cosimo Beverelli, Alexander Keck, Mario Larch, Yoto V. Yotov
Electronic copy available at: https://ssrn.com/abstract=3167749

Impressum:
CESifo Working Papers

ISSN 2364‐1428 (electronic version)
Publisher and distributor: Munich Society for the Promotion of Economic Research ‐ CESifo
GmbH
The international platform of Ludwigs‐Maximilians University’s Center for Economic Studies
and the ifo Institute
Poschingerstr. 5, 81679 Munich, Germany
Telephone +49 (0)89 2180‐2740, Telefax +49 (0)89 2180‐17845, email [email protected]
Editors: Clemens Fuest, Oliver Falck, Jasmin Gröschl
www.cesifo‐group.org/wp
An electronic version of the paper may be downloaded

∙ from the SSRN website: www.SSRN.com
∙ from the RePEc website: www.RePEc.org
∙ from the CESifo website: www.CESifo‐group.org/wp

CESifo Working Paper No. 6920
Category 8: Trade Policy
Institutions, Trade and Development:

A Quantitative Analysis
Abstract
We propose and apply methods to quantify the impact of national institutions on international
trade and development. We are able to identify the direct impact of country-specific institutions
on international trade within the structural gravity framework. Our approach naturally addresses
the prominent issue of endogenous institutions. The empirical analysis offers robust evidence
that stronger institutions promote trade. A counterfactual analysis reveals that the changes in
institutional quality in the poor countries in our sample between 1996 and 2006 have had, via
their impact on imports from rich countries, significant and heterogeneous real GDP effects,
varying between -5 and 5 percent. Our methods are readily applicable to identifying the impact
of a wide range of country-specific variables on international trade.
JEL-Codes: F130, F140, F160.
Keywords: institutional quality, international trade, development, structural gravity.
Cosimo Beverelli Alexander Keck

Economic Research and Statistics Division Economic Research and Statistics Division
World Trade Organization World Trade Organization
Genève / Switzerland Genève / Switzerland
[email protected] [email protected]
Maro Larch Yoto V. Yotov

Department of Law and Economics School of Economics
University of Bayreuth Drexel University
Bayreuth / Germany Philadelphia / PA / USA
[email protected] [email protected]
February 9, 2018

[I]nstitutions [...] are the fundamental cause of economic growth and develop-
ment dierences across countries
Acemoglu and Robinson, 2010
Domestic institutions can have profound eects on international trade
Nunn and Treer, 2014
1 Introduction
What determines dierences in real income across countries is one of the most fundamental
questions in international economics. In this paper, we seek to contribute to the debate by
quantifying the impact of institutional quality on development and cross-country real income
dierences through international trade.
The focus on trade and institutions is motivated by the abundant evidence that both are
determinants of income levels and growth. On the one hand, the recent trade and growth
literature overwhelmingly nds that trade has a positive causal eect on income levels and
1
growth. Among the mechanisms through which trade causes growth, several studies have
found a positive role of knowledge spillovers, import of new varieties and cost-reducing
2
innovation. On the other hand, studies such as North (1981; 1990), La Porta et al. (1997),
and Acemoglu et al. (2001) show that high-quality institutions are a primary determinant of
1 The seminal paper in the trade and growth literature is Frankel and Romer (1999). For more recent
contributions, see Feyrer (2009; 2011), Brueckner and Lederman (2015), and Donaldson (2015) for a review.
Bernhofen and Brown (2005), Wacziarg and Horn Welch (2008) and Bühler et al. (2011) show that trade
liberalization episodes cause economic growth.
2 Building on the literature on knowledge spillovers, Buera and Obereld (2016) construct a structural
model in which trade aects the diusion of ideas. The model explains over a third of the TFP growth in
China, Chinese Taipei and the Republic of Korea between 1962 and 2000. Broda et al. (2017) show that
new imported varieties account for 10-25 percent of countries' productivity growth. Impullitti and Licandro
(2017) nd that trade liberalization generates tougher rm selection. This increases rms' incentives to
innovate, thereby leading to a higher aggregate productivity growth rate. Calibrating their model to match
the US economy, they nd that moving from autarky to a 8.6 percent import penetration ratio is associated
with a 57 percent increase in aggregate productivity growth due to innovation by more productive rms
(exporters).

3
economic performance. Beyond their direct eects on economic growth, institutions are also
moderators of the relationship between trade and growth, as emphasized by Pascali (2017)
4
in a recent but already inuential contribution.
To estimate the impact that institutional quality has on countries' comparative devel-
opment through international trade, we employ a structural gravity methodology. This
methodology is based on identifying empirically the impact of domestic institutions on inter-
national trade ows (relative to intra-national trade ows). From a theoretical perspective,
there are at least two reasons why institutions may have a direct eect on trade. First,
5
they are a source of comparative advantage. Second, low-quality institutions act as trade
6
frictions in bilateral trade, raising the cost of international exchange.
We make four contributions to the literature, two methodological and two empirical. The
rst methodological contribution is to propose a simple approach to identify the direct eects
of country-specic national institutions on international trade within a structural gravity es-
timation framework. As famously argued by Anderson and van Wincoop (2003), failure to
control for the (unobservable) multilateral resistance terms (MRTs) in gravity estimations
leads to biased estimates of the coecients of the determinants of trade ows. To control
3 Nunn (2009) provides a review of this literature. North (1981) argues that both contracting institutions
(which support private contracts) and property rights institutions (which protect against unlawful expro-
priation) aect economic growth. Subsequent work by Acemoglu and Johnson (2005) shows that property
rights institutions are the ones that matter most for long-term growth.
4 Pascali (2017) estimates the eects of an exogenous increase in international trade (given by the asym-
metric changes in shipping costs that resulted from the introduction of the steamship) on per capita incomes
in the second half of the XIX century. In a restricted sample of 37 countries, he nds that trade reduced
per capita GDP growth rates by more than a third in countries characterized by an executive power with
unlimited authority, while it increased per capita GDP growth rates by almost one-tenth in countries in
which the executive power was obliged to respond to several accountability groups.
5 In particular, contracting institutions shape comparative advantage in complex products that require
relationship-specic investments, thereby giving rise to the hold-up problem (Nunn, 2007). Moreover, in-
stitutions associated with nancial development shape comparative advantage in industries that require
relatively more external nance (Manova, 2008). Finally, labor market-related institutions shape compara-
tive advantage in settings with endogenous division of labor (Costinot, 2009), hiring and ring costs (Cuñat
and Melitz, 2012), and search frictions in the labor market (Helpman and Itskhoki, 2010). For a review of
the literature on institutions and comparative advantage, see Nunn and Treer (2014).
6 In a seminal paper, Anderson and Marcouiller (2002) show analytically that the quality of the importing
country's institutions determines the amount of risk of insecurity, which operates as a mark-up on imported
goods. In a model featuring heterogeneous rms, Crozet et al. (2008) show that insecurity decreases the
volume of bilateral exports by reducing the number of exporters.

for the MRTs, the standard practice in the literature, following Hummels (2001) and Feen-
stra (2004), is to adopt an econometric treatment with exporter and importer xed eects.
Unfortunately, while very powerful and convenient from an econometric perspective, the
country-specic xed eects do not allow for identication of the impact of national insti-
tutions, since the latter are also country-specic and, therefore, they are perfectly collinear
with the xed eects.
To overcome this identication challenge, researchers have adopted two alternative ap-
proaches. Some studies, e.g. Anderson and Marcouiller (2002), Yu et al. (2015), and Álvarez
et al. (2018), have constructed bilateral institution variables as a combination of the insti-
tutional indexes on the importer and on the exporter side. The advantage of this approach
is that it respects the structural properties of the gravity model by allowing estimation with
7
the proper set of exporter and importer xed eects. The disadvantage of this method
is that it does not allow for direct identication of the impact of national institutions on
international trade and, therefore, poses a challenge with the interpretation of the estimates
of the impact of (bilateral) institutions on trade.
Other studies, e.g. de Groot et al. (2004), Dutt and Traca (2010), de Jong and Bogmans
(2011), Francois and Manchin (2013), Gil-Pareja et al. (2017), and Álvarez et al. (2018)
have been able to estimate the direct impact of national institutions on international trade,
8
however, at the expense of not properly controlling for the multilateral resistance terms. As
a consequence, the estimates of institutional quality on trade in such studies are potentially
9
biased, subject to the Anderson and van Wincoop (2003) MR critique.
7 In Table 3 of Álvarez et al. (2018), bilateral institution variables (constructed as the dierence between
importer and exporter in various measures of institutional quality) are used. However, only exporter xed
eect are included, while importer xed eects are omitted.
8 In Table 2 of Álvarez et al. (2018), institution variables for the importing country are used (rather than
bilateral institutions as in their Table 3). Only exporter xed eect are included, while importer xed eects
are omitted. Dutt and Traca (2010) and de Jong and Bogmans (2011) consider the eects of corruption
on bilateral trade ows. In their aggregate analysis with an exporter-importer-time panel, Dutt and Traca
(2010) include either importer and exporter non time-varying xed eects, or bilateral country-pair dummies.
They cannot, however, include importer-time and exporter-time dummies. Similarly, de Jong and Bogmans
(2011), who estimate a cross-sectional gravity equation, cannot include importer and exporter xed eects
to identify the coecients on importer's and exporter's corruption.
9 de Jong and Bogmans (2011) and Francois and Manchin (2013) approximate multilateral resistance

Our methods simultaneously address the deciencies of both of the above mentioned
approaches by delivering estimates of the impact of country-specic institutional measures
within a structural gravity framework with exporter and importer xed eects. To achieve
this goal, we extend on the identication strategy of Heid et al. (2017), who propose the
addition of intra-national trade ows to gravity estimations in order to be able to identify the
impact of non-discriminatory unilateral trade policies within the structural gravity frame-
work. Similar to Heid et al. (2017), we show that the introduction of intra-national trade
ows in gravity estimations allows for identication of the impact of country-specic national
institutions. However, unlike Heid et al. (2017), who are able to identify the impact of uni-
lateral trade policies separately on the importer and on the exporter side, we demonstrate
that it is not possible to identify the impact of national institutions (as well as of any other
country-specic variable of interest) separately on exports versus imports. Instead, we show
that it is only possible to obtain estimates of the impact of national institutional quality (or
any other country-specic variable) on international trade relative to internal trade.
Our second methodological contribution relates to the literature that accounts for endo-
geneity of institutions. A large body of historical evidence supports the idea that pivotal
10
institutional changes can occur as a result of economic integration. Moreover, a series of
prominent studies, e.g. Hall and Jones (1999), Acemoglu et al. (2001) and, more recently,
Auer (2013), recognize the fact that national institutions are, to some extent, determined
by income levels. In the light of the link between trade and income, this endogenous link
between institutions and income is another reason why the relationship between institutions
terms with the Baier and Bergstrand (2009)'s transformation. However, such transformation is an ap-
proximation centered around a world with symmetric and identical trade costs, which is not suited to the
counterfactual experiments we implement. Moreover, this approximation only applies to a log-linearized
model, and not to the PPML model that we use to estimate our gravity equation.
10 See, among several others, Puga and Treer (2014) for an analysis of merchants-led institutional change
in Venice during the Commercial Revolution of the early XI century; Acemoglu et al. (2005) for a study
on the establishment of strong property rights protections in XVI century Britain; Inikori (2003) and Nunn
and Wantchekon (2011) for analyses of the adverse consequences on institutional development of the slave
trade in Africa. A comprehensive review is provided by Nunn and Treer (2014). The conclusion of their
survey is that the impact of international trade on domestic institutions is the single most important source
of long-run gains from trade (Nunn and Treer, 2014, p. 309).

and international trade cannot be treated as exogenous.
Our approach to identify the eects of national institutional quality on international
trade relative to internal trade oers an alternative to deal with the endogeneity issue.
First, exporter and importer xed eects absorb and control for all observable and, more
importantly, unobservable country-specic links between trade and national institutions.
Further, our variable of interest is the interaction term between an ìnternational trade'
dummy (equal to one if the trade ow is international, zero if it is intra-national) and
institutions. As shown in a recent paper by Nizalova and Murtazashvili (2016), we will obtain
consistent estimates if our international trade dummy is not correlated with institutional
quality or the potentially omitted variables. This is the case, rst because the international
trade dummy is independent of any country choice, as it is one for all international ows and
zero else. Hence, it is exogenous and does not vary systematically with institutional quality;
and second because, on top of the exogeneity and the xed eects mentioned above, we
include standard country pair gravity controls (in cross-section estimations) and pair xed
eects (in panel estimations) that greatly reduce the scope for omitting variables. Therefore,
our methods should deliver proper estimates of the impact of national institutions on trade
11
without the need to use instrumental variables.
We also make two empirical contributions to the existing literature. First, we apply our
methods to obtain estimates of the impact of national institutions on international trade.
To that end, we employ (cross-section and panel) data sets that include consistently con-
structed international and intra-national manufacturing trade ows. We obtain positive,
large and statistically signicant estimates of the impact of national institutions on inter-
national trade across specications that employ dierent samples, alternative estimators,
and various measures of institutional quality. Thus, we contribute to the related literature
by oering robust support for a positive causal relationship between institutional quality
11 Despite this argument, in the empirical analysis we also employ IV estimators with leading instruments
from the existing literature. Reassuringly, the IV estimates are not statistically dierent from the results
that are obtained without instrumenting for potentially endogenous institutions.

and international trade. Prominent papers that also obtain positive estimates of the impact
of institutions on trade include Anderson and Marcouiller (2002), de Groot et al. (2004),
12
Francois and Manchin (2013), and Álvarez et al. (2018). The main dierence between
the above mentioned papers and our approach is that we are able to identify the direct
impact of country-specic institutions, while as discussed above the existing literature
either employs bilateral institutional measures or does not control properly for the structural
multilateral resistance terms.
The last main contribution of this paper is the specic focus on the impact of poor coun-
tries' institutions which vary substantially more over time than rich countries' institutions
in our sample. We nd a signicantly stronger impact of institutional quality on the imports
of poor countries from rich countries than on their exports to rich countries. This reects
the comparative advantage of the industrialized countries in the production and exports of
manufacturing goods, which are covered by our data. We use this result, in combination
with the general equilibrium properties of the structural gravity model, to demonstrate the
importance of institutional quality for economic development and for cross-country real in-
come dierences. To that end, we perform an ex-post counterfactual analysis of the impact
of observed changes in institutional quality in the poor countries in our sample over the
period 1996-2006. The main nding from the counterfactual experiment is that the changes
in institutional quality in the poor countries in our sample have had signicant and het-
erogeneous impact on their welfare, as measured by real GDP. For example, our estimates
reveal that, via its impact on trade, improvement in institutional quality increased real GDP
by almost 5 percent in the best performing country, while negative changes in institutional
quality in other countries led to decreases in real GDP of up to 5.2 percent. Importantly,
12 The empirical literature on corruption and trade obtains more ambiguous results. In Dutt and Traca
(2010), corruption taxes trade in the majority of cases, but in high-tari environments it is trade enhancing.
In de Jong and Bogmans (2011), measures of corruption in general are negatively associated with international
trade, while measures of corruption directly related to international trade, such as frequent payments to
customs, are positively associated with trade. Gil-Pareja et al. (2017) argue that, when using subjective
perception measures, corruption decreases bilateral trade ows, while the opposite is true when using an
objective measurement of corruption such as the Structural Corruption Index (SCI).

