Unit 4 Article Review: International Economics

O R I G I N A L P A P E R

Do regional trade agreements really boost trade? Evidence from agricultural products

Sébastien Jean1,2 • Jean-Christophe Bureau1,2

Published online: 20 April 2016

� Kiel Institute 2016

Abstract We evaluate the impact on trade of regional trade agreements (RTAs) using a panel data approach at the detailed product level which exploits exports to

third destinations and imports from third origins as benchmarks. This method is

robust to both endogeneity and heterogeneity across agreements and across prod-

ucts, and allows differentiation between the impacts of tariff provisions and non-

tariff provisions. The analysis covers agricultural and food products for 74 country

pairs linked by an agreement entered into force during the period 1998–2009. Our

estimate of the mean elasticity of substitution across imports at product level is

slightly below four. Counterfactual simulations suggest that RTAs have increased

partners’ bilateral agricultural and food exports by 30–40 % on average, with

marked heterogeneity across agreements. Also, RTAs are found to increase the

probability of exporting a given product to a partner country although this impact is

small. Finally, we found non-tariff provisions have no measurable trade impact.

Keywords Regional trade agreement � International trade � Agricultural products � Tariff protection

JEL Classfication F13 � Q17

Electronic supplementary material The online version of this article (doi:10.1007/s10290-016-0253-1) contains supplementary material, which is available to authorized users.

& Sébastien Jean [email protected]

1 CEPII, 113 rue de Grenelle, 75007 Paris, France

2 Economie publique, AgroParisTech, INRA, Université Paris-Saclay, 78850 Thiverval-Grignon,

France

123

Rev World Econ (2016) 152:477–499

DOI 10.1007/s10290-016-0253-1

1 Introduction

Although multilateralism has stalled since implementation of the Uruguay round

was completed in 2004, regional trade agreements (RTAs) have proliferated: at

April 7, 2015, the World Trade Organization (WTO) had received 612 notifications

of RTAs and estimates that 406 were in force, up from 180 in 2005. 1

In the

meantime, the share of world trade based on RTAs has grown steadily, increasing

by 30 % even when intra-EU trade is excluded (Bureau et al. 2016).

While RTAs are likely to produce a profound transformation in the established

order of international trade, their real impact on trade flows remains controversial.

Literature surveys show that the estimates vary widely—evidence of a lack of

robustness (Head and Mayer 2014; Cipollina and Salvatici 2010; Ghosh and

Yamarik 2004). Among the various problems, two seem especially important, those

of endogeneity and heterogeneity. Endogeneity stems from the fact that RTAs are

often motivated by missing variables which also affect the intensity of trade. Baier

and Bergstrand (2007) argue that the ensuing estimation bias is potentially large,

and difficult to adjust for but that a panel approach is a relevant solution because it

absorbs any pair-specific trade determinants when it becomes time invariant.

Heterogeneity refers to differences across both agreements and products. Most

studies involving large numbers of RTAs estimate their trade impact using a dummy

variable for the existence of an in force RTA. In most cases, qualitatively different

agreements such as those involving a customs union or a common market are treated

separately but beyond this limited categorical distinction, the methodology assumes

that all RTAs have the same impact on trade. Kohl (2014) refers to this as the

‘‘generalist’’ approach, which does not acknowledge the fact that agreements differ

widely in their provisions (a good overview is WTO 2011). For instance, according

to the tariff provisions under the Brazil–Mexico free trade agreement (FTA), 83 %

of bilateral trade benefits from a strictly positive preferential margin, and this

margin exceeds 20 % for 24 % of bilateral trade. In contrast, the Japan–Singapore

FTA gives rise to a positive preferential margin for only 3.1 % of bilateral trade

(WTO 2011, p. 77). These figures also provide evidence of the heterogeneity across

products: not all products benefit from a positive preferential margin, and the

magnitude of this margin can differ significantly, so that the ensuing trade impact is

bound to be uneven across products. Kohl (2014) emphasizes the need for

‘‘specialist’’ approaches that focus on individual agreements, and he combines

methodological refinements and wide range of generalist approaches (including

treatment for endogeneity), with an acknowledgement of each agreement’s

specificity. This aggregate approach confirms the heterogeneity of the impacts

across agreements. Orefice and Rocha (2014, focusing on network trade) and Kohl

et al. (2016) concur with this conclusion of heterogeneous impacts, and relate it to

differences in non-tariff provisions. Baier et al. (2015) also emphasize impact

heterogeneity, which according to them exceeds what can be explained by the

1 In this article, RTAs should be understood as including preferential agreements between countries in

different regions. Note that the WTO figures tend to overstate the number of active RTAs since those

covering liberalization of both goods and services are counted twice by the WTO. Taking account of

double-counting, the WTO registered 262 ‘‘physical’’ agreements in force in April 2015.

478 S. Jean, J.-C. Bureau

123

degree of liberalization. They relate it to differences in trade-cost elasticities, owing

to differences across country-pairs in fixed and variable export costs.

However, none of these analyses allow consideration of cross-product hetero-

geneity, which is large; we show below that among the RTAs studied in this paper,

the average preferential margin amounts to 16.8 percentage points for tobacco

products compared to only 5.4 % for non-food agricultural products.

The present paper proposes a joint study of those various issues using a product-

level panel data approach at to the trade impact of RTAs. Our reliance on yearly

product-level information on most favored nation (MFN) and preferential tariffs

requires no assumptions about homogeneity across products or across agreements,

or even over time. The panel approach adjusts for RTA endogeneity, and also

controls for any time-invariant determinants including those specific to exchanges

of a given product between a given pair of countries.

While our approach minimizes the scope for potentially uncontrolled trade

determinants, we cannot exclude the possibility of the exporter’s competiveness

varying over time for a given product, or for changes in the importer’s demand. To

control for such time-varying determinants, we transform the dependent variable, as

suggested by Romalis (2007) in the case of NAFTA and CUSFTA. This

transformation to the multiplicative structure of the model corresponds to a

difference-in-differences (DiD) specification of its logarithmic form, similar to the

methods in Hallak (2006), and in Head et al. (2010) which describes it as the

Tetrads method. Application of this method requires both exporter and importer

control groups. By choosing as control groups countries whose trade policy toward

signatories does not change during the study period, we can assume that bilateral

trade determinants with control groups remain unchanged, so that movements in the

DiD in trade flows can be explained by trade liberalization between the signatories.

This controls for time-varying, exporter-product and importer-product specific

determinants.

Our analysis covers 74 country pairs linked by a ‘‘simple’’ RTA (not including

any customs union or common market, see list in Online Appendix 1) which became

effective between 1998 and 2009. Since this approach is demanding in terms of

data, and relies primarily on tariffs, we focus on agricultural and food products

where tariffs are significantly higher than for manufacturing products (e.g.,

Guimbard et al. 2012). This is advantageous: based on previously funded projects

devoted to agricultural trade, the data on this sector are more detailed than for other

sectors. We built a dataset that measures tariffs and trade in a consistent way for

each product (at the 6-digit level of the harmonized system) and each year over the

1998–2009 period, for the 35 countries, including those involved in at least one of

these agreements and their main trading partners. In addition to MFN tariffs, the

detailed commitment schedules of each of these 74 RTAs are factored in, to

compute the corresponding preferential tariffs on a bilateral basis, year by year. To

our knowledge, this dataset is unique. It allows a depiction of the progressive

implementation of tariff cuts, product by product and year by year.

