Post on 31-Oct-2019
Tari¤s versus Non-Tari¤ Barriers
Hiau Looi Kee and Cristina Neagu
Development Research Group � Trade, The World Bank
Oct 2011Work in Progress and Preliminary
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 1 / 22
Motivations
With the dramatic decline in tari¤s through the successive rounds ofmultilateral negotiation and unilateral liberalizations, the main tradepolicy tools could be in the form of non-tari¤ barriers (NTBs)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 2 / 22
Examples on commonly used NTBs
Origin of materials and partsImport licence feesCustoms inspection feesTesting requirementInspection requirementDirect consignmentrequirementRequirement to pass throughspeci�ed portService chargesLabelling requirementsCerti�cation requirementProcessing historySystems Approach
Temporary geographicprohibition for SPConformity assessmentsrelated to SPSTraceability requirementCerti�cation required by theexportingStorage and transportsconditionsMicrobiological criteria on the�nal pQuarantine requirementTraceability informationrequirements
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 3 / 22
Million Dollar Question
Are tari¤s and NTBs substitutes or complements?
Scant and mixed empirical evidence:
Complements �Dean, Ludema, Signoret, Feinberg and Ferrantino(2009) based on cross country evidence of NTBsSubstitutes �Kee, Nicita and Olarreaga (2009); Bown and Crowley(2009) based on time series evidence on the use of antidumping dutiesin India
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 4 / 22
Million Dollar Question
Are tari¤s and NTBs substitutes or complements?
Scant and mixed empirical evidence:
Complements �Dean, Ludema, Signoret, Feinberg and Ferrantino(2009) based on cross country evidence of NTBsSubstitutes �Kee, Nicita and Olarreaga (2009); Bown and Crowley(2009) based on time series evidence on the use of antidumping dutiesin India
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 4 / 22
Million Dollar Question
Are tari¤s and NTBs substitutes or complements?
Scant and mixed empirical evidence:
Complements �Dean, Ludema, Signoret, Feinberg and Ferrantino(2009) based on cross country evidence of NTBs
Substitutes �Kee, Nicita and Olarreaga (2009); Bown and Crowley(2009) based on time series evidence on the use of antidumping dutiesin India
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 4 / 22
Million Dollar Question
Are tari¤s and NTBs substitutes or complements?
Scant and mixed empirical evidence:
Complements �Dean, Ludema, Signoret, Feinberg and Ferrantino(2009) based on cross country evidence of NTBsSubstitutes �Kee, Nicita and Olarreaga (2009); Bown and Crowley(2009) based on time series evidence on the use of antidumping dutiesin India
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 4 / 22
What we plan to do
Exploit a newly collected dataset of NTBs by the WB on a group ofdeveloping countries
Estimate bilateral AVEs at detailed product level as rigorously aspossible
Study the factors that in�uence the level of AVEs such as countrysizes, stage of development, and bilateral distance
Relate the estimated AVEs to bilateral tari¤s
Allow the degree of substitutability or complementarity betweenNTBs and tari¤s to be in�uenced by the characteristics of thebilateral country pair
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 5 / 22
What we plan to do
Exploit a newly collected dataset of NTBs by the WB on a group ofdeveloping countries
Estimate bilateral AVEs at detailed product level as rigorously aspossible
Study the factors that in�uence the level of AVEs such as countrysizes, stage of development, and bilateral distance
Relate the estimated AVEs to bilateral tari¤s
Allow the degree of substitutability or complementarity betweenNTBs and tari¤s to be in�uenced by the characteristics of thebilateral country pair
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 5 / 22
What we plan to do
Exploit a newly collected dataset of NTBs by the WB on a group ofdeveloping countries
Estimate bilateral AVEs at detailed product level as rigorously aspossible
Study the factors that in�uence the level of AVEs such as countrysizes, stage of development, and bilateral distance
Relate the estimated AVEs to bilateral tari¤s