our results reveal that the main driver of the changes in real GDP for the poor countries in
our sample was the change in consumer prices.
Stimulated by the large and heterogeneous estimates of the GE impact of institutional
quality on the welfare of the poor countries in our sample, we attempt to identify certain
pre-existing economic conditions that would facilitate or hinder the positive impact of in-
stitutional quality improvements on development through trade. To achieve this goal, we
undertake a second counterfactual analysis where we simulate a uniform increase in the
quality of institutions of the poor countries in our sample and exploit the variation in the
resulting real GDP changes to study their determinants. This analysis reveals that the initial
level of institutional quality as well as the initial level of national output are not signicant
factors in determining the eectiveness of institutional quality improvements. We view this
as an encouraging result because it suggests that not only countries that are richer or that
have better institutional quality initially, but also poorer countries and countries with poor
institutions can benet from institutional improvements. We also nd that trade openness
promotes the impact of institutional improvements. The implication of this result is that
there might be signicant benets from implementing a combination of trade liberalization
and institutional improvements.
The rest of the paper is organized as follows. Section 2 oers a brief review of the
structural gravity model (in Subsection 2.1) and presents our identication strategy (in Sub-
section 2.2). Section 3 describes our data set (in Subsection 3.1), obtains partial equilibrium
estimates of the impact of institutions on international trade as well as dierential estimates
of the eects of institutions for poor versus rich countries (in Subsection 3.2), and presents
general equilibrium welfare eects of institutional quality changes observed in poor countries
between 1996 and 2006, as well as results from a counterfactual experiment which hypo-
thetically improves the institutional quality in those countries (in Subsection 3.3). Section 4
concludes with a summary of our ndings and with a brief discussion of possible directions
for future research. Finally, an analytical exposition of our methodology and the results from

a series of robustness tests are respectively included in Appendices A and B.
2 Theoretical foundations and identication

We start, in Section 2.1, with a brief review of the structural gravity model as the workhorse
framework for partial and general equilibrium analysis of the determinants of bilateral trade
ows. Gravity theory will guide our estimation analysis and we will rely on it to perform the
general equilibrium counterfactual experiments. Then, in Section 2.2, we present our strategy
to identify the impact of country-specic national institutions on international trade.
2.1 Review of the structural gravity model

It is now well-established and understood that the following generic theory-founded struc-
13
tural gravity model can be derived from a very wide class of microeconomic foundations:
Yi Ej
Xij = Tij . (1)
Πi Pj
Here, Xij are bilateral trade ows from exporting country i to importing country j ; Tij de-
notes any determinants of trade between countries i and j , including bilateral trade barriers,
such as geographic distance and regional trade agreements, as well as country-specic trade-
related drivers, such as institutions. Yi denotes the total value of production of country i,
P
which can be calculated as total sales at home and abroad: Yi = Xii + j6=i Xij . Ej is the
expenditure in country j , which can be calculated for each country as sum across all bilateral
P
imports, including the domestic sales in country j : Ej = Xjj + i6=j Xij . Finally, Πi and Pj
denote the structural outward and inward multilateral resistance terms (MRTs) of Anderson
13 For expositional simplicity, equation (1) is presented in a cross-section form for a generic sector. However,
the same equation applies separately to sectoral data as well as to panel data. We refer the reader to Head
and Mayer (2014), Costinot and Rodríguez-Clare (2014) and Yotov et al. (2016) for recent surveys of the
empirical and the theoretical structural gravity literature.

and van Wincoop (2003):
X Tij Ej X Tij Yi
Πi = , Pj = . (2)
j
Pj i
Πi
The multilateral resistances in equation (2) have several appealing properties. Intuitively,
they capture the fact that bilateral trade between two countries depends not only on their
sizes and on the bilateral trade costs between them, but also on how isolated/remote each
country is from the rest of the world. All else equal, more multilaterally remote countries
would trade more with each other, see Anderson and van Wincoop (2003). From a structural
perspective, the MRTs are general equilibrium indexes, which decompose the incidence of
trade costs on the producers and on the consumers in each country as if they sell to and/or
buy from a unied world market, see Anderson and Yotov (2010) and Yotov et al. (2016).
As general equilibrium indexes, the MRTs capture the fact that a change in trade costs
between any two countries in the world would aect all other countries in the world. With
respect to decomposing the incidence of trade costs and their changes on producers and
consumers, the outward multilateral resistances can be used in combination with the market
clearing conditions to translate any changes in trade costs into changes in factory-gate prices
and, consequently, into changes in the value of national output. At the same time, the inward
multilateral resistances can be interpreted structurally as ideal consumer price indexes, see
Anderson and Yotov (2010). Thus, in combination, the changes in the inward and in the
outward multilateral resistance terms can be used to obtain the response of real GDP of each
14
country to any change in bilateral trade costs in the world economic system. We capitalize
14 Assuming an endowment economy with CES preferences, we can write nominal output as Yi =
P
Xij =
j
1−σ
P
j (αi pi ) Tij Ej /Pj , where pi denotes the price of the good in country i, αi is a CES preference parameter,
σ denotes the elasticity of substitution, and the replacement of Xij uses the solution for expenditures on
goods shipped from country i j
to country of the consumer's optimization problem. Then we can solve for
αi pi : (αi pi )1−σ = Yi /[ j Tij Ej /Pj ] = Yi /Ωi . Hence, αi pi = (Yi /Ωi )1/(1−σ) and the change in nominal
P
output between the baseline (denoted by superscript b) and the counterfactual (denoted by superscript
c) is given by Yic /Yib = αi pci /(αi pbi ) = pci /pbi = [(Yic /Ωci )1/(1−σ) ]/[(Yib /Ωbi )1/(1−σ) ]. Note that real gross
c c 1/(1−σ)
domestic product can then be calculated as Yi /(Pi ) . We refer the reader to Yotov et al. (2016) for a
detailed discussion of the multilateral resistances, their properties and construction, as well as more detailed
discussion about performing a counterfactual analysis. In addition, Head and Mayer (2014) and Costinot

on these properties in Section 3.3, where we study the general equilibrium implications of
stronger institutions in the poor countries in our sample.
2.2 National institutions and trade: an identication strategy

We develop and present our identication strategy in three steps. First, we show why existing
gravity studies are unable to obtain estimates of the eects of country-specic institutions on
international trade. While the collinearity issue that we describe at this stage is obvious, we
present this trivial step as an opportunity to introduce the design of our analysis as well as
some notation. Second, we demonstrate that the introduction of intra-national trade ows in
structural gravity estimations allows identication of the impact of country-specic national
institutions. Finally, we show that it is not possible to identify separate eects of institutions
on exports versus imports. All steps that we describe in this section extend directly to the
sectoral level and to analysis with panel data.
Start with the following traditional gravity model:
ln Xij = GRAVij β + β1 IQi + β2 IQj + ηi + µj + εij , ∀ i 6= j. (3)
Equation (3) is obtained after log-linearizing equation (1) and two additional steps. First,
we have replaced the bilateral trade costs variable Tij from specication (1) with a vector
of trade cost variables GRAVij , which may include any determinant of bilateral trade (e.g.
distance, trade agreements, etc.), and we have included explicitly the two variables of interest
that measure institutional quality on the exporter side (IQi ) and on the importer side (IQj ).
Second, we have introduced a set of exporter xed eects (ηi ), which will control for the value
of output and for the outward multilateral resistance in the exporter country, and a set of
importer xed eects (µj ), which will control for expenditure and for the inward multilateral
resistance in the importer country. εij is a remainder error term.
and Rodríguez-Clare (2014) oer very informative and insightful reviews of the cutting-edge approaches to
perform general equilibrium analysis with the structural gravity model.
10

Heid et al. (2017) propose the addition of intra-national trade ows to gravity estimations
in order to be able to identify the impact of country-specic variables within the structural
gravity framework. While Heid et al. (2017) suggest how to identify and to interpret the
trade impact of country-specic variables that apply dierently on exports and imports,
such as unilateral trade policies, we extend on their approach and show how to identify the
impact of country-specic variables that apply equally on exports and imports, such as (but
not only) domestic institutions. As shown in Appendix A with a specic data example, this
is not a trivial extension due to a series of collinearity issues discussed therein.
In particular, in Appendix A we show three results that will guide the empirical analysis
of Section 3. First, in a cross-section setting (data varying across importers and exporters),
when including intra- and international trade ows, it is possible to obtain estimates of the
impact of national institutional quality on international relative to internal trade, even in
estimations with importer- and exporter-specic eects. Second, it is not possible to identify
simultaneously the dierential trade eects of importer versus exporter institutions. This is
because, unlike unilateral trade policies that are potentially directional, i.e., specic for a
country as importer or exporter, the institutional quality is the same for a country, indepen-
dently of whether it acts as an exporter or importer. Third, if the analysis is conducted in a
panel setting (data varying across importers, exporters and years), the identication of the
impact of importer's (or exporter's) institutions is still possible after including importer-time,
exporter-time and country-pair xed eects.
In a panel gravity estimation, the use of country-pair xed eects is desirable for two
reasons. First, they would control comprehensively for all observable and unobservable
bilateral trade frictions. Second, as demonstrated by Baier and Bergstrand (2007), the pair
xed eects would mitigate possible endogeneity concerns with respect to the bilateral policy
covariates in gravity equations.
To identify the eect of national institutions on international trade in a gravity framework
with intra- and international trade ows, we will therefore use the following identication
11

strategy. In cross-sectional estimations, we will add importer and exporter xed eects, and
we will estimate the eect of an interaction between national institutions (of the importing
or of the exporting country) and an international trade dummy (equal to one when the
trade ow is international, as opposed to intra-national). In panel estimations, we will add
importer-time, exporter-time and (in some estimations) country-pair xed eects, and still
we will be able to identify the coecient on the interaction term. In both cross-sectional
and panel settings, the coecient of interest identies the impact of national institutional
quality on international relative to internal trade.
Finally, we will also show results of IV estimations that use instruments for institutional
quality from the literature on institutions and income.
3 Empirical analysis
This section has four objectives. First, we demonstrate the ability of our methods to identify
the eects of country-specic institutions on international trade. Second, we quantify the
direct, partial-equilibrium impact of national institutions on international trade. Third,
we study the dierential partial-equilibrium impact of institutions on trade of poor with
rich countries. All partial equilibrium analyses are presented in Section 3.2. Finally, in
Section 3.3, we perform general equilibrium counterfactual experiments, where we study the
links between institutional quality and development through trade. Before we present and
discuss our estimates of the partial and general equilibrium impact of institutional quality
on international trade, in Section 3.1 we describe our data.
3.1 Data: description and sources

Following the existing literature, our main estimates are obtained with a cross-section data
set. Data constraints predetermined 2006 as the year of our main sample. In the sensitivity
analysis we also experiment with panel estimations, which, due to availability of data on
12

institutional quality, cover either the period 1996-2006 or the period 1988-2006, depending
on the institutional index that we use. Availability of reliable data on intra-national trade
ows, which are crucial for the implementation of our methods, led to the use of data on total
manufacturing and, in combination with the rest of the data, we were able to obtain coverage
15
for sixty-three countries. The data that we use to obtain our results can be split into three
categories, which include: (i) data on trade ows (international and intra-national); (ii)
data on institutional quality and economic development; and (iii) data on standard gravity
variables. Next, we oer details on the construction of the variables in each category and we
describe our data sources in more detail.
International and intra-national trade ows. Two sources are used to obtain the data on
international trade ows. The rst source is the United Nations' Commodity Trade Statistics
Database (COMTRADE), and the second source is CEPII's Trade, Production and Bilateral
16
Protection (TradeProd) database. COMTRADE is the primary data source and TradeProd
is used for instances when it includes positive ows for observations when no trade ows are
reported in COMTRADE. Most important for the implementation of our methods, our
database includes consistently constructed intra-national trade ows observations, which are
calculated as the dierence between total manufacturing production and total manufacturing
17
exports. Importantly, we note that both of these variables are reported on a gross basis,
which ensures consistency between intra-national and international trade. Three original
15 The following is the list of the sixty-three countries in our sample (with their respective ISO country
codes in parentheses): Argentina (ARG), Australia (AUS), Austria (AUT), Belgium (BEL), Bolivia (BOL),
Brazil (BRA), Bulgaria (BGR), Cameroon (CMR), Canada (CAN), Chile (CHL), China (CHN), Colombia
(COL), Costa Rica (CRI), Cyprus (CYP), Denmark (DNK), Ecuador (ECU), Egypt (EGY), Finland (FIN),
France (FRA), Germany (DEU), Greece (GRC), Hungary (HUN), India (IND), Indonesia (IDN), Iran (IRN),
Ireland (IRL), Israel (ISR), Italy (ITA), Japan (JPN), Jordan (JOR), Kenya (KEN), Republic of Korea
(KOR), Kuwait (KWT), Malawi (MWI), Malaysia (MYS), Mauritius (MUS), Mexico (MEX), Morocco
(MAR), Nepal (NPL), Netherlands (NLD), Niger (NER), Nigeria (NGA), Norway (NOR), Panama (PAN),
Philippines (PHL), Poland (POL), Portugal (PRT), Romania (ROU), Senegal (SEN), Singapore (SGP),
South Africa (ZAF), Spain (ESP), Sri Lanka (LKA), Sweden (SWE), Switzerland (CHE), Tanzania (TZA),
Thailand (THA), Trinidad & Tobago (TTO), Tunisia (TUN), Turkey (TUR), United Kingdom (GBR),
United States (USA), and Uruguay (URY).
16 UN COMTRADE can be accessed at http://comtrade.un.org, while CEPII's TradeProd can be found
at http://www.cepii.fr/CEPII/en/bdd_modele/presentation.asp?id=5.
17 These data were constructed and kindly provided to us by Thomas Zylkin.
13

data sources are used to construct the production data, which are needed for the calculation
of intra-national trade. The main source is the United Nations' UNIDO Industrial Statistics
database. In addition, the UNIDO data are complemented with data from the CEPII's
TradeProd database and with data from the World Bank's Trade, Production and Protection
18
(TPP) database.
Institutional quality and development indexes. Our main data on institutional quality
come from the World Bank's World Governance Indicators (WGI) database, which includes
data on a series of indicators for institutional quality including Voice and Accountability,
Political Stability and Absence of Violence, Government Eectiveness, Regulatory Quality,
19
Rule of Law, and Control of Corruption. To obtain the main results, we construct an aggre-
gate institutional quality index, as the simple average of the six individual WGI categories.
In the sensitivity experiments, we also conrm the robustness of our main ndings with each
of the individual WGI indexes. In addition to using alternative measures from the WGI
database, we also experiment with two other measures of institutional quality. Specically
we employ (i) the Combined Polity Score index from the Polity IV project; as well as (ii)
an average index that we construct from the Political Rights and Civil Liberties indicators
20
of the Freedom House initiative. An advantage of the Polity IV and the Freedom House
indexes of institutional quality is that they are available for an extended period of time,
which enables us to study the impact of institutions over the period 1988-2006 (while the
WGI data only covers the period 1996-2006).
To obtain some of our results, we implement instrumental variable estimators. To do
so, in the main analysis, we rely on the distance-from-the-equator instrument of Hall and
Jones (1999). In addition, in the robustness experiments, we also employ the mortality-rate
18 The INDSTAT database can be found at http://stat.unido.org. The TPP data set can be accessed at
http://go.worldbank.org/4Z6UU7TO40. For further details on the construction of the intra-national trade
ows data we refer the reader to Baier et al. (2016).
19 We refer the reader to http://info.worldbank.org/governance/wgi/#doc for a detailed description of
the World Bank's WGI database.
20 More information about the original Polity IV data set and its description can be found at http:
//www.systemicpeace.org/polity/polity4.htm. For more information on the Freedom House indexes, we
refer the reader to https://freedomhouse.org/report/freedom-world-2016/methodology.
14

instrument from Auer (2013), who expands on the original `settler mortality' instrument
of Acemoglu et al. (2001) to construct a hypothetical mortality rate for a larger sample of
countries.
In addition to the data on institutional quality, we employ the World Bank's Country
and Lending Groups classication in order to identify the poor and the rich countries in
21
our sample. The poor countries in our sample are dened as those that are classied as
`low-income' or `lower-middle income' economies by the World Bank. The rich countries
are those classied as ùpper-middle income' or `high-income' economies. There are a total
22
of twenty-two `poor' countries and forty-one `rich countries' in the sample. In the main
empirical analysis, we demonstrate that the impact of institutions is much stronger for poor
countries.
Standard gravity variables. In addition to a very rich structure of xed eects, which
control for many observable and unobservable determinants of bilateral trade, in most of the
estimations we rely on proxies for bilateral trade costs that have been used standardly in the
gravity literature. Specically, we employ data on bilateral distance, colonial relationships,
common ocial language and contiguous borders. All of these gravity variables come from
23
the CEPII's GeoDist database. An important advantage of the distance variables from
the GeoDist database, especially for our analysis with intra-national trade ows, is that the
weighted-average methods that are used to construct distance ensure consistency between
the measures of intra-national and international distance. Specically, consistency is ensured
because each method uses population-weighted distances across the major economic centers
21 The World Bank's Country and Lending Groups data, along with a detailed description of the data set
and information about the construction of the data, can be found at https://datahelpdesk.worldbank.org/
knowledgebase/articles/906519-world-bank-country-and-lending-groups.
22 In 2006, the following eight countries belonging to our sample of sixty-three countries were classied
as `low-income' in the World Bank's Country and Lending Groups classication: India, Kenya, Malawi,
Nepal, Niger, Nigeria, Senegal, and Tanzania. The following fourteen countries were classied as `lower-
middle income': Bolivia, Cameroon, China, Colombia, Ecuador, Egypt, Indonesia, Iran, Jordan, Morocco,
Philippines, Sri Lanka, Thailand, and Tunisia. The remaining forty-one countries in the sample were either
classied as ùpper-middle income' (fteen countries) or `high-income' (twenty-six countries).
23 This database can be accessed at http://www.cepii.fr/cepii/en/bdd_modele/presentation.asp?id=6.
We refer the reader to Mayer and Zignago (2011) for detailed description of these commonly used gravity
covariates.
15

within or across countries, respectively. Finally, we also employ data on regional trade
agreements (RTAs). The original data on RTAs, including information about the agreement
name, status, date of notication and of signature, signatories countries and link to the text
of documents, is the WTO Regional Trade Agreements Information System (RTA-IS). We
use Mario Larch's Regional Trade Agreements Database from Egger and Larch (2008), which
24
is based on the original RTA data from the WTO.
3.2 The impact of national institutions on international trade