Consistent estimates based on the standard specification would require consid-

eration of a number of fixed effects (exporter–importer-product, export-product-

year, importer-product-year) which is intractable for such a detailed analysis

Do regional trade agreements really boost trade? Evidence… 479

123

covering a large number of agreement. Based on 35 countries, 700 products, and

12 years, the corresponding number of fixed effects would be 35 9 35 9 700 =

857,700 for exporter–importer-product, and 35 9 700 9 12 = 294,000 for expor-

ter-product-year and for importer-product-year. 2

The transformation method which

we describe as a ratio-of-ratios approach at product level, avoids this difficulty. We

also extend the method developed by Romalis (2007) in two ways. First, we apply it

to a large number of agreements rather than a bilateral comparison; second, our

estimation is based on a count model (Poisson) which avoids the bias inherent in

log-transformations of a heavily heteroskedastic model. In addition, we distinguish

between the impact of these agreements on pre-existing trade flows (the intensive

margin), and on the creation of new flows (the extensive margin).

Because RTAs include non-tariff provisions, their trade impact may go beyond

tariffs. Thus, we complement the estimation model by allowing for the non-tariff

impact of RTAs. The nature and depth of these non-tariff provisions vary across

agreements, with influence the ensuing trade consequences, as emphasized for

instance by Dür et al. (2014), Orefice and Rocha (2014) and Kohl et al. (2016). For

our product-level analysis, however, making sense of this heterogeneity would

require more information than we can gather. We thus assume the trade impact of

these non-tariff provisions to be homogenous. This contributes to the literature by

allowing the respective roles of tariff and non-tariff provisions to be disentangled,

but focusing on the average impact of non-tariff provisions is a simplifying

assumption that will require investigation in further research.

We find that trade agreements affect pre-existing bilateral trade flows with a

mean elasticity of substitution at product level of 4, so that a 1 % preferential

margin with respect to the MFN duty rate increases bilateral trade by slightly more

than 4 % on average. According to our counterfactual simulations, once fully

implemented the RTAs studied have increased bilateral agricultural and food

exports by 30–45 % on average. This effect is phased in gradually, with

approximately two-thirds of the final effect felt after 5 years. The differences

across agreements are large, reflecting uneven commitments; in some cases there is

almost zero impact, and in others initial bilateral trade flows double. Although

statistically significant, the impact on the extensive margin, that is new trade flows,

is very weak: a 1 % preferential margin increases the probability of exporting by

0.05 percentage points. Finally, for both the intensive and the extensive margins, we

do not find the RTA non-tariff provisions have any discernible impact upon bilateral

trade, over and above what tariff cuts allow.

The paper is organized as follows. Following a description of the methodological

set-up (Sect. 2) and the data (Sect. 3), Sect. 4 presents the descriptive statistics for

the importance of RTAs for the agricultural and food products sector. Section 5

presents the estimation results which are used for the counterfactual simulations in

Sect. 6.

2 Iterative methods have been developed to carry out multi-way fixed effect estimations with unbalanced

data and a large numbers of effects. However, we have in the present case both a very large number of

observations, due to our product-level approach, as well as three-way fixed effects. To the best of our

knowledge, this remains intractable using a Poisson regression model.

480 S. Jean, J.-C. Bureau

123

2 Methodological set-up

Gravity models are the most common approach used to measure the impacts of a

trade agreement. The basic form of the gravity model is Xijt ¼ GtSitMjt/ijt, where the indices i, j and t denote respectively the exporting country, the importing

country, and the year, X denotes the value of the trade flow, G is a time-specific

constant, S is a variable linked to the exporter’s attributes, and M is a variable linked

to the importer’s attributes, and the influence of the determinants of trade specific to

each country pair is represented by /. Unit-by-unit division of the basic models for the two exporting countries i and i’

and for a given product k, yields: Rii0jkt ¼ Xijkt=Xi0jkt ¼ Sikt/ijkt � �

= Si0kt/i0jkt � �

, where

R is the ratio of country j’s imports of k from suppliers i and i’. This formulation

allows the general term and the importer-specific term to be eliminated. If we let

exporter-specific attributes be invariant (or vary at the same relative rate regardless

of the sector), this equation allows identification of the evolution over time of the

determinants of the bilateral intensity of trade /, provided the bilateral determinants of trade are constant in the case of partner i’. Exporters’ attributes which change

over time, can be controlled for by examining the relative volume of imports from

suppliers i and i’ on markets j and j’:

RORii0jj0kt ¼ Rii0 jkt

Rii0 j0kt ¼

Xijkt

Xi0jkt

� � =

Xij0kt

Xi0j0kt

� � ¼

/ijkt /ij0kt

!

= /i0jkt /i0j0kt

!

: ð1Þ

Since this measure is obtained by dividing unit-by-unit the above-mentioned

ratios for market j and j’, we call it a ratio of ratios, or ROR. The ROR depend purely

on bilateral trade determinants. Clarifying how these determinants relate to tariffs

requires more specific information on demand. A standard assumption consistent

with the general equations presented above, is that the elasticity of substitution

between imports from different origins is constant. In other words, the bundle of

imports for a given product can be represented within consumer preferences through

a constant elasticity of substitution (CES) function over imports with different

origins. In practice, country j’s consumption sub-index specific to good k imports at

period t can be written as:

Cjkt ¼ Cjkt0 X

i 6¼j hijk

qijkt

qijkt0

� �r�1 r

" # r r�1

; ð2Þ

where qijkt refers to the quantity of good k originating in country i consumed in

country j at year t, t0 is the reference year, and r [ 0 is the elasticity of substitution between varieties of good k. Given the calibrated share form used,

hijk ¼ pijkt0 qijkt0P l 6¼j

pljkt0 qljkt0 , where pijkt refers to the tax-inclusive price of country j imports of

products k from country i at year t. Thus, hi is the value share of supplier i in country j imports of good k at year t0. The corresponding dual price index is:

Do regional trade agreements really boost trade? Evidence… 481

123

Pjkt ¼ Pjkt0 X

i 6¼j hijk

pijkt

pijkt0

� �1�r " # 1

1�r

: ð3Þ

Import demand in (tax-inclusive) value can be expressed as (4). Assuming mill

pricing, and assuming that bilateral transport costs (as well as regulations that might

impact on prices) do not change over time, tariffs are the only source of change in

prices, hence (5) holds:

pijktqijkt ¼ pijkt0 qijkt0 pijkt

pijkt0

� �1�r Pjkt0 Pjkt

� ��r Cjkt

Cjkt0

� � ; ð4Þ

Xijkt ¼ Xijkt0 sijkt sijkt0

� ��r Pjkt0 Pjkt

� ��r Cjkt

Cjkt0

� � ; ð5Þ

where, as before, X refers to tax-exclusive import values. sijkt - 1 is the ad valorem customs duty applied by country j to imports of good k from supplier i at time t. It

should be noted that the same equation would have been obtained, had CES

(constant elasticity of substitution) preferences been defined for a basket that

included domestic consumption of good k, in addition to imports. This equation

clarifies the link between applied tariffs and bilateral determinants of trade flows.

The ROR defined in (1) becomes:

RORii0 jj0 kt ¼ RORii0 jj0 kt0 sijkt0 si0 jkt0

! . sij0 kt0

si0 j0 kt0

!" #r sijkt si0 jkt

! . sij0 kt

si0 j0 kt

!" #�r ; ð6Þ

Now, let the indices i and j represent two signatory partners to a bilateral trade

agreement. Also, let j’ denote a control market, defined for each importer [for the

sake of brevity j0 = J(i) is referred to as J in what follows] as a representative set of countries whose trade policy regarding country i has not changed during the period

under study siJkt ¼ siJkt0½ �. 3

Finally, let i’ be the exporter control group [i0 = I(j), henceforth I] consisting of trading partners, such that the trade policy of both

country j and the control market J toward this group I remains unchanged during the

period under examination siJkt ¼ siJkt0 ; sIJkt ¼ sIJkt0½ �. 4

Under these conditions, the ratio of import duties applied by the reference group

of importers J to suppliers i and I does not change over time. Equation (6) can then

be rewritten as RORiIjJkt ¼ kijk sijkt sIjkt

� ��r where kijk ¼ RORii0 jj0 kt0 sijkt0 =si0 jkt0

� �h ir ,

I and J are respectively functions of j and i, so that ROR depends only on i, j, and k.