Allow the degree of substitutability or complementarity betweenNTBs and tari¤s to be in�uenced by the characteristics of thebilateral country pair
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 5 / 22
What we plan to do
Exploit a newly collected dataset of NTBs by the WB on a group ofdeveloping countries
Estimate bilateral AVEs at detailed product level as rigorously aspossible
Study the factors that in�uence the level of AVEs such as countrysizes, stage of development, and bilateral distance
Relate the estimated AVEs to bilateral tari¤s
Allow the degree of substitutability or complementarity betweenNTBs and tari¤s to be in�uenced by the characteristics of thebilateral country pair
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 5 / 22
What we plan to do
Exploit a newly collected dataset of NTBs by the WB on a group ofdeveloping countries
Estimate bilateral AVEs at detailed product level as rigorously aspossible
Study the factors that in�uence the level of AVEs such as countrysizes, stage of development, and bilateral distance
Relate the estimated AVEs to bilateral tari¤s
Allow the degree of substitutability or complementarity betweenNTBs and tari¤s to be in�uenced by the characteristics of thebilateral country pair
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 5 / 22
TheoriesOptimal level of protection
Countries may use trade policies to in�uence their terms of trade andtransfer rents from their trading partners
Level of protection depends on tari¤s and the ad valorem equivalent(AVE) of NTBs
tnij + AVEnij = T̄nij
Lowering tari¤s due to multilateral negotiation or unilateralliberalizations implies that NTBs will be increased to keep protectionat its optimal level ==> tari¤s and NTBs could be substitutes
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 6 / 22
TheoriesOptimal level of protection
Countries may use trade policies to in�uence their terms of trade andtransfer rents from their trading partners
Level of protection depends on tari¤s and the ad valorem equivalent(AVE) of NTBs
tnij + AVEnij = T̄nij
Lowering tari¤s due to multilateral negotiation or unilateralliberalizations implies that NTBs will be increased to keep protectionat its optimal level ==> tari¤s and NTBs could be substitutes
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 6 / 22
TheoriesOptimal level of protection
Countries may use trade policies to in�uence their terms of trade andtransfer rents from their trading partners
Level of protection depends on tari¤s and the ad valorem equivalent(AVE) of NTBs
tnij + AVEnij = T̄nij
Lowering tari¤s due to multilateral negotiation or unilateralliberalizations implies that NTBs will be increased to keep protectionat its optimal level ==> tari¤s and NTBs could be substitutes
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 6 / 22
TheoriesProtection for sale
Trade policies can be in�uenced by interests parties through lobbiesand governments care about social welfare as well as campaigncontributions
Level of protection depends on tari¤s and NTBs
tnij + AVEnij = Tnij
which increases with lobby contributions ==> tari¤s and NTBscould be complements
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 7 / 22
TheoriesProtection for sale
Trade policies can be in�uenced by interests parties through lobbiesand governments care about social welfare as well as campaigncontributions
Level of protection depends on tari¤s and NTBs
tnij + AVEnij = Tnij
which increases with lobby contributions ==> tari¤s and NTBscould be complements
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 7 / 22
Other factors that may in�uence government�s policychoices
Tari¤s are more transparent while NTBs are harder to implements,subject to corruptions
Some NTBs such as rules of origin requirement may a¤ect multipleindustries while tari¤s are more targeted
Tari¤s are under more scrutinized while NTBs are harder to monitored
Unlike NTBs, tari¤s generate revenues which could be important forsmall developing countries
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 8 / 22
Other factors that may in�uence government�s policychoices
Tari¤s are more transparent while NTBs are harder to implements,subject to corruptions
Some NTBs such as rules of origin requirement may a¤ect multipleindustries while tari¤s are more targeted