This section demonstrates the validity of our methods by obtaining partial equilibrium es-
timates of the impact of national institutions on international trade. In addition, we also
obtain dierential estimates of the direct impact of institutions on international trade be-
tween poor and rich countries. The econometric analysis relies on the identication strategy
that we presented in Section 2.2. In order to most clearly demonstrate the eectiveness of
our methods, we develop the analysis sequentially. We start with a standard gravity spec-
ication, which conrms the representativeness of our sample. Then, we show that we can
identify the impact of country-specic institutions within a structural gravity estimation
framework. We proceed by gradually introducing a series of adjustments from the related
literature. Our main estimates are presented in Table 1.
The estimates from column (1) of Table 1 are obtained with the OLS estimator based on
the following standard cross-section gravity specication that does not include a measure of
25
institutional quality:
ln Xij = β1 LN _DIST ij + β2 CN T Gij + β3 LAN Gij + β4 CLN Y ij + β5 RT Aij
+ β1 BRDRij + ηi + µj + εij , (4)
24 The original RTA data of the WTO can be found at http://rtais.wto.org/UI/PublicMaintainRTAHome.

aspx. Mario Larch's RTA database can be accessed at http://www.ewf.uni-bayreuth.de/en/research/
RTA-data/index.html.
25 We start with a cross-section OLS specication because this is the standard and most popular econo-
metric approach to study the impact of institutions in the existing literature. Below we demonstrate that
our cross-section OLS estimates are robust to the use of panel data and of the PPML estimator.
16

where we have used the standard series of observable variables to proxy for trade costs,
including: the logarithm of bilateral distance (LN _DISTij ), whether or not two trading
partners share a common border (CN T Gij ), whether i and j speak the same ocial lan-
guage (LAN Gij ), if two countries share any colonial relationships (CLN Y ij ), and whether
they have an RTA in force (RT Aij ). In addition, we have introduced an indicator variable
BRDRij , which takes a value of one for international trade, and it is equal to zero for intra-
national trade. (See Appendix A for details on why this variable is introduced). Finally,
we also include the sets of exporter xed eects (ηi ) and importer xed eects (µj ), which,
amongst other things, control for the multilateral resistance terms, and add a remainder
error term (εij ).
Without going into details, we note that the estimates from column (1) of Table 1 capture
the fact that distance and international borders are signicant impediments to international
trade, while sharing a common ocial language, sharing colonial ties, and having a regional
trade agreement promote bilateral trade ows, all else equal. This is indicated by the large
and signicant estimates on LN _DIST , BRDR, LAN G, CLN Y , and RT A, respectively.
The estimate of the impact of contiguous borders (CN T G) on international trade is also
positive, as expected, although not statistically signicant. Overall, the gravity estimates
from column (1) of Table 1 are readily comparable to the meta-analysis gravity indexes from
26
Head and Mayer (2014) and, therefore, they establish the representativeness of our sample.
The results from column (2) of Table 1 are obtained with the same cross-section OLS
specication as in column (1). However, in addition to the standard gravity covariates we
have introduced the additional regressor IQ_BRDR, which captures the impact of national
institutions on international trade. As discussed in Appendix A, IQ_BRDR can be de-
ned either on the exporter side (IQ_BRDR ≡ IQi × BRDRij ), or on the importer side
(IQ_BRDR ≡ IQj × BRDRij ), without any quantitative implications for the correspond-
26 Head and Mayer's results are based on more than 150 studies, including more than 2500 gravity esti-
mates.
17

27
ing estimates. In each case, by construction, the estimate on IQ_BRDR captures the
dierential impact of national institutions on international relative to internal trade. Im-
portantly, since the institution interaction term is set to zero for domestic sales, the impact
of institutions can be identied even in the presence of the exporter and the importer xed
eects (ηi and µj ), which apply to both internal and international sales.
Three main ndings stand out from the estimates in column (2) of Table 1. First, they
demonstrate that we were able to obtain an estimate of the impact of national institutions
in the presence of the full set of exporter and importer xed eects and without facing any
collinearity issues. This result is important from a methodological perspective because it
validates our methods in the case of national institutions, and also because it opens a uni-
verse of opportunities to study the dierential impact of any country-specic variable on
28
international relative to internal trade. Second, we obtain a large, positive and statisti-
cally signicant estimate on IQ_BRDR, which reveals that stronger national institutions
promote international trade. This result reinforces the argument for a positive relationship
between institutional quality and international trade from existing studies that use bilateral
institution variables (e.g. Anderson and Marcouiller, 2002) or do not appropriately control
for the multilateral resistances (e.g. de Groot et al., 2004). Third, we note that the esti-
mates of all gravity covariates, except for the estimate on BRDR, are virtually identical to
the corresponding estimates from column (1). As expected (based on the positive estimate
of the impact of institutions), the estimate on BRDR in column (2) is larger in absolute
value as compared to the BRDR estimate from column (1). The natural explanation for
this result is that the estimate of the border eects from column (1) combines the positive
impact of institutions with the negative impact of some unobservable impediments to trade.
In the robustness analysis that we present in Appendix B, we demonstrate that qualitatively
identical results are obtained with a series of alternative measures of institutional quality.
27 We refer the reader to Table B-1 from Appendix B, where we demonstrate empirically the equivalence
of the estimates of the impact of national institutions obtained (i) on the exporter side; (ii) on the importer
side; and (iii) for intra-national trade.
28 For example, value-added taxes, sales taxes, etc.
18

One of the most prominent and dicult challenges with the identication of the impact of
institutions on trade, growth, and development is to address properly potential endogeneity of
the institution regressor. As a result, some of the most inuential papers in the literature that
study the growth impact of institutions, e.g. Hall and Jones (1999), Acemoglu et al. (2001)
and, more recently, Auer (2013), have gained popularity due to the introduction of novel
and successful instruments to control for endogenous institutions. We obtain estimates of
the relative impact of national institutional quality on international trade relative to internal
trade, controlling for exporter and importer specic eects. Hence, in our specications, the
xed eects will absorb and control for all observable and, more importantly, unobservable
country-specic links between trade and national institutions. In addition, Nizalova and
Murtazashvili (2016) show that when interest lies in the estimation of the dierential impact
of a particular factor, consistent estimates can be obtained even when the particular factor
of interest is correlated with omitted variables. This is the case as long as the treatment is
uncorrelated with both the factor of interest as well as the omitted variables. This insight
can be applied to our setting by viewing the international trade dummy as the treatment and
the institutional quality as the factor of interest. Since the distinction of international trade
versus internal trade (not their levels!) should neither be correlated with the institutional
quality nor with the omitted variables (and therefore be exogenous), we will obtain consistent
estimates of the interaction term, i.e. of the eects of institutional quality on international
relative to internal trade.
Even tough, as we just discussed, our econometric approach should address the issue
of endogeneity of institutions in our setting automatically, we follow the prominent related
literature to also implement an instrumental variable strategy in our setting. Specically,
the estimates from column (3) of Table 1 are obtained after instrumenting for endogenous
institutions with the distance-from-the-equator instrument of Hall and Jones (1999). Two
main ndings stand out from column (3). First, as expected based on existing studies, the
Hall-Jones instrument is a strong predictor of institutional strength. The rst stage estimate
19

of the coecient on the Hall-Jones instrument is signicant at any level. Furthermore,
we obtain large under-identication canonical correlations χ2 test statistic of 766.58, (see
Anderson, 1951), and weak-identication Wald F estimates of 922.82 and 16.38 (constructed
following Cragg and Donald, 1993 and Stock and Yogo, 2005, respectively).
Second, and more important for our purposes, the estimate of the impact of institu-
tions in column (3) is large, positive, and statistically signicant at any conventional level.
Furthermore, we nd that, while a bit smaller, the IV estimate from column (3) is not sta-
tistically dierent from the corresponding OLS estimate in column (2), thus, reinforcing the
argument that the country-specic (exporter and importer) xed eects and the focus on
the interaction term avoid severe endogeneity issues. In the robustness analysis that we per-
form in Appendix B, we demonstrate that very similar results are obtained if, instead of the
Hall-Jones instrument, we employ the instrument from Auer (2013), which expands on the
original settler mortality instrument of Acemoglu et al. (2001) to construct a hypothetical
mortality rate for a larger sample of countries. Finally, we also employ a combination of
the Hall-Jones and the Auer instruments to, once again, conrm the robustness of our main
results. See Appendix Table B-4 for further details.
To obtain the estimates in column (4) of Table 1 we use an IV Poisson-Pseudo Maximum
Likelihood estimator (PPML). The PPML estimator has gained popularity for structural
gravity estimations due to three attractive properties. First, as demonstrated by Santos Silva
and Tenreyro (2006), the PPML estimator eectively addresses the issue of heteroskedastic-
ity, which often plagues international trade ows data. Second, since PPML is a multiplica-
tive estimator, it enables researchers to take into account the information that is contained
in the zero trade ows, which appear often in bilateral trade data. Finally, owing to its
additive property (see Arvis and Shepherd, 2013; Fally, 2015), the PPML estimator can be
used to perform structural general equilibrium analysis directly in standard software pack-
ages (e.g. Stata), see Anderson et al. (2016). Similar to the IV-OLS analysis from column
(3), we employ the Hall-Jones instrument to obtain the IV-PPML estimates from column
20

(4). As before, in Appendix B we also demonstrate that the results are robust to employing
the instrument from Auer (2013) as well as a combination of the Hall-Jones and the Auer
instruments. The IV-PPML estimates from column (4) are not statistically dierent from
the results that we already obtained in columns (2) and (3) and, therefore, they reinforce
the robustness of our ndings.
The last two columns of Table 1 report panel estimates over the period 1996-2006. The
panel treatment of the impact of institutions in international trade has at least three ad-
vantages as compared to the cross-section analysis that we employed thus far. First, from a
pure econometric perspective, the panel setting will improve the eciency of our estimates.
Second, the panel setting will enable us to capture the impact of changing institutional
quality over time. This is potentially important because, in principle, the cross-section es-
timates of institutional quality may be biased upwards due to endogeneity resulting from
reverse causality: better institutions may lead to more trade, but more trade could also
lead to better institutions. If this intuition is correct, then we would expect that the panel
estimates of the impact of institutions would be smaller as compared to their cross-section
counterparts. Third, the panel setting will enable us to control comprehensively for all time-
invariant bilateral frictions as well as to further mitigate endogeneity concerns through the
use of country-pair xed eects. Due to data availability constraints imposed by the WGI
database, the years included in our sample are 1996, 1998, 2000, and the period 2002-2006.
Proper treatment of the structural multilateral resistance terms in panel settings requires the
use of exporter-time and importer-time xed eects. Accordingly, we do include such xed
eects in the estimations of the last two columns of Table 1, although omit their estimates
29
for brevity in the table.
The estimates from column (5) of Table 1 are obtained with the standard gravity vari-
ables, where we also introduced a series of time-varying border dummies, which are designed
29 The panel estimates in Table 1 are obtained with the OLS estimator. PPML estimates, which conrm
the robustness of our main results, are presented in Table B-5 of Appendix B.
21

30
to capture any globalization eects. Two main results stand out from the estimates in
column (5). First, we nd that globalization has had a signicant impact on international
trade during the period 1996-2006. This is captured by the positive and increasing estimates
on the time-varying border dummies, which should be interpreted as deviations from the av-
31
erage BRDR estimate. Second, and more important for our purposes, the panel estimate
on IQ_BRDR is still large, positive, statistically signicant at any conventional level, and
it is not (statistically) dierent from the corresponding cross-section estimates from columns
(2)-(4).
The estimates from the last column of Table 1 are also obtained with panel data. How-
ever, in addition to the exporter-time and importer-time xed eects, in column (6) we also
include a complete set of country-pair xed eects. As summarized by Yotov et al. (2016),
the motivation for the inclusion of the country-pair xed eects in gravity estimations is
twofold. First, as demonstrated by Baier and Bergstrand (2007), the country-pair xed ef-
fects eectively control for potential endogeneity of any time-varying bilateral trade policy
variables, e.g. regional trade agreements, as they control for any non-time-varying unob-
servable characteristics correlated with the trade policy variables. Second, the country-pair
xed eects control more thoroughly for bilateral trade costs by absorbing all observable and
unobservable time-invariant bilateral determinants of trade. Accordingly, the results from
column (6) do not include estimates of the eects of distance, contiguity, common language,
colonial ties, and international borders, because these covariates have been absorbed by the
country-pair xed eects.
30 There are at least two important reasons for the introduction of the time-varying border dummies in
our setting. First, the variation in the estimates of the coecients of time-varying border dummies will
capture the impact of globalization on international trade, which, as noted by Coe et al. (2002), has been
missing from structural gravity estimations. Our estimates on the time-varying border dummies will reveal
that globalization is indeed present in gravity regressions. Second, Bergstrand et al. (2015) demonstrate that
the estimates of the eects of economic integration agreements, international borders, and bilateral distance
in recessions that do not explicitly control for globalization trends can be severely biased.
31 Our time-varying border estimates are comparable to the results from Bergstrand et al. (2015) and
conrm that structural gravity estimations can capture the impact of globalization, thus, oering evidence
against the famous `distance puzzle' (see Disdier and Head, 2008) and `missing globalization puzzle' (see Coe
et al., 2002) in international trade.
22