Traditionally, this type of equation is estimated in log-linear form:

3 In constructing the control groups, we interpret this condition of unchanged trade policy as meaning

that no RTA between the countries was signed or phased in. 4

A sufficient condition is that the ratio siJkt=sIJkt remains constant over time. Here again, we interpret this condition of unchanged trade policy as meaning that no RTA has been signed or phased in between

the countries.

482 S. Jean, J.-C. Bureau

123

ln RORijkt � �

¼ aijk � r ln sijkt sIjkt

� � þ fRTAij þ uijkt; ð7Þ

where u is an error term and aijk = ln(kijk). This specification includes one fixed effect specific to each exporter–importer-good triplet. This means that the elas-

ticity of substitution between imports from different origins r is estimated only on changes over time within each triplet. Fixed effects by exporter, importer, or

good, or any combination of two of these dimensions, are accounted for implic-

itly. 5

Note that Eq. (7) includes an additional variable, RTAij: this dummy variable

denotes the existence of an RTA in force between countries i and j. It is included

in the estimation equation to account for RTA non-tariff provisions which can

reduce the bilateral trade transaction costs. In this specification, the corresponding

impact on trade, measured by parameter f, is assumed to be constant across agreements.

Santos-Silva and Tenreyro (hereafter SS&T) demonstrate the bias inherent in

estimating those models in their logarithmic form (SS&T 2006), and argue that

gravity equations should be estimated in a multiplicative form, for instance

using a Poisson pseudo-maximum likelihood (PPML) estimation technique. The

arguments related to the standard gravity model apply to the present case which

is a DiD of the log-transformed model, or a ratio of ratios of the multiplicative

form. Also, no distributional assumption is needed to ensure the consistency of

the PPML estimator which requires only correct specification of the conditional

mean.

SS&T (2011) show the theoretical consistency and good statistical properties of

this estimator, and that it performs better than the proposed alternatives (see also

Fally 2015; Sun and Reed 2010). They show that Gamma PML is an efficient

estimator in this context. However, as discussed in Head and Mayer (2014), these

results are based on a data generating process which does not reflect modern

theories of international trade. Ultimately, the best estimator depends on the type of

heteroskedasticity since PPML is efficient under the assumption of a constant

variance-to-mean ratio, while Gamma PML is efficient under the assumption of a

constant coefficient of variation. The Manning and Mullahy (2001) test (referred to

as the ‘‘MaMu’’ test in Head and Mayer 2014, and the Park-type test in SS&T 2006)

allows the corresponding diagnostic to be made by studying the relationship

between the empirical counterparts of the error variance and the expected value of

the dependant variable. 6

While this test provides a biased estimate of the

corresponding coefficient, Head and Mayer (2014) show that it differentiates

between alternative settings. In our case, applying this test to the benchmark

estimation presented below delivers an estimated coefficient k̂ ¼ 1:37, with a

5 It is superfluous to incorporate them explicitly since they would be perfectly correlated with the fixed

effects already included. Two-way time-exporter or time-importer fixed effects are not needed either,

since the corresponding shocks should be absorbed by the dependent variable transformation. 6

In practice, the test is based on a regression where the dependent variable is the logarithm of the

squared error term, the latter being computed as the difference between trade in level and the exponential

of its fitted log-value. The fitted log-trade value is used as independent variable.

Do regional trade agreements really boost trade? Evidence… 483

123

standard error of 0.26. Accordingly, in what follows, our preferred estimator is the

PPML. 7

Our estimates are based on fixed-effect PPML estimates of the multiplicative

form of the model (8):

RORijkt ¼ exp aijk � r ln sijkt sIjkt

� � þ fRTAij

� � þ vijkt: ð8Þ

The dependent variable computation involves control group trade flows: for a given

importer, these flows are common to exporters; for a given exporter, they are common

to importers. This feature might produce residual correlation across observations for a

given exporter, or a given importer in a given year. Therefore, standard errors ideally

should be clustered by both importer-year and exporter-year. This is possible for

ordinary least square (OLS) estimates based on Cameron et al. (2011), but not for

PPML estimates. In the case of PPML estimates, we compute robust standard errors

clustered at panel level (i.e., by exporter–importer-product triplet), following

Wooldridge (1999). However, to allow us to compute clustered standard errors, we

also conduct OLS estimates of the log-transformed model based on (7).

As the recent international trade literature demonstrates, it is important to

account for trends in existing trade flows—the so-called intensive margin of trade as

well as developments in the number of goods traded at the extensive margin (e.g.,

Chaney 2008; Helpman et al. 2008). The estimates described so far focus on the

intensive margin (defined at the country-pair product level) since they deal with

changes over time in non-zero trade flows. To complement these, we estimate the

export probability. Econometric modeling of this probability requires accounting for

the determinants specific to each (potential) exporter–importer-good triplet.

Exporter-by-year and importer-by-year fixed effects are also important since the

corresponding idiosyncratic shocks are no longer controlled for by transformation of

the dependent variable. Given the large size of the sample, accounting for fixed

effects by panel unit is possible only by applying a within estimator—explicitly

incorporating all the dummies is numerically intractable.

Following Frazer and Van Biesebroeck (2010) and Head et al. (2010), we use a

linear model to estimate the probability of exporting. The estimation equation is:

P Xijkt [ 0 � �

¼ a 0

ijk þ b 0

it þ c 0

jt � r 0 ln

sijkt sIjkt

� � þ f0RTAij þ wijkt; ð9Þ

where w is an error term. Thus, our benchmark estimates, in addition to the panel-

specific fixed effects aijk 0 (81,570 in the benchmark estimates), include 419 reporter-

year fixed effects (35 countries 9 12 years, minus 1) and the same number of partner-

year fixed effects. Frazer and Van Biesebrock (2010: 132) argue that ‘‘the main dis-

advantage [of the linear probability model]—that predicted values are not restricted to

lie on the (0, 1) interval—is unlikely to be much of an issue as all coefficients are

7 The PPML estimator is efficient for k = 1, while Gamma PML is efficient for k = 2; however, Head

and Mayer (2014) show that OLS estimates of k such as those in this exercise, are significantly upward biased so that ‘‘estimates of k significantly below two were a near perfect predictor of a constant variance to mean ratio’’ in their simulations.

484 S. Jean, J.-C. Bureau

123

identified off the time variation within country-product categories.’’ In addition, Angrist

and Pischke (2009: 107) emphasize that linear estimates, while theoretically less well-

suited, are very close in practice to the marginal effects drawn from non-linear models.

3 Data

The method presented above requires panel data on bilateral trade and preferential

tariffs, at the product level. We rely first on a joint study OECD and Inter American

Development Bank study on the treatment of agriculture in trade agreements in order

to characterize concession schedules including phase-in patterns, product by product

(Fulponi et al. 2011). The data were codified and combined with assessments of ad

valorem equivalents of MFN tariff duties taken from the MacMap-HS6 database for

years 2001, 2004, and 2007 including both ad valorem and non-ad valorem duties (see

Guimbard et al. 2012). The resulting database gives ad valorem equivalents of

preferential duties for 74 RTAs, year by year over the period 1998–2009. Note that all

the agreements covered belong to the ‘‘simple’’ RTAs category, i.e., no customs union

or common market involved. Trade data (cost-insurance-freight, inclusive of annual

imports) are from the BACI database which provides a detailed description of the role

played by RTAs in global agricultural products trade during 1998–2009 (Gaulier and

Zignago 2010). However, the sample of RTAs analyzed in this paper study is

constrained by data availability and cannot be considered representative. For example,

Latin American countries are over-represented while African RTAs are grossly under-

represented. A broader application of the methodology would be desirable. However,

the large size of our sample makes it possible to obtain some useful insights

applicable to RTAs in general.