Tari¤s are under more scrutinized while NTBs are harder to monitored
Unlike NTBs, tari¤s generate revenues which could be important forsmall developing countries
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 8 / 22
Other factors that may in�uence government�s policychoices
Tari¤s are more transparent while NTBs are harder to implements,subject to corruptions
Some NTBs such as rules of origin requirement may a¤ect multipleindustries while tari¤s are more targeted
Tari¤s are under more scrutinized while NTBs are harder to monitored
Unlike NTBs, tari¤s generate revenues which could be important forsmall developing countries
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 8 / 22
Other factors that may in�uence government�s policychoices
Tari¤s are more transparent while NTBs are harder to implements,subject to corruptions
Some NTBs such as rules of origin requirement may a¤ect multipleindustries while tari¤s are more targeted
Tari¤s are under more scrutinized while NTBs are harder to monitored
Unlike NTBs, tari¤s generate revenues which could be important forsmall developing countries
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 8 / 22
Data on NTBs
Up to now the main data source is from TRAIN, which has not beenupdated since 2006
In 2010, World Bank collected detailed NTB data from about 20developing countries
Table 1 presents the share of products of each country that aresubjected NTBs and positive tari¤s
At sample mean, 41% of products face NTBs while 68% of productsface positive tari¤s
No clear correlation between the shares of products a¤ected by NTBsvs the shares of products a¤ected by positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 9 / 22
Data on NTBs
Up to now the main data source is from TRAIN, which has not beenupdated since 2006
In 2010, World Bank collected detailed NTB data from about 20developing countries
Table 1 presents the share of products of each country that aresubjected NTBs and positive tari¤s
At sample mean, 41% of products face NTBs while 68% of productsface positive tari¤s
No clear correlation between the shares of products a¤ected by NTBsvs the shares of products a¤ected by positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 9 / 22
Data on NTBs
Up to now the main data source is from TRAIN, which has not beenupdated since 2006
In 2010, World Bank collected detailed NTB data from about 20developing countries
Table 1 presents the share of products of each country that aresubjected NTBs and positive tari¤s
At sample mean, 41% of products face NTBs while 68% of productsface positive tari¤s
No clear correlation between the shares of products a¤ected by NTBsvs the shares of products a¤ected by positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 9 / 22
Data on NTBs
Up to now the main data source is from TRAIN, which has not beenupdated since 2006
In 2010, World Bank collected detailed NTB data from about 20developing countries
Table 1 presents the share of products of each country that aresubjected NTBs and positive tari¤s
At sample mean, 41% of products face NTBs while 68% of productsface positive tari¤s
No clear correlation between the shares of products a¤ected by NTBsvs the shares of products a¤ected by positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 9 / 22
Data on NTBs
Up to now the main data source is from TRAIN, which has not beenupdated since 2006
In 2010, World Bank collected detailed NTB data from about 20developing countries
Table 1 presents the share of products of each country that aresubjected NTBs and positive tari¤s
At sample mean, 41% of products face NTBs while 68% of productsface positive tari¤s
No clear correlation between the shares of products a¤ected by NTBsvs the shares of products a¤ected by positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 9 / 22
Table: Percentage of Products A¤ected by NTBs and Positive Tari¤s
Code NTB Tari¤s Code NTB Tari¤s
ARG 87.2 80.8 NAM 49.8 40.7BOL 6.1 85.5 PER 6.4 51.8COL 7.1 88.6 PHL 20.3 91.4ECU 43.3 73.8 PRY 25.1 88.1EGY 91.7 77.2 SYR 86.9 86.4IDN 42.0 73.9 TUN 22.2 68.5JPN 37.5 28.2 TZA 5.5 59.1KEN 83.1 56.0 UGA 94.0 59.1LBN 12.4 55.5 URY 6.8 80.0MEX 55.4 69.4 VEN 43.4 89.4MUS 41.6 15.8 MEAN 41.3 67.6Notes: Products are de�ned by HS6-exporter pair;
There are all together 1,234,555 products.