Importantly, once again, we obtain a large, positive and statistically signicant estimate
of the impact of institutions on international trade. We do note, however, that the estimate
on IQ_BRDR from column (6) is smaller in magnitude and less precisely estimated as
compared to the corresponding indexes from columns (2)-(5). The natural explanation for
this result is that, in the presence of the country-pair xed eects, identication of the eects
of institutions is due to the time variation of this variable, which, given the short sample
period and the nature of the institutional variable, is not very large. Further, the smaller
magnitude could be a result of controlling for reverse causality concerns by the country-pair
xed eects. We nd it quite encouraging that we obtain positive eects of institutional
quality even in a very demanding panel setting with such a rich structure of xed eects.
Furthermore, as noted earlier, we think that it is important to be able to identify the impact
of institutions due to changes in institutional quality over time. Therefore, we favor the panel
estimates with country-pair xed eects from column (6) over the rest of the specications
in Table 1.
Trade with rich economies may be an important driver of development and growth for
poor countries (see for instance Amiti and Konings, 2007). However, trade with less devel-
32
oped nations is hampered by signicant uncertainty, often due to poor institutional quality.
Intuitively, strong institutions should be viewed as an indicator for economic stability and
reliability. At the same time, institutional quality in poor countries varies signicantly more
33
than in developed nations. In combination, these arguments point to a potentially dier-
ential, and in particular stronger impact of institutional quality changes on trade between
32 For example, in an interview with Jakob Svensson, which aimed to quantify the experience of domestic
rms in an emerging economy with government regulations and corruption in the foreign trade sector, the
chief executive ocer of a successful [...] manufacturing rm exclaimed: Ì hope to be reborn as a custom
ocial' . When a well-paid CEO wishes for a job with low ocial pay in the government sector, corruption
is almost surely a problem! (Svensson, 2005, p. 19).
33 Inspection of the changes in institutional quality among the countries in our sample reveals that this is
indeed the case. For example, the three countries that have experienced the largest decrease in institutional
quality are countries in the low income and low-middle income group (worsening between -0.66 and -0.47).
On the opposite side of the spectrum we nd that the countries that experienced the biggest improvement in
institutional quality in our sample are also countries from the low income and the low-middle income group,
with improvements between 0.33 and 0.51. The change in the institutional quality in some of the major
developed countries are small in comparison.
23

poor and rich countries. We test this hypothesis in Table 2, where we allow for and esti-
mate a dierential impact of institutional quality on the exports of poor to rich countries
(IQP _BRDRP R ≡ IQP × BRDRP R ) as well as on the imports of poor nations from rich
countries (IQP _BRDRRP ≡ IQP × BRDRRP ), where subscript P denotes poor and sub-
34
script R denotes rich. In order to ease interpretation, we subtract the two new institution
variables from IQ_BRDR. Thus, the estimates on each institution covariate should be
interpreted independently (rather than as deviations).
35
The estimates in Table 2 are obtained with the OLS estimator and panel data. As
before, each specication includes the full set of exporter-time and importer-time xed ef-
fects. In addition, we also include a rich set of time-varying bilateral border dummies that
correspond to the newly-introduced institution variables, whose estimates are omitted for
brevity. Panel A of Table 2 oers estimates with the standard gravity variables. To ease
comparison, column (1) reproduces the estimates from column (5) of Table 1, while column
(2) introduces the new institutional quality variables for trade of poor with rich nations.
Two main ndings stand out from the results in column (2). First, we obtain positive
estimates of the impact of institutional quality on the exports of poor nations to rich nations
as well as on the imports of poor countries from rich nations. The estimates are large
and statistically signicant. Second, we document signicant asymmetries in the impact
of institutional quality on the exports vs. imports of poor nations with rich countries.
Specically, we nd that the imports of poor countries from rich countries will be aected
signicantly more. We nd this result intuitive and our interpretation is that it reects
the pronounced comparative advantage of the industrialized nations in the production and
exports of manufacturing goods, which are covered by our data.
34 Due to perfect multicollinearity, it is not possible to identify both the eect of the exports of poor
countries (to any country) and the eect of the imports of poor countries (from any country).
35 The main motivation for the use of panel data is that our estimates would reect the time-variation
in institutional quality in the poor nations in our sample over the period of investigation. OLS is used for
consistency with the literature and with our main results from Table 1. Table B-5 from Appendix B oers
corresponding estimates that are obtained with the PPML estimator, and which conrm the robustness of
the estimates that we present in Table 2.
24

Panel B of Table 2 oers estimates that are obtained with country-pair xed eects. Once
again, to ease comparison, column (3) reproduces the estimates from column (6) of Table 1,
while column (4) introduces the new institutional quality variables for trade of poor with rich
nations. Two main ndings stand out from the estimates in column (4). First, we see that
the estimates on the new institutional variables are signicantly smaller as compared to the
corresponding numbers that are obtained with the standard gravity variables in column (2).
A possible explanation of this result is that once the country-pair xed eects are introduced,
the impact of institutions is identied mostly of the time-variation in the institutional quality
indexes. Second, we still nd signicant asymmetries between the impact of institutions on
exports vs. imports of poor with rich nations. Consistent with our ndings from column (2),
the asymmetries are in favor of the imports of the poor from the rich countries. Finally, the
estimates from column (4) of Table 2 reveal that, while still positive and relatively large in
economic magnitude, our estimate of the impact of institutions on the exports of poor to
rich nations is no longer statistically signicant.
In sum, the empirical analysis that we presented in this section demonstrates the validity
and eectiveness of our methods to identify the impact of country-specic institutions on in-
ternational trade ows. To that end, we obtained large, positive, and statistically signicant
estimates of the impact of institutions across various specications with alternative esti-
mators, dierent samples, and dierent measures of institutional quality. (More sensitivity
experiments that demonstrate the robustness of our main results can be found in Appendix
B.) In addition, we found evidence that poor countries' institutions matter, particularly for
their imports from rich countries. Stimulated by this partial equilibrium result, in the next
section we quantify the general equilibrium impact of institutions in developing countries.
3.3 Institutions, trade, and development: a GE analysis

The objective of this section is twofold. First, it presents quantitative evidence for the
importance of trade as a channel for transmission of the impact of changes in institutional
25

quality on development. To achieve this goal, we oer ex-post analysis of the impact of
institutional quality changes observed in the poor countries in our sample over the period
1996-2006. Specically, we capitalize on the signicant partial equilibrium estimates from
the previous section (showing that the direct impact of institutional quality on trade of poor
with rich nations is driven by the poor countries' imports from the rich nations) in order
to quantify the general equilibrium impact of changes in institutional quality for the poor
countries in our sample due to their imports from rich nations. The second objective of this
section is to identify certain pre-existing economic conditions that would facilitate or hinder
the positive impact of institutional quality improvements on development through trade. To
achieve this goal, we undertake a second counterfactual analysis where we simulate a uniform
increase in the quality of institutions of the poor countries in our sample and exploit the
variation in the resulting real GDP changes (due to increased poor countries' imports from
rich countries) to study the determinants of such changes.
To perform the counterfactual analysis, we rely on our most demanding and most con-
servative partial equilibrium estimates from column (4) of Table 2. We only focus on the
signicant eect of institutions on imports of poor countries from rich countries, i.e., we use
only the estimate on IQP × BRDRRP in order to generate the initial trade cost shock to the
GE gravity system. We treat the last year in our sample (2006) as the baseline year, and
we change the institutional quality for the poor nations to the values from the initial year
36
in the sample (1996). Thus, in eect, our counterfactual experiment evaluates the impact
of changes in institutional quality over the period 1996-2006 in the poor nations, through
their imports from rich countries, on the welfare of the poor countries in our sample. As
we perform an ex-post analysis, the real GDP results from our GE analysis are reported as
percentage changes from the counterfactual to the baseline values:
1/(1−σ) b 1/(1−σ) c
(Yi /Pj ) − (Yi /Pj )
%∆rGDP = 1/(1−σ) c
× 100, (5)
(Yi /Pj )
36 Recall that the measure of institutional quality used in column (4) of Table 2 is the simple average of
the six individual WGI categories (see Section 3.1).
26

where superscript b denotes baseline and superscript c denotes counterfactual.
The main results from our ex-post counterfactual analysis are presented in Panel A of
Table 3. Column (2) of Table 3 lists the changes in institutional quality for the poor nations
in our sample during the period of investigation (1996-2006). Real GDP percentage changes,
which correspond to the institutional quality changes from column (2), are reported in column
(3). The main nding from column (3) is that the changes in institutional quality in the
poor countries in our sample have had signicant and heterogeneous impact on their real
GDPs. For example, our estimates reveal that, ceteris paribus, changes in real GDP due to
changes institutional quality observed between 1996 and 2006 range between 4.7 percent for
the best performer and -5.2 percent for the worst.
In order to better understand the channels through which changes in institutional quality
impact real GDP, we capitalize on the structural properties of the gravity model to decompose
the total real GDP change into an eect on producers, which is captured by the percentage
change in factory gate prices (%∆p = 100 × (pbj − pcj )/pcj ), reported in column (4) of Table
3; and into an eect on consumers, which is captured by the percentage changes in the inward
b c c
1/(1−σ) 1/(1−σ) 1/(1−σ)
multilateral resistances %∆IM R = 100 × Pj − Pj / Pj ,
reported in column (5) of Table 3. Several ndings are noteworthy concerning the price
eects. First, increases in institutions typically decrease producer and consumer prices, but
consumer prices decrease more than producer prices leading to an increase in real GDP. If
the institutional quality decreases, consumer and producer prices both tend to increase, with
a stronger increase of consumer prices leading to a decrease in real GDP.
We nd these results intuitive. For example, the reason for the inverse relationship
between improvements in institutional quality and consumer prices in the poor nations is
that, according to our partial estimates, better institutions will attract more imports and,
therefore, will result in lower consumer prices in the destination countries. Similar intu-
ition explains the fall in producer prices in the poor nations, i.e., imports lead to more
competitive pricing in destination countries. Finally, the fact that the positive impact of
27

lower consumer prices in response to trade liberalization outweighs the negative impact on
producers is a manifestation of standard gains-from-grade argument from introductory in-
ternational economics. The same logic, but in the opposite direction, applies to explain the
inverse relationship between worsening in institutional quality and the changes in consumer
and producer prices in the poor nations.
Even though, on average, most of the results from Panel A of Table 3 can be explained
intuitively, there are also some instances in which our estimates are exactly the opposite to
those that we just described and, therefore, they do not conform to the traditional intuitive
explanations that we oered above. For some countries we observe an increase in consumer
prices but a fall in producer prices in response to worsening institutional quality. For others,
we observe both lower producer and lower consumer prices in response to worsening insti-
tutions. The reason for the heterogeneous response of prices, and therefore real GDPs, to
changes in institutional quality is that we are investigating a complex general equilibrium
counterfactual scenario with dierent institutional changes for many countries that may im-
pact other countries directly and indirectly. Thus, in principle, the price eects can go in
37
either direction.
Another important and clear result that stands out from the real GDP estimates from
column (3) in Table 3, in combination with the institutional changes that we report in column
(2) of the same table, is the direct relationship between improvements in institutional quality
and real GDP changes. Specically, whenever institutions improved, poor countries gained in
terms of real GDP, while real GDP decreased when institutional quality deteriorated between
1996 and 2006. We highlight this relationship with a visual representation in Figure 1, where
we plot the percentage change in real GDP responses against the observed institutional
quality changes in the poor countries in our sample between 1996 to 2006. As can be seen
from the gure, there is a strong positive relationship between real GDP percentage changes
37 Note however that the eects of institutional quality changes on third countries, i.e. on the rich countries
that are considered not to change their institutional quality, are quite small. The natural explanation for
this result is that the poor nations are small. Thus, even the direct impact on rich countries, through the
increase in their exports to the poor nations, is small.
28

and the changes in institutional quality in the poor countries. However, in addition, Figure
1 also reveals that the relationship between the changes in institutional quality and the
corresponding percentage changes in real GDP is not perfect.
As noted earlier, the heterogeneous response in prices and real GDP to changes in in-
stitutional quality is due to a complex system of direct and indirect (general equilibrium)
relationships that cannot be characterized analytically. Nevertheless, in what follows, guided
by our theory we attempt to shed some light on the determinants of the eectiveness of
institutional quality changes on the welfare of poor nations. To isolate the impact of hetero-
geneous institutional quality changes (i.e., the fact that some countries experienced larger
institutional quality changes than others), we complement our ex-post counterfactual anal-
ysis of the impact of actual institutional quality changes with a hypothetical experiment.
In such experiment, once again we start from the values in 2006 but, dierently from the
previous analysis, we simulate the response in real GDP to a uniform change in institutional
quality across all poor countries. In particular, we use as an initial shock the largest positive
institutional quality change that we observe in our sample over the period 1996-2006 which
is an improvement of 0.5 in the WGI index. Thus, our new experiment can be interpreted
as one where all poor countries in the sample improved their institutional quality uniformly
by 0.5 during the period of investigation.
Estimation results from the new counterfactual experiment are presented in columns (6)-
(9) of Table 3. As expected, all real GDP percentage changes for the poor countries in our
sample, which we report in column (7) of Table 3, are positive. This conrms the direct
relationship between improvements in institutional quality and welfare. More importantly
for the current purposes, we note that the real GDP eects for the poor countries are quite
heterogeneous (varying between 6.36 and 0.33, despite the fact that in this scenario all
real GDP changes for the poor countries are generated in response to same uniform initial
institutional quality change (see column (6) of Table 3). In order to understand the driving
forces underlying the heterogeneity in the response of real GDP to institutional quality
29

changes, we rely on theory and our intuition to identify several potential candidates that
may aect the eectiveness of institutional quality improvements in increasing welfare. For
clarity and expositional simplicity, and since this analysis is only suggestive, we use graphical
presentation of our ndings.
First, in Figure 2 we plot the real GDP changes for the poor countries that we obtained in
column (7) of Table 3 against the initial level of the institutional quality in 2006. The gure
does not seem to capture any signicant relationship. Next, in Figure 3 we investigate the
relationship between the real GDP changes of column (7) of Table 3 and the level of outputs
in 2006. For the sake of better visualization, but without loss of generality, we exclude four
outliers from this gure. Once again, this gure does not capture a clear pattern. Last,
Figure 4 plots the real GDP changes of column (7) of Table 3 against the trade openness in
2006. Here, we see a positive relationship: countries that are more open prot more form
institutional improvements.
Based on these reduced-form analyses, we conclude that the initial level of institutional
quality as well as the initial level of national output are not signicant factors in determin-
ing the eectiveness of institutional quality improvements. We view this as an encouraging
result because it suggest that poorer countries and countries with poor institutions may
benet equally from institutional improvements as countries that are richer and with bet-
ter institutional quality initially. In other words, the initial conditions across these two
dimensions do not matter much. Our second main conclusion, which has possible important
policy implications, is that trade openness promotes the impact of institutional improve-
ments and, therefore, there might be signicant benets from implementing a combination
of trade liberalization and institutional improvements.
30

4 Conclusions
Institutions are deep determinants of economic exchange, both within and across borders.
Most of the literature on how institutions aect trade ows, however, has not satisfacto-
rily dealt with endogeneity issues. We simply do not know for sure whether country-level
institutional quality has a causal eect on trade ows.
We have proposed a novel methodology that allows to estimate the eects of institutional
quality on international trade relative to intra-national trade in a structural gravity frame-
work with the appropriate set of xed eects: importer and exporter xed eects in cross-
section; importer-time, exporter-time and even country-pair xed eects in panels. Since
consistently with theoretical gravity we employ both international and intra-national
trade ows, our approach allows to identify the eects of national institutional quality on
international trade relative to internal trade. Beyond providing a natural and intuitive inter-
pretation of the econometric coecient of interest, there are two other invaluable advantages
of our approach.
First, it also deals with endogeneity issues, allowing to obtain consistent estimates of the
(relative) eects of national institutional quality. Second, it opens the ground to studies of
the trade impact of any country-specic variable that is not directional, i.e., that takes the
same value both for observations in which the country is an importer and for observations
in which the country is an exporter. Future research could therefore apply our methodology,
for instance, to behind-the-border measures such as value-added taxes or sales taxes, or to
measures of infrastructure and trade facilitation.
The impact of national institutional quality on international trade ows (relative to
domestic trade) that we estimate is positive, signicant and economically relevant. Such
impact is particularly strong for trade ows that represent imports of the poor countries
from the rich countries in our sample. These results are robust to the use of alternative
estimators, dierent samples, dierent measures of institutional quality, and to the use of
instrumental variables for institutional quality that are standard in the institutions and
31

growth literature.
Endowed with consistent estimates of the eects of national institutional quality on inter-
national trade relative to internal trade, we have presented general equilibrium quantication
of the welfare impact of counterfactual changes in institutional quality in the poor countries
in our sample over the period 1996-2006, based on their imports from rich countries. In a
rst counterfactual, we have simply considered observed changes. We have shown that the
welfare (measured by real GDP) of countries that signicantly improved their institutional
quality also increased signicantly, generally (but not in all cases) through consumer prices
falling more than producer prices. In a second counterfactual, we have considered hypothet-
ical changes in which all poor countries experience the same improvement in institutional
quality as the best performer over the period 1996-2006. We have shown that all real GDP
eects are positive, but quite heterogeneous.
Encouragingly, neither the initial level of institutional quality nor the initial level of na-
tional output seem to explain the heterogeneity in counterfactual welfare changes. The factor
that seems to matter most is the initial level of trade openness: countries that were more
open to trade in the baseline year (2006) experienced the largest real GDP counterfactual
changes. The important policy implication of this result is that poor countries stand to
gain relatively more from a combination of institutional reform and trade liberalization, as
opposed to any of these two policies in isolation.
References
Acemoglu, Daron and James A. Robinson, The Role of Institutions in Growth and
Development, Review of Economics and Institutions, 2010, 1 (2).
and Simon Johnson, Unbundling Institutions, Journal of Political Economy, 2005,
113 (5), 949995.
32

, , and James A. Robinson, The Colonial Origins of Comparative Development:
An Empirical Investigation, American Economic Review, 2001, 91 (5), 13691401.
, , and , The Rise of Europe: Atlantic Trade, Institutional Change, and Economic
Growth, American Economic Review, 2005, 95 (3), 546579.
Álvarez, Inmaculada C., Javier Barbero, Andrés Rodríguez-Pose, and José L.