The methodology applied here requires the creation of control groups specific to

each importer and exporter. For a given country, the corresponding control group

consists of all the countries in our database which had not signed a preferential

agreement with that country by 2009. To improve robustness, this group includes

the entire population of countries which provides meaningful points of comparison.

This group composition is specific to each country, but stable over time. The

estimates are conducted at the 6-digit product code level of the harmonized system.

Working with product-level trade data introduces a potential for lack of

robustness which is a particular problem if the dependent variable is computed as a

ratio as in our case. We handle this by conducting various checks. First, to ensure

that trade flows with control groups are representative, we retain only those

exceeding USD 200,000. 8

Second, we drop observations where either one of the

8 This condition applies only to trade with control groups not trade between partner and reporter. The

threshold was chosen based on analysis of the degree of autocorrelation of product-level trade flows.

Control group trade flows are used as a benchmark to represent partner-specific trade determinants. Since

these determinants are likely to be highly correlated in practice (they are linked to variables such as

preferences and productivity), a low level of autocorrelation points to lack of representativeness which is

likely when flows are small for two reasons: their possible overdependence on incidental factors, and the

different statistical reporting thresholds used by countries. In fact, the autocorrelation of flows is

significantly lower at magnitudes of below USD 200,000. See Online Appendix 2.

Do regional trade agreements really boost trade? Evidence… 485

123

control ratios XiJkt=XIJkt and XIjkt=XIJkt deviates from its median over the period by a multiplicative factor of more than three: when the deviation exceeds this factor, we

consider that the instability of the ratio prevents it from being a reliable control.

Third, we test alternative assumptions in constructing the estimation sample, as

robustness checks. 9

Another concern is the measurement of protection during the first year of

enforcement of an RTA. In our database, protection is measured on January 1 each

year. Therefore, for an RTA that comes into force during a particular year, the tariff

cuts are not taken into account until the following year, resulting potentially in

serious mismeasurement of the tariffs applied in the first year. In addition, the trade

impacts of an RTA might be delayed compared to the date of enforcement, because

the economic agents need time to adapt to a new institutional context. For those

reasons, we consider the impact of RTAs on the first year of entry into force

separately.

4 RTAs and international trade flows

The share of global trade conducted between two parties to a trade agreement rose

from less than 24 % in 1998 to over 36 % in 2009 (the year that our data sample

ends), with no sign of slowing at the end of the period. An analysis by broad sectors

reveals that while the share of global trade between RTA partners was higher for

manufactured products in 1998, it was below the level for agro-food products in

2009. The accelerated growth agricultural production might at least in part, reflect

the greater intensity of agricultural trade between signatories to RTAs that came

into force between 1998 and 2009.

Implementation of RTAs typically extends over some 10 years, and for selected

products the transition period can exceed 15 years. A detailed examination of the

tariff concessions for agricultural and food products in the 74 RTAs allows the

average preference margin to be computed for each agreement, as a function of the

number of years since entry into force, based on the exact phase-in pattern. To

describe these patterns, we compute the ‘‘preferential margin’’, defined as the price

wedge (taxes included) attributable to preferential treatment. Note that this is

different from the reduction in tariff rates shown in Eq. (7). For ad valorem import

duties under the MFN system (tMFN) and the preferential system (tpref), this margin

is defined as m ¼ 1 � 1 þ tpref � �

= 1 þ tMFNð Þ, or in line with our earlier notation, m ¼ 1 � spref =sMFN . Thus, a 1 % preferential margin means that for the same producer price, a product benefiting from the preferential treatment would be sold at

a consumer price 1 % lower than a product sold under the MFN regime.

Over the whole sample, the mean preferential margin nearly doubles within

8 years of entry into force, rising from 4.3 % during the first year to 8.8 % 8 years

later. In contrast, the level of preferential margins hardly changes after 10 years of

entry into force of an agreement.

9 We ran robustness checks to deal with the minimum threshold and multiplicative factor chosen. See

Table A.2 in Online Appendix 3.

486 S. Jean, J.-C. Bureau

123

We can distinguish between two groups of partner countries: high-income OECD

member countries (which we describe as the ‘‘North’’), and others (the ‘‘South’’).

This reveals that preferential margins granted by agreements between South

countries at the end of the phase-in period (10.1 %) are near the mean calculated for

all agreements (9.5 %). Conversely, North–South agreements are asymmetric,

especially after several years: margins granted by South countries are higher

(10.4 % after 8 years, 11.2 % once fully phased in) than those granted by North

countries (6 %, with little variation over time). Since our sample includes only one

North–North agreement, in what follows we do not consider this category

separately.

The differences between the chapters of the Harmonized Commodity Description

and Coding System (HS) are presented in a similar calculation in Table 1. They are

quite significant; the average preferential margin at the end of the phase-in period

ranges from 5.4 % for non-food agricultural products to as high as 14.0 % for dairy

products, eggs and honey (Chapter 2), and 16.8 % for tobacco products (Chapter 24).

The preferential margin in more than half agricultural sectors exceeds ten points.

Cross-sectoral differences in the preferential margin tend to increase over the time

since implementation, confirming the well-known fact that transition periods are

longer for more sensitive products.

Because of marked cross-country differences in import growth trends it is

difficult to show how RTAs might be related to trade flows. Focusing on the share of

each signing country in its RTA partner’s total imports allows us to dispense with

these influences. Figure 1 shows that for both agricultural and food products, this

market share increases by 15–20 % on average over the first 4 years of enforcement

of an agreement, suggesting that RTAs have an impact on bilateral trade flows.

5 Estimation results

To evaluate the impact of RTAs on international trade in agricultural products, the

above described method is applied to the 74 agreements for which we were able to

obtain complete information on both custom duty ad valorem equivalents and

concession schedules for the 1998–2009 period.

5.1 Intensive margin

Our benchmark import substitution elasticity estimate, obtained by applying the

PPML estimator to the whole sample, gives a statistically significant value of 3.99 at

least 1 year after implementation (Table 2, estimate 1 row 5). This means that an

RTA that reduces the relevant tariff by 1 % results in a 4.07 % [exp(0.0399) - 1]

increase in bilateral exports compared to flows between the signatory parties and the

control groups. This increase is based on the ratio of bilateral exports to the controls,

referred to above as bi-ratios which means it could result from an increase in

bilateral exports (trade creation) or a decline in trade with third countries (trade

diversion), or both. Although our methodology does not allow us to disentangle

these effects, the counterfactual simulations below provide some insights.

Do regional trade agreements really boost trade? Evidence… 487

123

The specificity of the first year of RTA implementation being linked to

measurement issues combined with adjustment delays is confirmed because the

corresponding estimated elasticity (2.54) is about half the eventual size achieved,

and is only statistically significant at the 10 % level. 10

This would tend to confirm

that analyzing the year of implementation based on annual data is difficult.