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 10 / 22
Figure: Share of products a¤ected by NTBs vs Tari¤s
MUSJPN
NAM
PERLBN
KEN
UGA
TZA
TUN
MEX
ECUIDN
EGY
URY
ARG
BOL
SYR
PRY
COL
VEN
PHL
.4.2
0.2
.4.6
e( N
TMsh
are
| X )
.6 .4 .2 0 .2e( postariffshare | X )
coef = .04772611, se = .34338015, t = .14
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 11 / 22
Estimating AVE of NTBs
Modify the framework of Kee, Nicita and Olarreaga (2009) toestimate bilateral AVEs
Gravity model at product level: Country i 0s import of product n fromcountry j (Mnij ) , depends on GDP of the two countries (Yi ,Yj ),distance in kilometers (Dij ), border (Bij ), tari¤s (tnij ) and thepresence of NTB (NTBni )
lnMnij = βn + β1 lnYi + β2 lnYj + β3 lnDij + β4Bij +
β5 ln (1+ tnij ) + βnijNTBni
While NTBs are importer-product speci�c, the e¤ect of NTBs onbilateral trade may depend on the characteristics of the importer andthe exporter, proxied by the total imports of product n in country ifrom the rest of the world (MniROW ) and the total exports of productn from country j to the rest of the world (EnROWj )
βnij = � exp (γn + γ1 lnMniROW + γ2 lnEnROWj )
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 12 / 22
Estimating AVE of NTBs
Modify the framework of Kee, Nicita and Olarreaga (2009) toestimate bilateral AVEsGravity model at product level: Country i 0s import of product n fromcountry j (Mnij ) , depends on GDP of the two countries (Yi ,Yj ),distance in kilometers (Dij ), border (Bij ), tari¤s (tnij ) and thepresence of NTB (NTBni )
lnMnij = βn + β1 lnYi + β2 lnYj + β3 lnDij + β4Bij +
β5 ln (1+ tnij ) + βnijNTBni
While NTBs are importer-product speci�c, the e¤ect of NTBs onbilateral trade may depend on the characteristics of the importer andthe exporter, proxied by the total imports of product n in country ifrom the rest of the world (MniROW ) and the total exports of productn from country j to the rest of the world (EnROWj )
βnij = � exp (γn + γ1 lnMniROW + γ2 lnEnROWj )
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 12 / 22
Estimating AVE of NTBs
Modify the framework of Kee, Nicita and Olarreaga (2009) toestimate bilateral AVEsGravity model at product level: Country i 0s import of product n fromcountry j (Mnij ) , depends on GDP of the two countries (Yi ,Yj ),distance in kilometers (Dij ), border (Bij ), tari¤s (tnij ) and thepresence of NTB (NTBni )
lnMnij = βn + β1 lnYi + β2 lnYj + β3 lnDij + β4Bij +
β5 ln (1+ tnij ) + βnijNTBni
While NTBs are importer-product speci�c, the e¤ect of NTBs onbilateral trade may depend on the characteristics of the importer andthe exporter, proxied by the total imports of product n in country ifrom the rest of the world (MniROW ) and the total exports of productn from country j to the rest of the world (EnROWj )
βnij = � exp (γn + γ1 lnMniROW + γ2 lnEnROWj )
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 12 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
By construction βnij varies by product, importer and exporter
Negativity constraint on βnij so that NTB always decrease trade
Negativity constraint on β5 when necessary
Estimated with non-linear LS
The resulting AVE estimate is
AVEnij =
(exp(βnij)�1
εnij, if NTB = 1
0, if NTB = 0,
where εnij is the country i 0s import demand elasticity of product nfrom country j (from Kee, Neagu and Nicita, ReStat, forthcoming)
If Mnij = 0, then use εni from Kee, Nicita and Olarreaga (ReStat,2008)
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 13 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not ideal