Zofío, Does Institutional Quality Matter for Trade? Institutional Conditions in a Sectoral
Trade Framework, World Development, 2018, 103, 7287.
Amiti, Mary and Jozef Konings, Trade liberalization, intermediate inputs, and produc-
tivity: Evidence from Indonesia, American Economic Review, 2007, 97 (5), 16111638.
Anderson, James E. and Douglas Marcouiller, Insecurity and the Pattern of Trade:
An Empirical Investigation, Review of Economics and Statistics, 2002, 84 (2), 342352.
and Eric van Wincoop, Gravity with Gravitas: A Solution to the Border Puzzle,
American Economic Review, 2003, 93 (1), 170192.
and Yoto V. Yotov, The Changing Incidence of Geography, American Economic
Review, 2010, 100 (5), 21572186.
, Mario Larch, and Yoto V. Yotov, On the Growth Eects of the Transatlantic
Trade and Investment Partnership, 2016. Unpublished manuscript, available for down-
load at https://editorialexpress.com/cgi-bin/conference/download.cgi?db_name =EEAE-
SEM2016&paper_id=2227, February.
Anderson, Theodore W., Estimating Linear Restrictions on Regression Coecients for

Multivariate Normal Distributions, Annals of Mathematical Statistics, 1951, 22, 327351.
Arvis, Jean-Francois and Ben Shepherd, The Poisson Quasi-Maximum Likelihood
Estimator: A Solution to the Adding Up Problem in Gravity Models, Applied Economics
Letters, 2013, 20 (6), 515519.
33

Auer, Raphael A., Geography, Institutions, and the Making of Comparative Develop-
ment, Journal of Economic Growth, 2013, 18 (2), 179215.
Baier, Scott L. and Jerey H. Bergstrand, Do Free Trade Agreements Actually In-
crease Members' International Trade?, Journal of International Economics, 2007, 71 (1),
7295.
and , Bonus Vetus OLS: A Simple Method for Approximating International Trade-
cost Eects using the Gravity Equation, Journal of International Economics, 2009, 77
(1), 7785.
, Yoto V. Yotov, and Thomas Zylkin, On the Widely Diering Eects of Free Trade
Agreements: Lessons from Twenty Years of Trade Integration, 2016. Working Paper No.
2016-15, School of Economics Working Paper Series, LeBow College of Business, Drexel
University, October.
Bergstrand, Jerey H., Mario Larch, and Yoto V. Yotov, Economic Integration
Agreements, Border Eects, and Distance Elasticities in the Gravity Equation, European
Economic Review, 2015, 78, 307327.
Bernhofen, Daniel M. and John C. Brown, An Empirical Assessment of the Compar-
ative Advantage Gains from Trade: Evidence from Japan, American Economic Review,
2005, 95 (1), 208225.
Broda, Christian, Joshua Greeneld, and David E. Weinstein, From Groundnuts

to Globalization: A Structural Estimate of Trade and Growth, Research in Economics,
2017, 71 (4), 759783.
Brueckner, Markus and Daniel Lederman, Trade Openness and Economic Growth:
Panel Data Evidence from Sub-Saharan Africa, Economica, 2015, 82 (s1), 13021323.
34

Buera, Paco and Ezra Obereld, The Global Diusion of Ideas, 2016. NBER Working
Paper No. 21844, National Bureau of Economic Research, Cambridge, MA, January.
Bühler, Stefan, Marco Helm, and Michael Lechner, Trade Liberalization and
Growth: Plant-Level Evidence from Switzerland, 2011. Economics Working Paper No.
1133, School of Economics and Political Science, University of St. Gallen, August.
Coe, David T., Arvind Subramanian, Natalia T. Tamirisa, and Rikhil Bhav-
nani, The Missing Globalization Puzzle, 2002. IMF Working Paper No. WP/02/171,
International Monetary Fund, Washington, DC, October.
Costinot, Arnaud, On the Origins of Comparative Advantage, Journal of International
Economics, 2009, 77 (2), 255264.
Costinot, Arnaud and Andrés Rodríguez-Clare, Trade Theory with Numbers: Quan-
tifying the Consequences of Globalization, Chapter 4 in the Handbook of International
Economics Vol. 4, eds. Gita Gopinath, Elhanan Helpman, and Kenneth S. Rogo, Oxford:
Elsevier, 2014, pp. 197261.
Cragg, John C. and Stephen G. Donald, Testing Identiability and Specication in
Instrumental Variables Models, Econometric Theory, 1993, 9 (2), 222240.
Crozet, Matthieu, Pamina Koenig, and Vincent Rebeyrol, Exporting to Insecure
Markets: A Firm-Level Analysis, 2008. CEPII Working Paper No. 2008-13, Centre
d'Etudes Prospectives et d'Informations Internationales, Paris, September.
Cuñat, Alejandro and Marc J. Melitz, Volatility, Labor Market Flexibility, and the
Pattern of Comparative Advantage, Journal of the European Economic Association, 2012,
10 (2), 225254.
35

de Groot, Henri L. F., Gert-Jan Linders, Piet Rietveld, and Uma Subramanian,
The Institutional Determinants of Bilateral Trade Patterns, KYKLOS, 2004, 57 (1),
103123.
de Jong, Eelke and Christian Bogmans, Does Corruption Discourage International
Trade?, European Journal of Political Economy, 2011, 27 (2), 385398.
Disdier, Anne-Célia and Keith Head, The Puzzling Persistence of the Distance Eect
on Bilateral Trade, Review of Economics and Statistics, 2008, 90 (1), 3748.
Donaldson, Dave, The Gains from Market Integration, Annual Review of Economics,
2015, 7 (1), 619647.
Dutt, Pushan and Daniel Traca, Corruption and Bilateral Trade Flows: Extortion or
Evasion?, Review of Economics and Statistics, 2010, 92 (4), 843860.
Egger, Peter and Mario Larch, Interdependent Preferential Trade Agreement Member-
ships: An Empirical Analysis, Journal of International Economics, 2008, 76 (2), 384399.
Fally, Thibault, Structural Gravity and Fixed Eects, Journal of International Eco-
nomics, 2015, 97 (1), 7685.
Feenstra, Robert C., Advanced International Trade: Theory and Evidence, Princeton NJ:
Princeton University Press, 2004.
Feyrer, James, Trade and Income Exploiting Time Series in Geography, 2009. Un-
published manuscript, available for download at http://www.dartmouth.edu/~jfeyrer/
Feyrer_AirSea2009_10_21.pdf , October 2009.
, Distance, Trade, and Income The 1967 to 1975 Closing of the Suez Canal as a
Natural Experiment, 2011. Unpublished manuscript, available for download at https:
//economicdynamics.org/meetpapers/2011/paper_1438.pdf , March 2011.
36

Francois, Joseph and Miriam Manchin, Institutions, Infrastructure, and Trade, World
Development, 2013, 46, 165175.
Frankel, Jerey A. and David H. Romer, Does Trade Cause Growth?, American
Economic Review, 1999, 89 (3), 379399.
Gil-Pareja, Salvador, Rafael Llorca-Rivero, and José Antonio Martínez-Serrano,

Corruption and International Trade: A Comprehensive Analysis with Gravity, Working
Paper WPAE-2017-05, University of Valencia, 2017.
Hall, Robert E. and Charles I. Jones, Why do Some Countries Produce So Much
More Output Per Worker than Others?, Quarterly Journal of Economics, 1999, 114 (1),
83116.
Head, Keith and Thierry Mayer, Gravity Equations: Workhorse, Toolkit, and Cook-
book, Chapter 3 in the Handbook of International Economics Vol. 4, eds. Gita Gopinath,
Elhanan Helpman, and Kenneth S. Rogo, Oxford: Elsevier, 2014, pp. 131195.
Heid, Benedikt, Mario Larch, and Yoto V. Yotov, Estimating the Eects of Non-
discriminatory Trade Policies within Structural Gravity Models, CESifo Working Paper
No. 6735, 2017.
Helpman, Elhanan and Oleg Itskhoki, Labor Market Rigidities, Trade and Unemploy-
ment, Review of Economic Studies, 2010, 77 (3), 11001137.
Hummels, David L., Toward a Geography of Trade Costs, 2001. Unpublished
manuscript, available for download at http://www.krannert.purdue.edu/faculty/ hum-
melsd/research/toward/tgtc.pdf, September.
Impullitti, Giammario and Omar Licandro, Trade, Firm Selection, and Innovation:
The Competition Channel, Economic Journal, 2017, in press.
37

Inikori, Joseph, The Struggle Against the Atlantic Slave Trade: The Role of the State,
Chapter 11 in Fighting the Slave Trade: West African Strategies, ed. Sylviane A. Diouf,
Athens, OH: Ohio University Press, 2003, pp. 170198.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W.

Vishny, Legal Determinants of External Finance, Journal of Finance, 1997, 52 (3),
11311150.
Larch, Mario, Joschka Wanner, Yoto V. Yotov, and Thomas Zylkin, The Currency
Union Eect: A PPML Re-assessment with High-Dimensional Fixed Eects, School of
Economics Working Paper Series 2017-7, LeBow College of Business, Drexel University
April 2017.
Manova, Kalina, Credit Constraints, Equity Market Liberalizations and International
Trade, Journal of International Economics, 2008, 76 (1), 3347.
Mayer, Thierry and Soledad Zignago, Notes on CEPII's Distances Measures: The
GeoDist Database, 2011. CEPII Working Paper No. 2011-25, Centre d'Etudes Prospec-
tives et d'Informations Internationales, Paris, December.
Nizalova, Olena Y. and Irina Murtazashvili, Exogenous Treatment and Endogenous

Factors: Vanishing of Omitted Variable Bias on the Interaction Term, Journal of Econo-
metric Methods, 2016, 5 (1).
North, Douglass C., Structure and Change in Economic History, New York: W W Norton
& Co. Inc., 1981.
, Institutions, Institutional Change and Economic Performance, Cambdridge (UK): Cam-
bridge University Press, 1990.
Nunn, Nathan, Relationship-Specicity, Incomplete Contracts, and the Pattern of Trade,

Quarterly Journal of Economics, 2007, 122 (2), 569600.
38

, The Importance of History for Economic Development, Annual Review of Economics,
2009, 1 (1), 6592.
and Daniel Treer, Domestic Institutions as a Source of Comparative Advantage,
Chapter 5 in the Handbook of International Economics Vol. 4, eds. Gita Gopinath, Elhanan
Helpman, and Kenneth S. Rogo, Oxford: Elsevier, 2014, pp. 263316.
and Leonard Wantchekon, The Slave Trade and the Origins of Mistrust in Africa,
American Economic Review, 2011, 101 (7), 32213252.
Pascali, Luigi, The Wind of Change: Maritime Technology, Trade, and Economic Devel-
opment, American Economic Review, 2017, 107 (9), 28212854.
Puga, Diego and Daniel Treer, International Trade and Institutional Change: Me-
dieval Venice's Response to Globalization, Quarterly Journal of Economics, 2014, 129
(2), 753821.
Santos Silva, João M. C. and Silvana Tenreyro, The Log of Gravity, Review of
Economics and Statistics, 2006, 88 (4), 641658.
Stock, James H. and Motohiro Yogo, Testing for Weak Instruments in Linear IV
Regression, in Donald W. K. A Andrews and James H. Stock, eds., Identication and
Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, Cambridge
(UK): Cambridge University Press, 2005, pp. 80108.
Svensson, Jakob, Eight Questions about Corruption, Journal of Economic Perspectives,
2005, 19 (3), 1942.
Wacziarg, Romain and Karen Horn Welch, Trade Liberalization and Growth: New
Evidence, World Bank Economic Review, 2008, 22 (2), 187231.
Yotov, Yoto V., Roberta Piermartini, José-Antonio Monteiro, and Mario Larch,
An Advanced Guide to Trade Policy Analysis: The Structural Gravity Model, Geneva,
39

Switzerland: United Nations and World Trade Organization. Available for download at
https://vi.unctad.org/tpa/web/vol2/vol2home.html, 2016.
Yu, Shu, Sjoerd Beugelsdijk, and Jakob de Haan, Trade, Trust and the Rule of Law,
European Journal of Political Economy, 2015, 37, 102115.
40

Tables and Figures
Table 1: National institutions and international trade

(1) (2) (3) (4) (5) (6)
OLS IQ IV_OLS IV_PPML PNL_GRV PNL_FES
LN_DIST -1.226 -1.223 -1.223 -1.224 -1.241
(0.061)∗∗ (0.057)∗∗ (0.043) ∗∗ (0.042) ∗∗ (0.049)∗∗
CNTG 0.196 0.195 0.195 0.474 0.169
(0.223) (0.226) (0.154) (0.157) ∗∗ (0.217)
LANG 0.718 0.731 0.730 0.614 0.720
(0.110)∗∗ (0.109)∗∗ (0.080)∗∗ (0.084)∗∗ (0.094)∗∗
CLNY 0.500 0.469 0.472 0.624 0.478
(0.158)∗∗ (0.157)∗∗ (0.154)∗∗ (0.131)∗∗ (0.145)∗∗
RTA 0.193 0.185 0.186 0.245 0.079 0.039
(0.075)∗∗ (0.073)∗ (0.062)∗∗ (0.063)∗∗ (0.057) (0.050)
BRDR -3.466 -4.256 -4.197 -4.391 -4.638
(0.361)∗∗ (0.345)∗∗ (0.295) ∗∗ (0.275) ∗∗ (0.297)∗∗
IQ_BRDR 1.906 1.764 1.015 2.011 0.616
(0.291)∗∗ (0.443)∗∗ (0.485)∗ (0.277)∗∗ (0.301)∗
BRDR_1998 0.118 0.142
(0.062)+ (0.059)∗
BRDR_2000 0.098 0.141
(0.073) (0.066)∗
BRDR_2002 0.136 0.142
(0.085) (0.073)+
BRDR_2003 0.251 0.241
(0.088)∗∗ (0.078)∗∗
BRDR_2004 0.292 0.287
(0.086)∗∗ (0.078)∗∗
BRDR_2005 0.361 0.341
(0.093)∗∗ (0.090)∗∗
BRDR_2006 0.438 0.402
(0.093)∗∗ (0.089)∗∗
N 3880 3880 3880 3969 30753 30753
R2 0.864 0.868 0.868 0.869 0.932
Notes : This table reports estimation results from a series of econometric models that study the
impact of national institutions on international trade. All estimates are obtained with exporter
and importer xed eects, which also vary over time in the panel specications. Estimates of the
xed eects, including the constant term, are omitted for brevity. Column (1) reports standard
gravity estimates that are obtained with the OLS estimator and data for 2006. Column (2)
replicates the specication from column (1) after adding a measure of institutional quality. Column
(3) employs an instrumental variable estimator to re-evaluate the results from column (2). Column
(4) replicates the estimates from column (3) with an IV PPML estimator. Finally, columns
(5) and (6) report OLS estimates that are obtained with panel data over the period 1996-2006
with standard gravity variables (in column 5) and with country-pair xed eects (in column 6),
+
respectively. Standard errors, clustered by country pair, are reported in parentheses. p < 0.10,
∗ ∗∗
p < .05, p < .01. See text for further details.
41