Non-tariff provisions, estimated here through the RTA dummy, do not have a

significant impact on bilateral trade flows: in other words, there is no trade gain

Table 1 Mean base rate and preferential margin by harmonized system chapter and by time elapsed since entry into force of the agreement (percentages). Source: Calculated by the authors from BACI

(CEPII), Comtrade (UN), MAcMap-HS6, and IDB data

Chapter Base Preferential margin

Rate Year 1 Year 5 Full

01—Live animals 9.4 5.3 6.0 6.5

02—Meat and edible meat offal 21.8 5.1 8.4 11.8

04—Dairy produce; eggs; honey 27.9 7.0 10.4 14.0

05—Products of animal origin, NES 5.8 4.1 5.0 5.6

06—Live trees and other plants 10.1 5.2 6.8 7.6

07—Vegetables 13.3 5.9 8.8 10.5

08—Fruits 11.6 6.3 8.8 10.4

09—Coffee, tea, spices 10.0 4.9 7.1 8.4

10—Cereals 17.1 5.6 7.7 10.1

11—Products of the milling industry 17.1 4.4 8.4 11.6

12—Oil seeds and oleaginous fruits 7.0 4.0 5.2 5.9

13—Lac; gums, resins 5.9 3.5 5.0 5.7

14—Vegetable plaiting materials 6.0 3.6 5.1 5.9

15—Animal or vegetable fats and oils 10.8 3.7 6.4 8.9

16—Preparations of meat and fish 20.1 4.9 9.0 12.7

17—Sugars and sugar confectionery 18.3 4.9 7.7 10.9

18—Cocoa and cocoa preparations 11.2 4.9 7.4 9.8

19—Preparation of cereals 13.6 4.8 8.5 11.4

20—Preparations of vegetables and fruits 15.0 6.1 10.0 12.7

21—Miscellaneous edible preparation 13.5 6.0 9.5 12.0

22—Beverages, spirits and vinegar 23.6 6.1 10.5 13.5

23—Food residues and waste 9.3 4.1 6.1 7.6

24—Tobacco 23.4 9.5 13.6 16.8

Non-food agricultural products 5.9 3.2 4.5 5.4

All products 13 5.0 7.5 9.5

‘‘Year 1’’ refers to the year following entry into force of the agreement, ‘‘Year 5’’ refers to the fifth year

after entry into force, ‘‘Full’’ refers to full implementation of RTAs after the phase-in period. ‘‘Base rate’’

refers to the duty rate on which the agreement is based. The mean is computed over all the agreements

covered in the paper (see list in Online Appendix 1)

10 With some exceptions, the estimated elasticities for the first year of implementation (not discussed or

presented in this paper) are small and statistically not significant.

488 S. Jean, J.-C. Bureau

123

apart from that linked to tariff cuts. In fact, this effect is negative but is not

significant, and holds across all the estimates presented below. While a negative

sign might seem counterintuitive, it may simply reflect the fact that an

agreement tends to favor those exporters and those products that benefit from a

positive preferential margin. As soon as supply is constrained or the supply

curve is upward-sloping, sales of products not benefiting from a preferential

margin may be affected negatively by a bilateral agreement. In our case, this

effect is not significant but would make sense if non-tariff provisions do not

provide benefit. Our main conclusion from these estimates is that non-tariff

provisions are not an efficient way to deliver trade gains over and above what

can be achieved by tariff cuts.

The base OLS estimate of the import-substitution elasticity (after the first year of

entry into force of a preferential agreement) is also statistically significant but to a

lower degree than in the PPML estimate (1.24, see Table 2, estimate 2). It has been

acknowledged that OLS estimates are biased in such contexts but the method allows

clustered standard errors to be computed. These are rather small, corresponding to a

significance level comparable to the levels in the PPML estimates. As a robustness

check, we applied the estimation technique to a sample that excluded panels (i.e.,

reporter–partner-product triplets) where the average value of non-zero trade flows

over the 1998–2009 period is below USD 100,000. Indirectly, this controls for

possible lack of reliability of statistics when trade flows are low. In the succeeding

2. 8

3 3.

2 3.

4

-2 -1 0 1 2 3 -2 -1 0 1 2 3

Agriculture Food products S

ha re

in R

T A

p ar

tn er

's im

po rt

s (%

)

Years since RTA

Fig. 1 Average share in RTA partner’s imports, before and after RTA enforcement. Scope: 384 ordered country pairs for which an RTA entered into force between 2000 and 2006. Source: Calculated by the authors from the BACI (CEPII) database, Comtrade (UN), MAcMap-HS6, and IDB data

Do regional trade agreements really boost trade? Evidence… 489

123

tables, these estimates are categorized as ‘‘Excluding low mean panels’’. This

restriction barely changes the PPML estimate (Table 2, 4.19, see estimate 3) but it

increases the OLS estimate by almost half (1.73), suggesting that the latter is more

sensitive to small-flow observations. To further check robustness, we restricted the

sample to the 2001–2009 period on the grounds that information on MFN duties was

first available directly (estimates referred to as ‘‘Excluding low mean panels post

2000’’ in the following tables). This did not substantially change the value of the

PPML estimate (3.60, see Table 2, estimate 5), but the OLS estimate increased to

1.93. Robustness was checked by restricting the control group to countries on the

same continent or within the same income category; alternatively, regressions were

weighted by the log of the value of trade flows (see Table A.3 in Online Appendix 3).

Overall, the PPML but not the OLS estimates appear consistent and robust. It

could be that since the OLS estimates are based on log transformation of the model,

the lack of consistency increases if the analysis is carried out at product level where

Table 2 Estimated trade impacts of RTAs’ tariff and non-tariff provisions. Source: Authors’ estimates based on IDB, MAcMap-HS6 (ITC and CEPII), BACI (CEPII) and Comtrade (UN Statistics Division)

data

Dependent variable: exports, ratio of ratios

Benchmark

estimates

Excluding ‘‘low-mean’’

panels

Excluding ‘‘low-mean’’ panels,

post-2000

(1) (2) (3) (4) (5) (6)

Log price wedge resulting from preferential duties

RTA’s first year 2.54* 0.29 2.68* 0.79* 2.40* 1.08**

(1.39) (0.40) (1.50) (0.46) (1.45) (0.49)

Subsequent years 3.99*** 1.24*** 4.19*** 1.73*** 3.60*** 1.93***

(1.29) (0.30) (1.40) (0.37) (1.27) (0.39)

Non-tariff provisions (RTA dummy)

RTA’s first year -0.07 -0.03 -0.08 -0.01 -0.05 -0.01

(0.10) (0.02) (0.11) (0.03) (0.12) (0.03)

Subsequent years -0.09 -0.05** -0.09 -0.01 -0.02 0.05

(0.09) (0.02) (0.10) (0.03) (0.10) (0.03)

Estimation method PPML OLS PPML OLS PPML OLS

Observations 80,641 63,122 47,127 43,128 36,658 33,876

Panel units 10,703 9462 5888 5678 5624 5433

All estimates include three-way partner–reporter-product fixed effects. Estimates are based on Eq. (8) for

PPML estimates, and on Eq. (7) for OLS estimates. The coefficients reported refer to the estimated value

of r, interacted with a dummy indicating whether the RTA is in its first year of implementation (1st row) or not (2nd row). ‘‘Low-mean panels’’ are those partner–reporter-product triplets for which the mean

value of non-zero trade flows is lower than USD 100,000 (see text). The estimation period is 1998–2009,

except in columns (5) and (6) where it is restricted to 2001–2009. Robust standard errors are reported in

parentheses, clustered at the panel (reporter–partner-product) level for PPML estimates, at both reporter-

year and partner-year level for OLS estimates. Asterisks denote statistical significance levels 10 % (*),

5 % (**) and 1 % (***) respectively

Scope: Agricultural products

490 S. Jean, J.-C. Bureau

123

it is even more pronounced than at cross country level. 11

This seems to matter even

if identification is based on out of time variations within reporter–partner-product

triplets.