tried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)
tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importer
could instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
Caveats
Zero�s in trade:
we add 1 to all the trade variables before taking logs ==> not idealtried Poisson estimations but cannot introduce negativity constraints(nearly half of the estimated βnij�s are positive)tried non-linear LS in level but very di¢ cult to reach convergence
Mnij = Yβ1i Y
β2j D
β3ij exp
�β4Bij
� �1+ tnij
�β5 exp�
βnijNTBni�
Endogeneity of NTB:
not as bad as it seems since the dependent variables and regressionerrors vary by product-importer-exporter, while NTBs only vary byproduct-importercould instrument for NTBs using a Heckman�s selection equation
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 14 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates
Nearly 26 millions AVEs of NTBs are estimated (21 importers � 245exporters � 5039 HS 6)
The average AVE and tari¤ is 11% and 10%
The average AVE when AVE is positive is 28%; it is 15% for tari¤
Only about 870 thousands observations have positive trade
Among the observations with positive trade, the average AVE andtari¤ is 6.5% and 9.4%
If only focus on observations with NTBs, the average AVE is 15%;the average tari¤ is 14% among observation with positive tari¤s
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 15 / 22
AVE Estimates Assessment
Despite the large number of AVE being estimated, the results lookreasonable
AVEs of NTBs look comparable with tari¤s in terms of size on average
When NTBs are binding, AVE is almost twice as large as tari¤
Both AVE and tari¤ are lower among trading partners
AVE is lower than tari¤ among trading partners
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 16 / 22
AVE Estimates Assessment
Despite the large number of AVE being estimated, the results lookreasonable
AVEs of NTBs look comparable with tari¤s in terms of size on average
When NTBs are binding, AVE is almost twice as large as tari¤
Both AVE and tari¤ are lower among trading partners
AVE is lower than tari¤ among trading partners
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 16 / 22
AVE Estimates Assessment
Despite the large number of AVE being estimated, the results lookreasonable
AVEs of NTBs look comparable with tari¤s in terms of size on average
When NTBs are binding, AVE is almost twice as large as tari¤
Both AVE and tari¤ are lower among trading partners
AVE is lower than tari¤ among trading partners
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 16 / 22
AVE Estimates Assessment
Despite the large number of AVE being estimated, the results lookreasonable
AVEs of NTBs look comparable with tari¤s in terms of size on average
When NTBs are binding, AVE is almost twice as large as tari¤
Both AVE and tari¤ are lower among trading partners
AVE is lower than tari¤ among trading partners
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 16 / 22
AVE Estimates Assessment
Despite the large number of AVE being estimated, the results lookreasonable
AVEs of NTBs look comparable with tari¤s in terms of size on average
When NTBs are binding, AVE is almost twice as large as tari¤
Both AVE and tari¤ are lower among trading partners
AVE is lower than tari¤ among trading partners
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 16 / 22
Figure: Distribtion of AVEs for Traded Products with NTBs
05
1015
Den
sity
0 .2 .4 .6 .8 1ave_core_nl
kernel = epanechnikov, bandwidth = 0.0072
Kernel density estimate