Table 2: Institutions and international trade of poor nations
Panel A: Gravity Variables Panel B: Fixed Eects
(1) (2) (3) (4)
MAIN POOR_RICH MAIN POOR_RICH
LN_DIST -1.241 -1.243
∗∗ ∗∗
(0.049) (0.049)
CNTG 0.169 0.156
(0.217) (0.216)
LANG 0.720 0.724
∗∗ ∗∗
(0.094) (0.093)
CLNY 0.478 0.477
∗∗ ∗∗
(0.145) (0.146)
RTA 0.079 0.051 0.039 0.028
(0.057) (0.058) (0.050) (0.051)
BRDR -4.638 -4.623
∗∗ ∗∗
(0.297) (0.299)
IQ_BRDR 2.011 1.984 0.616 0.608
∗∗ ∗∗ ∗ ∗
(0.277) (0.277) (0.301) (0.308)
IQP _BRDRP R 1.919 0.436
∗∗
(0.312) (0.314)
IQP _BRDRRP 2.277 0.813
∗∗ ∗∗
(0.306) (0.314)
N 30753 30753 30753 30753
R2 0.869 0.869 0.932 0.932
Notes : This table reports estimation results from econometric models that study the dierential
impact of national institutions on international trade of poor countries with rich countries. All
estimates are obtained in panel settings with exporter-time and importer-time xed eects. Es-
timates of the xed eects, including the constant term, as well as estimates of all time-varying
border variables, are omitted for brevity. Column (1) is a replication of column (5) from Table
1, which reports estimates of the eect of institutions on trade that are obtained with standard
gravity variables, the OLS estimator, and panel data. Column (2) replicates the specication
from column (1) after adding border and institution variables for exports and imports between
the poor and the rich countries in our sample. The estimates in column (3) reproduce the results
from column (6) of Table 1, which are obtained with country-pair xed eects, the OLS estimator,
and panel data. Finally, column (4) adds border and institution variables for exports and imports
between the poor and the rich countries in our sample. Standard errors, clustered by country
+ ∗ ∗∗
pair, are reported in parentheses. p < 0.10, p < .05, p < .01. See text for further details.
42

Table 3: Institutions, trade and development: GE analysis
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: Observed ∆IQ 1996-2006 Panel B: Uniform IQ Change

Country ∆Inst %∆rGDP %∆p %∆IMR ∆Inst %∆rGDP %∆p %∆IMR
ARG 0.00 0.02 -0.28 -0.30 0.00 -0.04 0.68 0.72
AUS 0.00 0.01 -0.38 -0.39 0.00 -0.02 0.91 0.93
AUT 0.00 0.01 -0.17 -0.18 0.00 -0.02 0.40 0.42
BEL 0.00 0.01 -0.21 -0.22 0.00 -0.01 0.50 0.51
BGR 0.00 0.01 -0.20 -0.20 0.00 -0.01 0.48 0.49
BOL -0.36 -1.68 0.59 2.31 0.50 2.52 -0.53 -2.98
BRA 0.00 0.01 -0.29 -0.30 0.00 -0.01 0.69 0.70
CAN 0.00 0.00 -0.34 -0.34 0.00 -0.01 0.81 0.83
CHE 0.00 0.00 -0.21 -0.21 0.00 -0.01 0.49 0.50
CHL 0.00 0.01 -0.31 -0.32 0.00 -0.03 0.72 0.75
CHN -0.14 -0.42 -0.86 -0.45 0.50 1.75 2.71 0.94
CMR 0.27 1.60 -0.31 -1.87 0.50 4.06 1.24 -2.71
COL 0.18 0.62 -0.70 -1.31 0.50 2.19 -0.02 -2.17
CRI 0.00 0.00 -0.33 -0.33 0.00 -0.01 0.76 0.77
CYP 0.00 -0.00 -0.24 -0.24 0.00 -0.00 0.56 0.56
DEU 0.00 -0.12 -0.12 0.00 0.00 0.28 0.28 0.00
DNK 0.00 0.01 -0.21 -0.22 0.00 -0.01 0.51 0.52
ECU -0.31 -1.33 0.59 1.94 0.50 2.25 -0.88 -3.07
EGY -0.37 -1.93 -0.04 1.92 0.50 2.79 0.43 -2.30
ESP 0.00 0.01 -0.21 -0.22 0.00 -0.01 0.50 0.51
FIN 0.00 0.01 -0.23 -0.24 0.00 -0.02 0.56 0.58
FRA 0.00 0.00 -0.22 -0.22 0.00 -0.02 0.51 0.53
GBR 0.00 0.01 -0.24 -0.25 0.00 -0.01 0.57 0.58
GRC 0.00 0.01 -0.21 -0.22 0.00 -0.01 0.51 0.52
HUN 0.00 0.01 -0.17 -0.18 0.00 -0.02 0.42 0.44
IDN -0.10 -0.56 -0.66 -0.09 0.50 3.37 2.58 -0.76
IND 0.02 0.06 -0.28 -0.34 0.50 1.42 1.12 -0.30
IRL 0.00 0.00 -0.25 -0.25 0.00 -0.00 0.60 0.60
IRN -0.18 -0.27 0.75 1.02 0.50 0.88 -1.89 -2.75
ISR 0.00 0.00 -0.28 -0.28 0.00 -0.01 0.65 0.66
ITA 0.00 0.01 -0.21 -0.22 0.00 -0.02 0.50 0.52
JOR -0.04 -0.05 -0.17 -0.12 0.50 1.07 -0.66 -1.71
JPN 0.00 0.02 -0.35 -0.37 0.00 -0.04 0.88 0.92
KEN 0.07 0.10 -0.88 -0.98 0.50 1.12 -2.62 -3.69
KOR 0.00 0.02 -0.36 -0.38 0.00 -0.05 0.89 0.94
KWT 0.00 0.02 -0.33 -0.35 0.00 -0.05 0.80 0.85
LKA 0.02 0.08 -0.29 -0.37 0.50 2.07 1.15 -0.90
MAR -0.31 -2.23 -0.65 1.62 0.50 3.95 1.40 -2.45
MEX 0.00 -0.00 -0.34 -0.34 0.00 -0.00 0.80 0.80
MUS 0.00 -0.00 -0.26 -0.25 0.00 -0.02 0.56 0.58
MWI -0.01 -0.05 -0.30 -0.24 0.50 2.43 0.01 -2.36
Continued on next page
43

Table 3 Continued from previous page
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Country ∆Inst %∆rGDP %∆p %∆IMR ∆Inst %∆rGDP %∆p %∆IMR

MYS 0.00 0.05 -0.34 -0.39 0.00 -0.09 0.83 0.92
NER 0.50 4.68 -0.43 -4.88 0.50 6.36 1.71 -4.37
NGA 0.06 0.03 -0.86 -0.89 0.50 0.33 -3.33 -3.64
NLD 0.00 0.02 -0.21 -0.23 0.00 -0.03 0.50 0.53
NOR 0.00 -0.00 -0.24 -0.24 0.00 -0.00 0.57 0.57
NPL -0.66 -1.45 -0.10 1.37 0.50 0.94 0.43 -0.50
PAN 0.00 0.02 -0.32 -0.34 0.00 -0.04 0.77 0.82
PHL -0.47 -5.23 -3.04 2.31 0.50 6.05 3.90 -2.02
POL 0.00 0.00 -0.18 -0.18 0.00 -0.01 0.42 0.43
PRT 0.00 0.01 -0.20 -0.21 0.00 -0.01 0.48 0.49
ROM 0.00 0.00 -0.20 -0.20 0.00 -0.01 0.47 0.48
SEN -0.04 -0.07 0.03 0.10 0.50 1.20 -2.88 -4.03
SGP 0.00 0.08 -0.33 -0.41 0.00 -0.16 0.81 0.97
SWE 0.00 -0.00 -0.23 -0.23 0.00 -0.01 0.54 0.55
THA -0.53 -4.16 -2.62 1.60 0.50 4.07 3.05 -0.98
TTO 0.00 -0.01 -0.33 -0.32 0.00 0.00 0.74 0.74
TUN 0.02 0.19 -0.14 -0.33 0.50 5.15 2.32 -2.69
TUR 0.00 0.01 -0.20 -0.21 0.00 -0.02 0.48 0.50
TZA 0.33 1.84 -1.22 -3.00 0.50 3.80 0.09 -3.57
URY 0.00 0.01 -0.29 -0.30 0.00 -0.01 0.74 0.75
USA 0.00 0.00 -0.34 -0.34 0.00 -0.01 0.81 0.82
ZAF 0.00 -0.00 -0.29 -0.29 0.00 -0.02 0.63 0.65
Notes : This table reports results from our counterfactual analyses. Column (1) lists the
ISO country codes (in bold for poor countries). Columns (2) and (6) report the changes in
institutional quality that are used to generate the corresponding general equilibrium eects.
Columns (3) to (5) report the results from an ex-post evaluation of the impact of institutional
quality in poor countries during the period (1996-2006), through changes in their imports
from rich nations. Column (3) presents real GDP percentage changes, while columns (4)
and (5) decompose their incidence on the consumers and on the producers in the world.
Columns (7)-(9) report results that correspond to the estimates in columns (3)-(5), but this
time in response to a uniform increase in the institutional quality in the poor nations in our
sample, as depicted in column (6). See text for further details.
44

Figure 1: Impact of change in institutions on real GDP
5
NER
real GDP change (in %)

TZA
CMR
COL
TUN
KEN
LKA
IND
NGA
JOR
MWI
SEN
0
IRN
CHN
IDN
NPL ECU
BOL
EGY
MAR
THA
-5
PHL
-1 -.5 0 .5
change in institutions
Notes : This gure plots, for the directly aected countries, the
changes in real GDP obtained in column (3) of Table 3 against the
observed changes from 1996 to 2006 of the institutions variable.
The considered counterfactual scenario is the one of column (2)
of Table 3, where the change in the institutions variable for im-
ports of the poor countries from the rich countries is equal to the
observed changes from 1996 to 2006. See text for further details.
Figure 2: Impact of initial level of institutions on real GDP
NER
PHL
6
TUN
CMR THA
4
MAR
TZA
IDN
EGY
BOL MWI
ECU COL
LKA
2
CHN
IND
KEN SEN
JOR
IRN NPL
NGA
0
-1 -.5 0
level of institutions in 2006
observed level of the institutions variable in 2006. The considered
counterfactual scenario is the one of column (6) of Table 3, where
the change in the institutions variable for imports of the poor
countries from the rich countries is uniform across poor countries
and equal to the one of the best performer (0.50) from 1996 to
2006. See text for further details.
45

Figure 3: Impact of initial level of output on real GDP
NER
PHL
6
TUN

CMR
4
MAR
TZA
EGY
BOL
MWI
ECU COL
LKA
2
SENKEN
JOR
NPL IRN
NGA
0
0 20000 40000 60000 80000 100000

output in 2006
observed level of output in 2006. The considered counterfactual
scenario is the one of column (6) of Table 3, where the change
in the institutions variable for imports of the poor countries from
the rich countries is uniform across poor countries and equal to
the one of the best performer (0.50) from 1996 to 2006. See text
for further details.
Figure 4: Impact of initial level of trade openness on real GDP
NER
PHL
6
TUN
THA CMR
4
MAR
TZA
IDN
EGY
MWI BOL
COL ECU
LKA
2
CHN
IND
KEN SEN
JOR
IRN NPL
NGA
0
0 .5 1 1.5 2
trade openness (exports+imports over output)
observed level of trade openness in 2006. The considered coun-
terfactual scenario is the one of column (6) of Table 3, where the
change in the institutions variable for imports of the poor coun-
tries from the rich countries is uniform across poor countries and
equal to the one of the best performer (0.50) from 1996 to 2006.
See text for further details.
46

Appendices
A Identication of country-specic institutions in gravity

We present the identication challenges and the solutions that we propose with a specic
data example. Assume availability of a cross-section bilateral trade data set with trade ows
between three countries {A, B, C}, which, as traditionally used in the gravity literature,
includes data on international trade ows but no data on intra-national trade ows. The
goal is to show whether, and under what circumstances, one can identify the impact of
country-specic national institutions and, ideally, to separate the impact on the importer
and on the exporter side. The relevant part of estimation data set looks as follows:
# i j η1 η2 µ1 µ2 µ3 IQj
1 A B 1 0 0 1 0 IQB 
2 A C 1 0 0 0 1 IQC 
 
3 B A 0
 1 1 0 0 IQA 
4 B C 0
 1 0 0 1 IQC  (A-1)
5 C A 0 0 1 0 0 IQ 
A
6 C B 0 0 0 1 0 IQB
Column (1) of matrix (A-1) numbers the observations. Columns i and j denote the generic
exporting and importing country. Columns η 's and µ's denote the exporter and the importer
xed eects, respectively. As with any dummy variable, we have to omit one category from
the xed eects and, without loss of generality, we drop the xed eect η3 , i.e. the xed
eect for country C as an exporter. Finally, for expositional simplicity, at this stage, we
only include one of the trade cost covariates, namely the institution variable on the importer
side (IQj ), which is of central interest to us. First, we demonstrate the obvious result that
estimations with data sets which only include international trade ows and control for the
structural MRTs with exporter and importer xed eects are unable to deliver estimates of
the eects of any country-specic determinant of trade. To see this formally in the case of
national institutions, note that IQj can be expressed as a linear combination of the xed
eects in matrix (A-1):
IQj = µ1 IQA + µ2 IQB + µ3 IQC . (A-2)
Equation (A-2) reveals that the importer institutions variable (IQj ) is perfectly collinear with
the importer xed eects and, therefore, the coecient of this variable cannot be identied
47

separately. Identical intuition and argumentation applies in the case of institutions on the
exporter side.
Next, we follow the recommendation of Heid et al. (2017) to estimate gravity equations
with intra-national as well as international trade ows. As we demonstrate next, similar to
Heid et al. (2017), who target identication of the impact of non-discriminatory trade policies,
we can show that the introduction of intra-national trade ows allows for identication of
the impact of country-specic national institutions. However, unlike Heid et al. (2017), who
are able to identify the impact of unilateral trade policies separately on the importer and
on the exporter side, we also demonstrate that it is not possible to identify the impact of
national institutions (and, in general, of any country-specic variable) separately on exports
versus imports. The reason is that unilateral trade policies are potentially directional, i.e.,
specic for a country as importer or exporter, while the institutional quality is the same for
a country, independently whether it acts as an exporter or importer.
Instead, we show that it is only possible to obtain estimates of the impact of national
institutional quality on international relative to internal trade. To see these arguments,
consider the following updated version of the data matrix (A-1):
# i j η1 η2 µ1 µ2 µ3 BRDRij IQj × BRDRij