Since our methodology is based in part on Romalis (2007), that paper provides a

natural point of comparison for our results. Romalis’s estimates vary between 6.3

and 9.4 for United States imports from Canada, between 9.6 and 10.9 for US

imports from Mexico, between 2.8 and 5.5 for Canadian imports from the United

States, between 6.6 and 8.1 for Canadian imports from Mexico, between 2.0 and 2.5

for Mexican imports from the United States, and between 0.5 and 0.7 (not

significant) for Mexican imports from Canada. Our estimates would seem consistent

with these figures, although are somewhat lower in value. This is borne out by

previous detailed estimates although none of them is specific to agriculture. Using a

very different methodology, Broda and Weinstein (2006) find that the unweighted

mean of the elasticities of substitution estimated for the United States between 1990

and 2001 is approximately 12.6 for goods at the 10-digit product code level of the

HS (compared to only 4.0 for the 3-digit product code level), for a median of 3.1

(2.2 at three digits). Simonovska and Waugh (2014) find a mean of approximately 4

for the same product code level employed in our paper—HS6. Import-demand

elasticities (at HS6 level) estimated by Kee et al. (2008) equal 3.1 on average for all

products. Head and Mayer’s (2014) meta-analysis of estimates based on aggregate

date shows a median of 5 for structural estimates using tariffs or freight rates as the

identifying variable. Overall, these comparisons suggest that our estimates are in

line with the literature, although perhaps on the low side.

To investigate whether these effects depend on the partners’ income levels, we

grouped the agreements according to the partner country categorization of North

and South. In each case, the estimated elasticity is significantly 12

higher for North–

South flows than for South–South and South–North flows (respectively 5.00, 3.91

and 3.42 in the benchmark estimation, see Table 3, estimates 1–3). The better

capacity of Northern exporters to take advantage of the opportunities created by an

agreement might explain this difference. Across all estimates, the trade impact of

non-tariff provision remains statistically insignificant, and this is true also if their

impact is allowed to vary across agreement types.

We also investigate the stability of elasticities over different ranges of

preferential margins (recall that ‘‘preferential margin’’ is used here to describe the

price wedge including taxes, attributable to preferential treatment). Separate

consideration of those cases where the preferential margin is lower than 5 %,

between 5 and 10 %, and higher than 10 %, yields statistically significant 13

differences. Although less precisely estimated, elasticities are found to be higher for

low preferential margins. This might be explained by the supply constraints, or an

11 Zero flows are more widespread although in our context this is not immediately apparent since the

dependent variable also is frequently missing, due to zeros in the control group countries’ trade flows

(or—as mentioned above—to values too low to be considered representative). 12

A likelihood-ratio test rejects the equality of the corresponding coefficients (v2 = 12.4, p value \1 % in the benchmark estimation). 13

A likelihood-ratio test rejects the equality of the corresponding coefficients (v2 = 45.2, p value \1 % in the benchmark estimation).

Do regional trade agreements really boost trade? Evidence… 491

123

upward-slopping supply curve, making it difficult to increase exports proportion-

ately in cases where a large preferential margin is available.

5.2 Extensive margin

To estimate the incidence of preferential tariffs on the extensive margin, we employ

a linear model of the probability of exporting, as described above. However, our

sample is nearly ten times larger because we no longer need to use representative

control-group trade flows. Our estimates confirm the impact of preferential tariffs on

the probability of exporting. After the first year (for which the unreported estimated

Table 3 Estimated trade impacts of RTAs’ tariff and non-tariff provisions, by type of agreement, by level of preference margin and by HS section. Source: Authors’ estimates based on IDB, MAcMap-HS6,

BACI and Comtrade data

Dependent variable: exports, ratio of ratios

Benchmark

estimates

Excluding ‘‘low-

mean’’ panels

Excluding ‘‘low-

mean’’ panels, post-

2000

(1) (2) (3) (4) (5) (6)

Log price wedge resulting from preferential duties

South–South 3.91** 4.11** 3.54*

(1.84) (1.97) (2.00)

North–South 5.00*** 5.57*** 4.13**

(1.45) (1.71) (1.92)

South–North 3.42*** 3.43*** 3.79***

(1.15) (1.17) (1.07)

0 \ preferential margin \5 % 6.49* 7.32* 6.57

(3.64) (3.88) (4.17)

5 % \ preferential margin \10 % 3.86** 4.25** 5.36***

(1.81) (1.95) (2.00)

10 % \ preferential margin 4.16*** 4.40*** 3.47***

(1.29) (1.39) (1.11)

Non-tariff provisions (RTA dummy)

RTA’s first year -0.07 -0.01 -0.07 -0.02 -0.06 0.03

(0.10) (0.12) (0.11) (0.13) (0.11) (0.13)

Subsequent years -0.09 -0.11 -0.09 -0.12 -0.03 -0.05

(0.09) (0.09) (0.10) (0.10) (0.09) (0.10)

Observations 80,641 80,641 47,127 47,127 36,658 36,658

Panel units 10,703 10,703 5888 5888 5624 5624

All results are PPML estimates based on Eq. (8). The coefficients reported refer to the estimated value of

r in cases where the RTA is beyond its first year of implementation. All estimates include three-way partner–reporter-product fixed effects. The estimation period is 1998–2009, except in columns (7), (8)

and (9) where it is restricted to 2001–2009. Asterisks denote statistical significance levels 10 % (*), 5 %

(**) and 1 % (***) respectively

492 S. Jean, J.-C. Bureau

123

effect is insignificant), the estimated effect across all agreements is relatively weak

(0.049), suggesting that a preferential margin that lowers the tariff-inclusive price

by 10 % increases the probability of exporting by 0.5 % (Table 4, estimation 1).

Distinguishing across categories of agreements shows that this effect is only

significant for North–South trade flows, for which the corresponding elasticity

(0.13) is three times bigger than the average (Table 4, estimation 2). Note that this

category shows the lowest estimated intensive-margin elasticity. Distinguishing by

preferential margin level also shows that the elasticity is lower for high preferential

margin levels, suggesting that the potential for trade preferences to create new trade

flows does not increase much beyond a certain level of preferential margin.

5.3 Preferential margin categories

The impact of preferential treatment can also be assessed by categorizing exporter–

importer-good triplets based on whether an agreement is in force, and if so, the

magnitude of the preferential margin. We create five categories: (i) no agreement in

place; (ii) an agreement in force but the preferential margin is zero for this product;

(iii) preferential margin greater than zero but less than 5 %; (iv) preferential margin

between 5 and 10 %; and (v) preferential margin greater than 10 %. In the

econometric specification defined by Eq. (8), the tariff level is replaced by a series

of dummies to indicate belonging to one of these categories, with the first group

serving as the reference. As before, the first year of implementation of each

agreement is handled separately. For clarity, only the coefficients of the effects

during subsequent years are presented.

The estimates indicate that an agreement does not affect trade materially for

products that do not benefit from a preferential margin (Table 5, estimates 1–3, first

line), which is consistent with the previous finding of a insignificant impact for non-

tariff provisions. The trade impact is also not significant when the preferential margin

is lower than 5 %. However, we find that entry into force of a trade agreement, in the

case of a good benefitting from a preferential margin larger than 5 %, translates into a

strong and significant increase in bilateral exports involving third countries.

According to our benchmark estimates, the average increase is approximately

27 % [exp(0.24)-1 = 27 %] if this margin is between 5 and 10 %, and is 60 % if

the preferential margin exceeds 10 %. If we apply the same approach to the export

probability the impact is statistically significant only for products where the margin

is between 5 and 10 %, in which case the probability increases by 0.8 % (Table 5,

estimation 4). This confirms the limited impact of RTAs on the extensive margin in

the case of agricultural and food products.