1 A B 1 0 0 1 0 1 IQB 
2 A C 1 0 0 0 1 1 IQC 
3 B A 0 1 1 0 0 1 IQA
 

 
4 B C 0
 1 0 0 1 1 IQC 

5 C A 0 0 1 0 0 1 IQA 
(A-3)
 
6 C B 0
 0 0 1 0 1 IQB 

7 A A 1 0 1 0 0 0 0
 

8 B B 0 1 0 1 0 0 0
 
9 C C 0 0 0 0 1 0 0
To obtain matrix (A-3), we added observations that correspond to intra-national trade.
In addition, we introduced a dummy variable BRDRij , which takes a value of one for
international trade, and it is equal to zero for intra-national trade. Finally, we re-dened
the institutional variable as the product between the original institution variable and the
border dummy, IQj × BRDRij . Thus, by construction, the estimate of the coecient on
IQj ×BRDRij should capture the dierential impact of importer institutions on international
48

trade relative to domestic trade.
We demonstrate the linear independence of IQj × BRDRij from the country-specic
xed eects by contradiction. If IQj × BRDRij were perfectly collinear with the dummies,
we should be able to express it as a linear combination of them:
IQj × BRDRij = α1 η1 + α2 η2 + α3 µ1 + α4 µ2 + α5 µ3 + α6 BRDRij , (A-4)
where α1 to α6 denote coecients to be estimated. Focus on specic rows. First, from
observation 9 in matrix (A-3) it follows that α5 has to be equal to zero. To fulll equation
(A-4) for observation 8, α2 = −α4 has to hold. From observation 7 it follows that α1 = −α3 .
Using these relationships, we can express equation (A-4) in matrix form as:
# i j α1 η1 + α2 η2 + α3 µ 1 + α4 µ2 + α5 µ3 + α6 BRDRij = IQj × BRDRij

1 A B  α1  +  0  +  0  +  −α2  +  0  +  α6  =  IQB 
2 A C  α1  +  0  +  0  +  0  +  0  +  α6  =  IQC 
3 B A  0  +  α  +  −α  +  0  +  0  + α6 = IQA
 2  1
   
          
4 B C  0 
  +  α 
 2  +  0 
  +  0 
  +  0 
  + 
 α6 
 = 
 IQC 

5 C A  0  +  0  +  −α1  +  0  +  0  + α6 = IQA
             
   
 −α2  (A-5)
             
6 C B  0  +  0  +  0  + +  0  +  α6  =  IQB 
             
7 A A  α1  +  0  +  −α1  +  0  +  0  +  0  =  0 
             
8 B B 0 + α2 + 0 + −α2 + 0 + 0 = 0
9 C C 0 + 0 + 0 + 0 + 0 + 0 = 0
It is obvious that equations corresponding to observations 7 to 9 in (A-5) fulll the condition
for perfect collinearity. Lines 1 and 6 can only simultaneously hold if α1 = 0. However, if
α1 = 0, it follows that the equations given in lines 2 and 4 can only simultaneously hold
if α2 = 0. Then, the only non-zero coecient left is α6 . In order to satisfy equations 1 to
6, α6 would have to take three dierent values. Hence, after ruling out trivial solutions for
multicollinearity due to missing variation in IQj × BRDRij , the only solution for the above
system of equations is α1 = ... = α6 = 0. Thus, we have proven that the variable IQj ×
BRDRij is linearly independent from the set of exporter and importer dummies when intra-
national trade ows observations are included in the estimating sample. Therefore, IQj ×
BRDRij can indeed be used to identify the eect of national institutions on international
trade. Identical analysis can be used to demonstrate that, instead of identifying the impact
of importer institutions on trade, we can identify the impact of exporter institutions on
trade.
49

It is tempting to assume that similar intuition and analysis can be used to show that
one can identify simultaneously the dierential eects on trade of importer versus exporter
institutions. Unfortunately, as we demonstrate next, this is not possible. To make this
argument clear, we further extend our data matrix to include an interaction term IQi ×
BRDRij that would capture the impact of national institutions on exports:
# i j η1 η2 µ1 µ2 µ3 BRDRij IQi × BRDRij IQj × BRDRij

1 A B 1 0 0 1 0 1 IQA IQB 
2 A C 1 0 0 0 1 1 IQA IQC 
3 B A 0 1 1 0 0 1 IQB IQA
 

 
4 B C 0
 1 0 0 1 1 IQB IQC 

5 C A 0 0 1 0 0 1 IQC IQA 
(A-6)
 
6 C B 0
 0 0 1 0 1 IQC IQB 

7 A A 1 0 1 0 0 0 0 0
 

8 B B 0 1 0 1 0 0 0 0
 
9 C C 0 0 0 0 1 0 0 0
Perfect collinearity would hamper identication if IQj × BRDRij can be expressed as a
linear combination of the dummies and IQi × BRDRij :
IQj × BRDRij = α1 η1 + α2 η2 + α3 µ1 + α4 µ2 + α5 µ3 + α6 BRDRij + α7 IQi × BRDRij . (A-7)
Focus on the observations in matrix (A-6). From observation 9 in matrix (A-6) it follows
that α5 has to be equal to zero. To fulll equation (A-7) for observation 8, α2 has to be
equal to −α4 , and from observation 7 it follows that α1 = −α3 . Using these relationships,
we can re-express equation (A-7) in matrix form as follows:
1 A B α1 + 0 + 0 + −α2 + 0 + α6 + α7 IQA = IQB

               
2 A C α1 
  + 0
  +  0 
  +  0 
  + 0
  + α6 
  + α7 IQA 
  = IQC 
 
3 B A 0 + α2  + −α1  +  0  + 0 + α6  + α7 IQB  = IQA 
               
4 B C 0
  + α 
 2 +  0 
  +  0 
  + 0
  + α 
 6 + α IQ 
 7 B = IQ 
 C
5 C A 0 + 0 + −α1  +  0  + 0 + α6  + α7 IQC  = IQA 
               
               
6 C B 0 + 0 +  0  + −α2  + 0 + α6  + α7 IQC  = IQB  (A-8)
               
7 A A α1  + 0 + −α1  +  0  + 0 + 0 +  0  =  0 
               
8 B B 0 + α  +  0  + −α  + 0 + 0 +  0  =  0 
2 2
9 C C 0 + 0 + 0 + 0 + 0 + 0 + 0 = 0
Subtract observation 6 from observation 1 in matrix (A-8) to obtain α1 = α7 (IQC − IQA ).

Subtract observation 5 from observation 3 to obtain α2 = α7 (IQC − IQB ). These two
50

expressions for α1 and α2 along with setting α7 = 1 and α6 = 0 are enough to express
IQi × BRDRij as function of the exporter and importer dummies and IQj × BRDRij .38
Hence, even with intra-national trade ows, one cannot simultaneously obtain separate eects
of the impact of institutions for exports and for imports. Instead, one can only identify
the impact of one of the two institution variables at a time. Importantly, however, since
IQi ×BRDRij and IQj ×BRDRij are perfectly collinear, the corresponding estimates on the
two institution variables from separate estimations will be identical and each of them should
be interpreted as the sum of the eect of institutions on exports and imports (international
trade) relative to intra-national trade. Therefore, in the subsequent empirical analysis we
dene the institutional regressor as IQ × BRDRij , where IQ can either be the institution
index on the importer side IQj or the institution index on the exporter side IQi .39
The analysis and discussion that we presented in this section naturally carry over to a
setting with panel data. Intuitively, one can view the panel setting as sequence of cross-
section data sets such as the one used in the presentation so far. More interesting from an
econometric perspective, a panel setting would enable us to employ country-pair xed eects.
Importantly, the introduction of country-pair xed eects would not lead to any additional
collinearity challenges with the identication of the impact of national institutions. To see
that this is indeed the case, one can consider a simple panel setting with two periods and
then take rst dierences. The rst dierencing would wipe out the bilateral xed eects
while the remaining variables can be expressed as changes between the two periods. For
identication purposes, the resulting system collapses to the cross-section setting that we
presented earlier, where, instead of levels the system includes the changes of the variables.
This however, has no impact and implications for any of our identication arguments.
38 This can easily be checked by plugging in these expressions for α and α into equation (A-7).
1 2
39 To gain further intuition for this result, assume that instead of dening BRDR as dummy taking value
ij
one for international trade, we dened a dummy IN T RAii that is equal to one for intra-national trade, i.e.,
IN T RAii = 1 − BRDRij , and BRDR and IN T RA both capture deviations between international and
intra-national trade. They are merely dierently dened dummies. However, using IN T RA, it is obvious
that we can not identify dierential eects of institutions for importers and exporters.
51

B Robustness analysis
This Appendix reports the results from a series of sensitivity experiments that are designed
to test the robustness of our methods and main ndings.
Table B-1 reports estimates from three alternative specications depending on the deni-
tion of the interaction between the institution quality index and the border dummy variable.
The estimates from column (1) of Table B-1 reproduce our main results, which are obtained
with an institution variable on the exporter side dened as IQi × BRDRij , which we now
explicitly label W GI _EXP ORT ER. The estimates in column (2) of Table B-1 are ob-
tained with an institutional variable dened on the importer side as W GI _IM P ORT ER ≡
IQj × BRDRij . Consistent with the analysis from Appendix A, the estimates from columns
(1) and (2) of Table B-1 are identical. The estimates in column (3) of Table B-1 are ob-
tained with an institutional variable dened as W GI _IN T RAN AT L ≡ IQi × IN T RAii ,

where IN T RAii is an indicator variable that is equal to one for intra-national trade, i.e.,
IN T RAii ≡ 1 − BRDRij . As expected (see footnote 39), the estimates from column (3) are
exactly the opposite as compared to those from columns (1) and (2). Thus, the results from
Table B-1 support empirically our theoretical analysis from Appendix A.
Table B-2 reproduces our main estimates with alternative measures of institutional qual-
ity. For comparison purposes, the estimates in column (1) of Table B-2 replicate our main
results from column (2) of Table 1. Then, in columns (2)-(7), we use the individual institu-
tional quality indexes from the World Bank's World Governance Indicators database, which
include Voice and Accountability (in column (2)), Political Stability and Absence of Violence
(in column (3)), Government Eectiveness (in column (4)), Regulatory Quality (in column
(5)), Rule of Law (in column (6)), and Control of Corruption (in column (7)). Without any
exception, we obtain positive, large, and statistically signicant estimates for each individ-
ual WGI indicator. The WGI estimates vary between 1.356 (std.err. 0.359) for Voice and
Accountability and 2.003 (std.err. 0.277) for Regulatory Quality and, therefore, they conrm
and support our main ndings.
52

The estimates from columns (8) and (9) of Table B-2 are obtained with institution mea-
sures from two alternative databases. Specically, we employ the Combined Polity Score
index from the Polity IV project to obtain the estimates in column (8). The estimate on
the new institution covariate P OLIT Y _BRDR is positive, however, it is not statistically
signicant. We oer a possible explanation for this result when we discuss the ndings from
our next experiment, reported in Table B-3. The estimates from column (9) are obtained
with an average institution index that is constructed from the Political Rights and Civil
Liberties indicators of the Freedom House initiative. The estimate on F RDM HS _BRDR
is positive, signicant and comparable to the rest of the institution estimates from Table
B-2.
An advantage of the Polity IV and the Freedom House indexes of institutional quality is
that they are available for an extended period of time, which enables us to study the impact
of institutions over the period 1988-2006. Table B-3 capitalizes on the longer time coverage
of our sample when the institutional quality indexes are constructed based on the Polity IV
and the Freedom House databases. Columns (1)-(3) obtain estimates with standard gravity
variables. The estimates of the exporter-time and of the importer-time xed eects, as well
as the estimates of the border dummies are omitted for brevity. For comparison purposes,
column (1) of Table B-3 replicates the panel results from column (5) of Table 1 with the
main WGI institution variable. Columns (2) and (3) obtain estimates with the institution
indexes from the Polity IV project and from the Freedom House database, respectively.
While the estimates on P OLIT Y _BRDR (in column (2)) and F RDM HS _BRDR (in
column (3)) are smaller in magnitude as compared to the main estimate on W GI _BRDR
(column (1)), we nd that the estimates on the new institution variables in Table B-3 are
large, positive, and statistically signicant. Columns (4)-(6) of Table B-3 conrm the results
from columns (1)-(3) with country-pair xed eects. As before, we nd that the estimates on
the institution variables are signicantly smaller once bilateral xed eects are added to the
estimating specication. However, once again, we obtain positive and statistically signicant
53

estimates on the two new institution variables. In fact, the estimates on P OLIT Y _BRDR
and F RDM HS _BRDR from Table B-3 are more precisely estimated as compared to the
corresponding estimate on W GI _BRDR. Furthermore, we see that the insignicant cross-
section estimate on P OLIT Y _BRDR from column (8) of Table B-2 has become signicant
with the panel data in Table B-3. The larger time period, which allows for more variation
in the institution indexes, is a natural explanation for the signicant and more precisely
estimated eects on P OLIT Y _BRDR and F RDM HS _BRDR.

The estimates in Table B-4 demonstrate the robustness of our main results to the use of
alternative instrumental variables. Columns (1)-(3) of Table B-4 report IV-OLS estimates.
For comparison purposes, column (1) of Table B-4 replicates the IV-OLS results from column
(3) of Table 1, where we employed the distance-from-the-equator instrument of Hall and
Jones (1999). The estimates from column (2) of Table B-4 are obtained with the instrument
from Auer (2013), which expands on the original `settler mortality' instrument of Acemoglu
et al. (2001) to construct a hypothetical mortality rate for a larger sample of countries.
Our results conrm the positive and signicant relationship between institutional quality
and international trade. The estimate on IQ_BRDR with the Auer instrument is a bit
larger (2.412 std.err. 0.443) as compared to our main estimate of 1.906 (std.err. 0.291) from
column (2) of Table 1. However, the two estimates are not statistically dierent from each
other. Furthermore, with a highly statistically signicant rst-stage estimate and with an
under-identication χ2 test statistic of 768.37 (Anderson, 1951) and weak-identication Wald

F estimates of 925.507 (Cragg and Donald, 1993) and 16.38 (Stock and Yogo, 2005), the
mortality-rate instrument is performing well.
The estimates in column (3) of Table B-4 are obtained with the use of both the Hall and
Jones and the Auer instruments. Once again, these estimates conrm the robustness of our
main results. Estimates from the rst-stage IV regression reveal that the instruments are
strong predictors of the endogenous variable (Anderson χ2 = 1560.52, Cragg-Donald Wald F
= 1260.47, and Stock-Yogo Wald F = 19.93). In addition, in combination, the instruments
54

jointly pass the overidentication test with a Sargan statistic χ2 = 1.053 (p-val. 0.3048).
Columns (4)-(6) of Table B-4 report IV-PPML estimates, which, in terms of the denitions
of the instrumental variables, correspond to the IV-OLS estimates from Columns (1)-(3) of
Table B-4. The IV-PPML estimates are also in support of our main ndings.
Finally, the specications in Table B-5 use the PPML estimator to reproduce our main
OLS results. Each PPML specication in Table B-5 corresponds directly to an OLS specica-
tion from the main text. Specically, column (1) of Table B-5 reproduces the OLS estimates
with gravity variables from column (5) of Table 1. Similar to our main OLS results, the
PPML estimates from column (1) deliver a large, positive, and statistically signicant esti-
mate of the impact of institutional quality on international trade. We do note, however, that
the PPML estimate is signicantly smaller in magnitude as compared to the corresponding
OLS estimate. Possible explanations for the dierence between the PPML and the OLS esti-
mates in gravity regressions include heteroskedasticity and the inclusion of zero trade ows,
see Santos Silva and Tenreyro (2006), as well as the dierent estimation weights attached to
nations with dierent size, see Larch et al. (2017).
Column (2) of Table B-5 reproduces the estimates of the impact of institutional quality
on poor countries' trade with rich countries from column (2) of Table 2. Similar to our OLS
ndings, we obtain positive estimates of the eects of institutional quality on the exports
and on the imports of poor countries with rich countries. However, only the estimate on the
poor countries' imports is statistically signicant. This conrms the signicantly asymmetric
impact of institutional quality on the exports vs. imports of the poor to and from the rich
countries, respectively.
The estimates in columns (3) and (4) of Table B-5 reproduce the corresponding OLS
estimates with country-pair xed eects from column (6) of Table 1 and from column (4)
of Table 2, respectively. The main dierence between the PPML results and our OLS
ndings is that the average impact of institutions in columns (3) and (4) of Table B-5 is not
signicant. A possible explanation for this result is that, on average, the time variation in the
55

institutional quality changes over the period of investigation is not sucient for identication
purposes. This is conrmed in column (4), where we focus on the trade of poor countries,
which experience signicant variation in institutional quality. We obtain a large, positive
and statistically signicant estimate of the impact of institutions on the imports of poor
countries from rich countries. This result is consistent with our OLS ndings and reinforces
the focus of our general equilibrium counterfactual analysis.
Based on the results from this Appendix, we view our main ndings as robust and represen-
tative.
56