6 Counterfactual simulations of the trade impacts of RTAs

The above estimates allow the partial equilibrium trade impacts of each RTA to be

simulated very simply, product by product. A partial equilibrium framework seems

appropriate for agricultural and food products. In addition, given the negligible

quantitative importance of the estimated impact of RTAs at the extensive margin,

Do regional trade agreements really boost trade? Evidence… 493

123

the simulations are focused on the intensive margin which keeps the theoretical

framework simple and transparent. This section discusses the corresponding

theoretical underpinnings, and summarizes the results obtained.

6.1 Underpinnings and assumptions

A fully consistent assessment of the impact of an RTA on trade flows would require

a general-equilibrium analysis taking account, inter alia, of supply responses and

changes in import demand. We make two simplifying assumptions. First, supply

response is not considered. In most cases, the RTAs being considered are unlikely

substantially to alter production prices since they involve only a limited share of

each partner country’s exports. Second, we assume total import demand by product

(PjktCjkt) to remain constant in (tax-inclusive) value. Therefore, we ignore the

impact of RTAs on income, and possible substitution between imports and domestic

Table 4 Estimated impact of preferential duties on the export probability. Source: Calculated by the authors from IDB, MAcMap-HS6, BACI and Comtrade data

Dependent variable: export probability

(1) (2) (3)

Log price wedge resulting from preferential duties

All agreements, all products 0.049***

(0.014)

South–South 0.013

(0.017)

North–South 0.133***

(0.033)

South–North 0.044

(0.033)

Elasticity, preferential margin 0–5 % 0.140***

(0.047)

Elasticity, preferential margin 5–10 % 0.134***

(0.025)

Elasticity, preferential margin [10 % 0.047***

(0.014)

Non-tariff provisions (RTA dummy) -0.003 -0.003 -0.005*

(0.003) (0.003) (0.003)

Observations 978,840 978,840 978,840

Panel units 81,570 81,570 81,570

All estimates use a linear probability model, based on Eq. (9). The coefficients reported refer to the

estimated value of r0, in cases where the RTA is past its first year of implementation. For reasons of space, the coefficients for the first year of implementation are not reported; all are statistically

insignificant. All estimates include three-way partner–reporter-product fixed effects, as well as two-way

partner-year and reporter-year fixed effects. The estimation period is 1998–2009. Robust standard errors,

clustered at panel (reporter–partner-product) level, are reported in parentheses. Asterisks denote statistical

significance levels 10 % (*), 5 % (**) and 1 % (***) respectively

494 S. Jean, J.-C. Bureau

123

products (an alternative assumption is considered below). Under these assumptions,

it follows from Eq. (5) that import demand in value can be expressed as:

Xijkt ¼ Xijkt0 sijkt sijkt0

� ��r Pjkt0 Pjkt

� �1�r : ð10Þ

This equation enables counterfactual simulations of the trade impact of RTAs. Let

us consider an RTA implemented in year t0 ? 1. Assuming all other trade deter-

minants remain unchanged, preferential tariff cuts in subsequent years are the only

source of trade changes, hence:

~Xijkt ¼ Xijkt0 sijkt sijkt0

� ��r hijk

sijkt sijkt0

� �1�r þ 1 � hijk � �

" #�1 ; ð11Þ

where, as before, X refers to tax-exclusive import values. Ceteris paribus, RTA

implementation shifts trade from Xijkt0 to ~Xijkt. An alternative assumption would be

Table 5 Estimation of the trade impact of preferential agreements, by level of preferential margin. Source: Authors’ estimates based on IDB, MacMap-HS6, BACI and Comtrade data

Dependant variable: Diff in diff log exports Export

probability

Sample: All Excl. low-mean

panels

Excl. low,

post-2000

All

(1) (2) (3) (4)

Independent variable:

Dummy variable indicating

PTA in force, preferential margin =0

for this product

0.00 0.02 0.08 -0.004

(0.11) (0.13) (0.13) (0.003)

0 \ preferential margin \5 % 0.16 0.19 0.21* 0.003

(0.12) (0.13) (0.12) (0.002)

5 % \ preferential margin \10 % 0.24* 0.27* 0.41** 0.008***

(0.14) (0.15) (0.16) (0.003)

10 % \ preferential margin 0.47*** 0.51*** 0.36** 0.004

(0.17) (0.18) (0.15) (0.003)

Estimation method PPML PPML PPML Linear

probability

model

Observations 80,641 47,127 36,658 978,840

Panel units 10,703 5888 5624 81,570

Estimates (1)–(3) are based on Eq. (8), column (4) is based on Eq. (9). The coefficients reported refer to

the estimated value of r and r’, respectively, in cases where the RTA is beyond its first year of implementation. In each case, the tariff duty variable is replaced by a set of dummy variables, as indicated

in the text. All estimates include three-way partner–reporter-product fixed effects. Estimate (4) also

includes two-way partner-year and reporter-year fixed effects. The estimation period is 1998–2009,

except for (3), where it is 2001–2009. Robust standard errors, clustered at the panel (reporter–partner-

product) level, are reported in parentheses

Do regional trade agreements really boost trade? Evidence… 495

123

the Armington assumption in its initial form, namely the assumption that products

are differentiated by country of origin, with domestic products included in the same

bundle (i.e., the same CES function) as other imports. In this case, our approach

would be similar to the one above but including domestic consumption. Accord-

ingly, the hs should be defined as value shares in total consumption, not just in imports. This information is not available at product level but an upper bound for

the corresponding impact can be computed by assuming that this share of imports

from the partner in total domestic consumption of the corresponding good is neg-

ligible. 14

The counterfactual value of trade in this case is:

~~Xijkt ¼ Xijkt0 sijkt sijkt0

� ��r : ð12Þ

Given the assumptions stated above, the conservative counterfactual refers to a case

where increases in bilateral exports reflect pure trade diversion, since the total value

of imports remains constant. In contrast, under the upper-bound counterfactual there

is an implicit assumption that trade creation will dominate.

Applying these formulas product by product for each exporter–importer pair, and

aggregating the results, allows the consequences of RTAs on agricultural and food

exports to be deduced. The results presented are based on our benchmark estimate

(r = 3.99). Among the 74 agreements studied, only the 57 that entered into force after 1998 are considered in the simulations, due to the need to use trade flows in the

year preceding entry into force as the benchmark.

6.2 Simulation results

As expected, the profile of trade impacts over time and across agreement types is

similar to that for preferential margins (Fig. 2). On average, across all the

agreements considered, ceteris paribus, RTAs increase agricultural and food

exports by 22–31 % after 5 years (as indicated respectively by the conservative

and upper-bound counterfactuals respectively), and by 30–45 % when they are

fully phased in.

The heterogeneity across agreement types when the agreement is being enforced,

grows over time. Complete implementation boosts bilateral exports by 13–15 % for

South–North flows, compared to 20–34 % for flows from the North to the South,

and 41–60 % for flows between South countries. Separate results by RTA confirm

this cross-agreement heterogeneity: some agreements have very little impact on

bilateral trade in agricultural and food products, while others result in an estimated

minimum twofold increase (see Table A.4 in Online Appendix 4). Trade impacts are

larger in the case of agreements granting higher preferential margins, particularly

when the partner’s initial market share is low.

14 The simulations in Fig. 1 depict the first calculation as the ‘‘conservative counterfactual’’, and the

second as the ‘‘upper-bound counterfactual’’.