Table B-1: Institutions and trade: collinearity
(1) (2) (3)
IQ_EXP IQ_IMP IQ_INTRA
LN_DIST -1.223 -1.223 -1.223
∗∗ ∗∗ ∗∗
(0.057) (0.057) (0.057)
CNTG 0.195 0.195 0.195
(0.226) (0.226) (0.226)
LANG 0.731 0.731 0.731
∗∗ ∗∗ ∗∗
(0.109) (0.109) (0.109)
CLNY 0.469 0.469 0.469
∗∗ ∗∗ ∗∗
(0.157) (0.157) (0.157)
RTA 0.185 0.185 0.185
∗ ∗ ∗
(0.073) (0.073) (0.073)
BRDR -4.256 -4.256 -4.256
∗∗ ∗∗ ∗∗
(0.345) (0.345) (0.345)
WGI_EXPORTER 1.906
∗∗
(0.291)
WGI_IMPORTER 1.906
∗∗
(0.291)
WGI_INTRANATL -1.906
∗∗
(0.291)
N 3880 3880 3880
R2 0.868 0.868 0.868
Notes : This table reports estimates from three alternative specications depending on
the denition of the interaction between the institution quality index and the border
dummy variable. Column (1) reproduces the main results with an institution variable
on the exporter side. The estimates in column (2) are obtained with an institution
variable on the importer side. The estimates in column (3) are obtained with an insti-
tutional variable that interacts the country-specic institution index with an indicator
variable for internal trade. All estimates are obtained with the OLS estimator and
+ ∗
with exporter and importer xed eects, which are omitted for brevity. p < 0.10,
∗∗
p < .05, p < .01. See text for further details.
57

Table B-2: Institutions and trade: alternative institutional quality indexes
(1) (2) (3) (4) (5) (6) (7) (8) (9)
WGI WGI_VA WGI_PV WGI_GE WGI_RQ WGI_RL WGI_CC POLITY FRDMHS
LN_DIST -1.223 -1.217 -1.227 -1.226 -1.225 -1.224 -1.224 -1.221 -1.219
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.057) (0.063) (0.058) (0.056) (0.057) (0.057) (0.057) (0.062) (0.063)
CNTG 0.195 0.201 0.189 0.193 0.196 0.196 0.193 0.198 0.198
(0.226) (0.226) (0.224) (0.226) (0.226) (0.226) (0.225) (0.224) (0.225)
LANG 0.731 0.728 0.724 0.732 0.731 0.732 0.730 0.722 0.726
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.109) (0.110) (0.109) (0.109) (0.109) (0.109) (0.109) (0.110) (0.110)
CLNY 0.469 0.479 0.491 0.467 0.462 0.466 0.472 0.494 0.484
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.157) (0.157) (0.158) (0.157) (0.157) (0.157) (0.157) (0.158) (0.158)
RTA 0.185 0.196 0.191 0.177 0.179 0.184 0.184 0.198 0.198
∗ ∗∗ ∗∗ ∗ ∗ ∗ ∗ ∗∗ ∗∗
(0.073) (0.075) (0.073) (0.072) (0.072) (0.073) (0.072) (0.075) (0.076)
BRDR -4.256 -4.045 -3.520 -4.540 -4.525 -4.257 -4.176 -4.054 -2.036
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.345) (0.405) (0.325) (0.327) (0.355) (0.348) (0.354) (0.614) (0.521)
WGI_BRDR 1.906
∗∗
(0.291)
WGI_VA_BRDR 1.356
∗∗
(0.359)
WGI_PV_BRDR 1.449
∗∗
(0.305)
WGI_GE_BRDR 1.931
∗∗
58
(0.253)
WGI_RQ_BRDR 2.003
∗∗
(0.277)
WGI_RL_BRDR 1.715
∗∗
(0.262)
WGI_CC_BRDR 1.489
∗∗
(0.256)
POLITY_BRDR 0.442
(0.333)
FRDMHS_BRDR 1.210
∗∗
(0.459)
N 3880 3880 3880 3880 3880 3880 3880 3880 3880
R2 0.868 0.866 0.867 0.868 0.868 0.868 0.867 0.865 0.866
: This table reports estimation results with alternative measures of institutional quality. Column (1) replicates the main

Notes
estimates from column (2) of Table 1. Columns (2)-(7), use the individual institutional quality indexes from WGI, including:
Voice and Accountability (column 2), Political Stability and Absence of Violence (column 3), Government Eectiveness (column 4),
Regulatory Quality (column 5), Rule of Law (column 6), and Control of Corruption (column 7). The estimates from columns (8)
and (9) are obtained with the Combined Polity Score index from the Polity IV project, and with an average institution index that is
constructed from the Political Rights and Civil Liberties indicators of the Freedom House initiative. All estimates are obtained with
+ ∗ ∗∗
the OLS estimator and with exporter and importer xed eects, which are omitted for brevity. p < 0.10, p < .05, p < .01. See
text for further details.
Table B-3: Institutions and trade: long panel data
Gravity Variables Country-Pair Fixed Eects
WGI POLITY FRDMHS WGI POLITY FRDMHS
(1) (2) (3) (4) (5) (6)
LN_DIST -1.241 -1.204 -1.206
(0.049)∗∗ (0.054)∗∗ (0.055)∗∗
CNTG 0.169 0.198 0.196
(0.217) (0.206) (0.207)
LANG 0.720 0.705 0.705
(0.094)∗∗ (0.091)∗∗ (0.091)∗∗
CLNY 0.478 0.601 0.598
(0.145)∗∗ (0.146)∗∗ (0.146)∗∗
RTA 0.079 0.036 0.032 0.039 0.222 0.222
(0.057) (0.057) (0.057) (0.050) (0.044)∗∗ (0.044)∗∗
WGI_BRDR 2.011 0.616
(0.277)∗∗ (0.301)∗
POLITY_BRDR 0.830 0.220
(0.234)∗∗ (0.076)∗∗
FRDMHS_BRDR 1.408 0.374
(0.310)∗∗ (0.136)∗∗
N 30753 76914 76914 30753 76914 76914
R2 0.869 0.854 0.855 0.932 0.916 0.916
Notes : This table reports panel estimates of the impact of institutions on international trade. Column
(1) replicates the panel results from column (5) of Table 1 with the main WGI institution variable.
Columns (2) and (3) obtain estimates with the institution indexes from the Polity IV project and from
the Freedom House database, respectively. Columns (4)-(6) replicate the estimations of columns (1)-(3)
with the addition of country-pair xed eects. All estimates are obtained with the OLS estimator and
+ ∗
with exporter-time and importer-time xed eects, which are omitted for brevity. p < 0.10, p < .05,
∗∗
p < .01. See text for further details.
59

Table B-4: Institutions and trade: alternative instruments
IV-OLS IV-PPML
MAIN AUER HJ_AUER MAIN AUER HJ_AUER
(1) (2) (3) (4) (5) (6)
IQ_BRDR 1.764 2.412 2.088 1.015 0.902 0.945
∗∗ ∗∗ ∗∗ ∗ ∗ ∗∗
(0.443) (0.443) (0.310) (0.485) (0.434) (0.278)
LN_DIST -1.223 -1.222 -1.223 -1.224 -1.224 -1.224
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.043) (0.043) (0.043) (0.042) (0.042) (0.042)
CNTG 0.195 0.194 0.195 0.474 0.470 0.472
∗∗ ∗∗ ∗∗
(0.154) (0.154) (0.154) (0.157) (0.156) (0.156)
LANG 0.730 0.735 0.733 0.614 0.613 0.613
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.080) (0.080) (0.080) (0.084) (0.084) (0.084)
CLNY 0.472 0.461 0.466 0.624 0.627 0.626
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.154) (0.154) (0.154) (0.131) (0.131) (0.131)
RTA 0.186 0.183 0.184 0.245 0.246 0.245
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.062) (0.062) (0.062) (0.063) (0.063) (0.063)
BRDR -4.197 -4.466 -4.332 -4.391 -4.395 -4.394
∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗
(0.295) (0.295) (0.264) (0.275) (0.290) (0.283)
N 3880 3880 3880 3969 3969 3969
R2 0.868 0.867 0.868
Notes : This table demonstrates the robustness of our main results to the use of alternative
instrumental variables. Columns (1)-(3) of Table B-4 report IV-OLS estimates. Column (1)
replicates the IV-OLS results from column (3) of Table 1 with the distance-from-the-equator
instrument of Hall and Jones (1999). Column (2) uses the mortality-rate instrument from Auer
(2013). Column (3) uses a combination of both the Hall and Jones and the Auer instruments.
Columns (4)-(6) report IV-PPML estimates, which, in terms of the denitions of the instrumental
+ ∗ ∗∗
variables, correspond to the IV-OLS estimates from columns (1)-(3). p < 0.10, p < .05,
p < .01. See text for further details.
60

Table B-5: Institutions and trade: PPML estimates
(1) (2) (3) (4)
PNL_GRV POOR PNL_FES POOR
LN_DIST -0.723 -0.714
∗∗ ∗∗
(0.082) (0.079)
CNTG 0.512 0.524
∗∗ ∗∗
(0.157) (0.149)
LANG 0.197 0.185
(0.134) (0.130)
CLNY -0.054 -0.038
(0.123) (0.122)
RTA 0.085 0.110 0.131 0.101
+ ∗
(0.117) (0.115) (0.067) (0.049)
IQ_BRDR 0.650 0.694 0.129 0.128
∗∗ ∗∗
(0.102) (0.116) (0.100) (0.092)
IQP _BRDRP R 0.369 0.084
(0.258) (0.146)
IQP _BRDRRP 0.892 0.393
∗∗ ∗∗
(0.244) (0.113)
N 31752 31752 31752 31752
Notes : This table reports estimation results from econometric models that employ
the PPML estimator. All estimates are obtained in panel settings with exporter-time
and importer-time xed eects. Estimates of the xed eects, including the constant
term, as well as estimates of all time-varying border variables (whenever applicable),
are omitted for brevity. Each PPML specication in this table corresponds directly
to an OLS specication from the main text. Specically, column (1) reproduces the
OLS estimates with gravity variables from column (5) of Table 1, while column (2)
reproduces the estimates for poor countries' trade from column (2) of Table 2. Similarly,
column (3) reproduces the OLS estimates with xed eects from column (6) of Table
1, while column (4) reproduces the estimates for poor countries' trade from column (4)
+ ∗ ∗∗
of Table 2. p < 0.10, p < .05, p < .01. See text for further details.
61

SSRN 3167749

Uploaded by

Copyright:

Available Formats

SSRN 3167749

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN 3167749

Uploaded by

Copyright:

Available Formats

6920

Institutions, Trade and

Electronic copy available at: https://ssrn.com/abstract=3167749

CESifo Working Papers

An electronic version of the paper may be downloaded

Electronic copy available at: https://ssrn.com/abstract=3167749

Institutions, Trade and Development:

Cosimo Beverelli Alexander Keck

Maro Larch Yoto V. Yotov

Electronic copy available at: https://ssrn.com/abstract=3167749

ment dierences across countries

Acemoglu and Robinson, 2010

Domestic institutions can have profound eects on international trade

Nunn and Treer, 2014

questions in international economics. In this paper, we seek to contribute to the debate by

dierences through international trade.

Electronic copy available at: https://ssrn.com/abstract=3167749

opment through international trade, we employ a structural gravity methodology. This

methodology is based on identifying empirically the impact of domestic institutions on inter-

of country-specic national institutions on international trade within a structural gravity es-

Electronic copy available at: https://ssrn.com/abstract=3167749

with the xed eects.

of the impact of (bilateral) institutions on trade.

Electronic copy available at: https://ssrn.com/abstract=3167749

approaches by delivering estimates of the impact of country-specic institutional measures

country-specic variable of interest) separately on exports versus imports. Instead, we show

any other country-specic variable) on international trade relative to internal trade.

Electronic copy available at: https://ssrn.com/abstract=3167749

Our approach to identify the eects of national institutional quality on international

importantly, unobservable country-specic links between trade and national institutions.

methods to obtain estimates of the impact of national institutions on international trade.

structed international and intra-national manufacturing trade ows. We obtain positive,

Electronic copy available at: https://ssrn.com/abstract=3167749

impact of country-specic institutions, while  as discussed above  the existing literature

multilateral resistance terms.

Electronic copy available at: https://ssrn.com/abstract=3167749

our sample was the change in consumer prices.

Stimulated by the large and heterogeneous estimates of the GE impact of institutional

stitutional quality improvements on development through trade. To achieve this goal, we

undertake a second counterfactual analysis where we simulate a uniform increase in the

factors in determining the eectiveness of institutional quality improvements. We view this

there might be signicant benets from implementing a combination of trade liberalization

and institutional improvements.

estimates of the impact of institutions on international trade as well as dierential estimates

Electronic copy available at: https://ssrn.com/abstract=3167749

2 Theoretical foundations and identication

to identify the impact of country-specic national institutions on international trade.

2.1 Review of the structural gravity model

Electronic copy available at: https://ssrn.com/abstract=3167749

Electronic copy available at: https://ssrn.com/abstract=3167749

stronger institutions in the poor countries in our sample.

2.2 National institutions and trade: an identication strategy

structural gravity estimations allows identication of the impact of country-specic national

sectoral level and to analysis with panel data.

Start with the following traditional gravity model:

ln Xij = GRAVij β + β1 IQi + β2 IQj + ηi + µj + εij , ∀ i 6= j. (3)

resistance in the importer country. εij is a remainder error term.

Electronic copy available at: https://ssrn.com/abstract=3167749

is not a trivial extension due to a series of collinearity issues discussed therein.

ment dierences across countries

Domestic institutions can have profound eects on international trade

Nunn and Treer, 2014

dierences through international trade.

of country-specic national institutions on international trade within a structural gravity es-

with the xed eects.

approaches by delivering estimates of the impact of country-specic institutional measures

country-specic variable of interest) separately on exports versus imports. Instead, we show

any other country-specic variable) on international trade relative to internal trade.

Our approach to identify the eects of national institutional quality on international

importantly, unobservable country-specic links between trade and national institutions.

structed international and intra-national manufacturing trade ows. We obtain positive,

impact of country-specic institutions, while as discussed above the existing literature

factors in determining the eectiveness of institutional quality improvements. We view this

there might be signicant benets from implementing a combination of trade liberalization

estimates of the impact of institutions on international trade as well as dierential estimates

2 Theoretical foundations and identication

to identify the impact of country-specic national institutions on international trade.

2.2 National institutions and trade: an identication strategy

structural gravity estimations allows identication of the impact of country-specic national

exporter-time and country-pair xed eects.

To identify the eect of national institutions on international trade in a gravity framework

trade ow is international, as opposed to intra-national). In panel estimations, we will add

the eects of country-specic institutions on international trade. Second, we quantify the

we study the dierential partial-equilibrium impact of institutions on trade of poor with

Political Stability and Absence of Violence, Government Eectiveness, Regulatory Quality,

gravity literature. Specically, we employ data on bilateral distance, colonial relationships,

the measures of intra-national and international distance. Specically, consistency is ensured

identify the impact of country-specic institutions within a structural gravity estimation

and signicant estimates on LN _DIST , BRDR, LAN G, CLN Y , and RT A, respectively.

specication as in column (1). However, in addition to the standard gravity covariates we

dierential impact of national institutions on international relative to internal trade. Im-

verse of opportunities to study the dierential impact of any country-specic variable on

literature to also implement an instrumental variable strategy in our setting. Specically,

we obtain large under-identication canonical correlations χ2 test statistic of 766.58, (see

original settler mortality instrument of Acemoglu et al. (2001) to construct a hypothetical

the robustness of our ndings.

country-pair xed eects.

and statistically signicant. Second, we document signicant asymmetries in the impact

rich nations is no longer statistically signicant.