496 S. Jean, J.-C. Bureau

123

7 Conclusions

Our analyses used a unique database that allowed the impact on trade of RTAs to be

assessed, adjusting for endogeneity and heterogeneity across both agreements and

products. We considered tariff cuts separately from non-tariff provisions. Our

estimations show that at least after the year of entry into force, RTAs boost trade

significantly based on the resulting preferential tariff margins, whose assessment is

problematic when using yearly data. We traced the impact product by product and

year by year using relevant tariff information. Our estimates suggest that on

average, a 1 % preferential margin increases pre-existing trade flows by 4 %, and

increases the probability of exporting a given product to a partner country by

0.05 %. We found moderately higher elasticity for relatively low preferential

margins and for North–South trade flows. Since the preferential margin for

agricultural and food products is often large, substantial increases in bilateral trade

flows ensue for these products, between 30 and 45 % on average. While these

results are in line with many estimates in the literature, they reveal large differences

across agreements.

0

20

40

60

0

20

40

60

0 3 6 9 12 15 0 3 6 9 12 15

All South-South

North-to-South South-to-North

Conservative counterfactual Upper-bound counterfactual

Years elapsed since entry into force

Fig. 2 Counterfactual simulations of the impact of RTAs on bilateral exports of agricultural and food products, by type of agreement and years elapsed since entry into force (%). Note The dots correspond to the full implementation of RTAs at the end of the phase-in period. The results cover agricultural and food products for 57 agreements that entered into force after 1998. Tariff-rate quota products are included in the simulations, using the inside-quota tariff rate as a measure of protection (excluding them makes little difference, except for South–North, where simulated increases are approximately 25 % lower—results available from the authors upon request). Source: Authors’ computations based on Eqs. (11) and (12) and on the benchmark estimate above

Do regional trade agreements really boost trade? Evidence… 497

123

While non-tariff provisions increasingly are seen as central trade policy issues, in

the context of RTAs we did not find a significant impact on trade. This result could be

interpreted in two ways. First, the corresponding provisions remain weakly binding

in most of the RTAs studied in this paper. This applies also to many past agreements,

especially those signed between developing countries. This leaves open the question

of whether more recent agreements, especially those involving rich countries, differ

from previous agreements in this respect. Second, agricultural and food products

might exhibit less sensitivity than manufacturing products to non-tariff provisions.

However, given inter alia the importance of sanitary and phytosanitary measures for

agricultural and food products this might seem problematic but we cannot exclude

the possibility that manufacturing products react differently to these provisions.

Further research could address both these issues.

References

Angrist, J., & Pischke, J.-S. (2009). Mostly harmless econometrics: An empiricist’s companion.

Princeton: Princeton University Press.

Baier, S. L., & Bergstrand, J. H. (2007). Do free trade agreements actually increase members’

international trade? Journal of International Economics, 71(1), 72–95.

Baier, S. L., Bergstrand, J. H. & Clance, M. W. (2015). Heterogeneous economic integration agreement

effects (CESifo Working Papers 5488). Munich: Ifo Institute for Economic Research.

Broda, C., & Weinstein, D. E. (2006). Globalization and the gains from variety. The Quarterly Journal of

Economics, 121(2), 541–585.

Bureau, J.-C., Guimbard, H., & Jean, S. (2016). What has been left to multilateralism to negotiate On?

(CEPII Working Paper, forthcoming). Paris: Centre d’Etudes Prospectives et d’Informations

Internationales.

Cameron, A. C., Gelbach, J. B., & Miller, D. L. (2011). Robust inference with multiway clustering.

Journal of Business and Economic Statistics, 29(2), 238–249.

Chaney, T. (2008). Distorted gravity: Heterogeneous firms, market structure and the geography of

international trade. American Economic Review, 98(4), 1707–1721.

Cipollina, M., & Salvatici, L. (2010). Reciprocal trade agreements in gravity models: A meta-analysis.

Review of International Economics, 18(1), 63–80.

Dür, A., Baccini, L., & Elsig, M. (2014). The design of international trade agreements: Introducing a new

dataset. The Review of International Organizations, 9(3), 353–375.

Fally, T. (2015). Structural gravity and fixed effects. Journal of International Economics, 97(1), 76–85.

Frazer, G., & Van Biesebroeck, J. (2010). Trade growth under the african growth and opportunity act.

Review of Economics and Statistics, 92(1), 128–144.

Fulponi, L., Shearer, M. & Almeida, J. (2011). Regional trade agreements—treatment of agriculture.

OECD Food, Agriculture and Fisheries Papers 44. Paris: Organisation for Economic Co-operation

and Development.

Gaulier, G. & Zignago, S. (2010). BACI: International trade database at the product-level. The

1994–2007 version (CEPII Working Paper 2010–23). Paris: Centre d’Etudes Prospectives et

d’Informations Internationales.

Ghosh, S., & Yamarik, S. (2004). Are regional trading arrangements trade creating?: An application of

extreme bounds analysis. Journal of International Economics, 63(2), 369–395.

Guimbard, H., Jean, S., Mimouni, M., & Pichot, X. (2012). MAcMap-HS6 2007, an exhaustive and

consistent measure of applied protection in 2007. International Economics, 2, 99–121.

Hallak, J. C. (2006). Product quality and the direction of trade. Journal of International Economics, 68(1),

238–265.

Head, K., & Mayer, T. (2014). Gravity equations: Workhorse, toolkit, and cookbook. In G. Gopinath, E.

Helpman, & K. Rogoff (Eds.), Handbook of international economics (Vol. 4, pp. 131–195).

Amsterdam: Elsevier-North Holland.

498 S. Jean, J.-C. Bureau

123

Head, K., Mayer, T., & Ries, J. (2010). The erosion of colonial trade linkages after independence. Journal

of International Economics, 81(1), 1–14.

Helpman, E., Melitz, M., & Rubinstein, Y. (2008). Estimating trade flows: Trading partners and trading

volumes. Quarterly Journal of Economics, 123(2), 441–487.

Kee, H.-L., Nicita, A., & Olaerreaga, M. (2008). Import demand elasticities and trade distortions. Review

of Economics and Statistics, 90(4), 666–682.

Kohl, T. (2014). Do we really know that trade agreements increase trade? Review of World Economics,

150(3), 443–469.

Kohl, T., Brakman, S., & Garretsen, H. (2016). Do trade agreements stimulate international trade

differently? Evidence from 296 trade agreements. The World Economy, 39(1), 97–131.

Manning, W., & Mullahy, J. (2001). Estimating log models: To transform or not to transform? Journal of

Health Economics, 20(4), 461–494.

Orefice, G., & Rocha, N. (2014). Deep integration and production networks: An empirical analysis. The

World Economy, 37(1), 106–136.

Romalis, J. (2007). NAFTA’s and CUSFTA’s impact on international trade. Review of Economics and

Statistics, 89(3), 416–435.

Santos Silva, J. M. C., & Tenreyro, S. (2006). The log of gravity. Review of Economics and Statistics,

88(4), 641–658.

Santos Silva, J. M. C., & Tenreyro, S. (2011). Further simulation evidence on the performance of the

Poisson pseudo-maximum likelihood estimator. Economics Letters, 112(2), 220–222.

Simonovska, I., & Waugh, M. E. (2014). The elasticity of trade: Estimates and evidence. Journal of

International Economics, 92(1), 34–50.

Sun, L., & Reed, M. R. (2010). Impacts of free trade agreements on agricultural trade creation and trade

diversion. American Journal of Agricultural Economics, 92(5), 1351–1363.

Wooldridge, J. M. (1999). Distribution-free estimation of some nonlinear panel data models. Journal of

Econometrics, 90(1), 77–97.

WTO. (2011). World Trade Report 2011. The WTO and preferential trade agreements: From co-

existence to coherence. Geneva: WTO Publications, World Trade Organization.

Do regional trade agreements really boost trade? Evidence… 499

123

Review of World Economics is a copyright of Springer, 2016. All Rights Reserved.