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 17 / 22
Table: Average Estimated AVEs of NTBs and Tari¤s
Code AVEs of NTBs Tari¤sAll > 0 Traded All > 0 Traded
ARG 24.9 30.3 12 12.0 13.8 11.1BOL 1.6 25.6 0.5 8.0 8.7 6.2COL 2.0 28.2 0.6 11.9 12.5 10.7ECU 5.9 14.1 3.1 11.0 14.0 9.7EGY 23.0 26.8 12.8 14.5 17.5 12.7IDN 11.2 27.4 3.3 6.7 8.5 5.6JPN 18.7 49.4 6.9 4.0 13.4 4KEN 34.4 44.3 13 11.8 19.6 12.1LBN 2.0 17.0 1.6 6.4 10.7 8.3MEX 16.5 31.1 5.6 10.5 14.1 6.1MUS 5.3 13.1 3.5 3.4 20.0 5.7
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 18 / 22
Table: Average Estimated AVEs of NTBs and Tari¤s
Code AVEs of NTBs Tari¤sAll > 0 Traded All > 0 Traded
NAM 8.1 17.8 7.6 7.4 17.0 6.3PER 2.3 34.3 1.1 6.0 10.7 6.7PHL 4.4 25.5 2 6.2 6.4 6PRY 14.9 61.6 2.1 11.4 12.0 9.4SYR 19.8 22.7 17.8 13.2 14.3 6.4TUN 6.2 28.3 2.6 21.5 27.0 21.9TZA 1.2 21.9 0.4 12.5 19.8 13.1UGA 16.1 18.2 34.6 12.0 18.9 12.2URY 2.6 36.7 0.6 10.3 12.0 9.6VEN 9.2 22.6 4.6 12.8 13.3 12.3
MEAN 11.0 28.4 6.5 10.2 14.5 9.4
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 19 / 22
Figure: Partial Correlations by Importers at Sample Mean
MUS
IDNPHL
LBN
JPNNAM
PER
BOL
SYRMEX
KEN
TZA
UGA
EGYURYECUPRY
ARG
COL
VENTUN
.10
.1.2
e( a
ve_c
ore_
nl |
X )
.05 0 .05 .1e( tariff | X )
coef = .07952134, se = .54985548, t = .14
MUS
LBN
URYNAM
TUN
KEN
TZA
BOL
PRY
PER
UGA
ECU
ARG
PHL
EGYMEX
IDNCOL
JPN
VEN
SYR
.10
.1.2
e( a
ve_c
ore_
nl |
X )
3 2 1 0 1e( lGDPi | X )
coef = .01474484, se = .01579062, t = .93
UGA
TZA
IDN
KEN
PHLBOL
EGYSYR
PRY
PER
COL
MEXECU
ARG
VENNAMJPN
URY
TUNLBN
MUS
.10
.1.2
.3e(
ave
_cor
e_nl
| X
)
2 1 0 1 2e( lGDPpci | X )
coef = .03308281, se = .02140866, t = 1.55
MUS
TZA
MEXLBNIDN
KENJPN
ARG
UGA
URY
TUNPHL
EGYNAM
ECUCOLBOL
VENPRY
SYR PER
.10
.1.2
e( a
ve_c
ore_
nl |
X )
.4 .2 0 .2 .4e( lGDPj | X )
coef = .00840381, se = .10611599, t = .08
SYR
VENPRY
COL
NAM
BOL
ECU
UGA
URY
ARG
MEXEGYJPN
PHLTUN
TZA
PER
IDN
KEN
LBN
MUS
.10
.1.2
e( a
ve_c
ore_
nl |
X )
.2 .1 0 .1 .2e( lGDPpcj | X )
coef = .19126077, se = .22355714, t = .86
SYR
LBNTUN
EGY
PHL
NAMPERJPN
VEN
UGA
IDN
KENMUS
BOL
ECU
TZA
COLPRY
MEXURY
ARG
.10
.1.2
e( a
ve_c
ore_
nl |
X )
.6 .4 .2 0 .2 .4e( lkm1 | X )
coef = .0553735, se = .06877168, t = .81
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 20 / 22
Figure: Dependent variable: ln�1+ AVEnij
�(1) (2) (3) (4) (5) (6) (7) (8)
Tariffs 0.006*** 0.020*** 0.024*** 0.181*** 0.050 0.039*** 0.101 0.060(0.002) (0.002) (0.002) (0.030) (0.042) (0.004) (0.074) (0.085)
Tariff*GDPi 0.004*** 0.016*** 0.003* 0.006***(0.001) (0.002) (0.002) (0.002)
Tariff*GDPj 0.014*** 0.003*** 0.007*** 0.006**(0.001) (0.001) (0.002) (0.003)
Tariff*GDPPCi 0.014*** 0.060*** 0.009*** 0.004(0.003) (0.003) (0.003) (0.004)
Tariff*GDPPCj 0.000 0.002* 0.002 0.005*(0.000) (0.001) (0.003) (0.003)
Tariff*Distanceij 0.005*** 0.001 0.001 0.000(0.001) (0.001) (0.005) (0.004)
GDPi 0.008***(0.000)
GDPj 0.002***(0.000)
GDPPCi 0.012***(0.000)
GDPPCj 0.000***(0.000)
Distanceij 0.003***(0.000)
Elasticity 0.000*** 0.000*** 0.000*** 0.000***(0.000) (0.000) (0.000) (0.000)
Product FE Yes Yes No Yes No NoImporter FE Yes No No No No NoExporter FE Yes No No No No NoImporterProduct FE No No Yes Yes Yes No NoImporterExporterFE No No No No No Yes YesTariffs at sample mean 0.025*** 0.039***
(0.002) (0.003)Observations 850640 826563 850640 826563 826563 850640 826563 826563
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 21 / 22
Preliminary Conclusions
Tari¤s and NTBs could be substitutes within importer-HS6 products,comparing across exporters
Preferential tari¤s often come with rules of origin requirements
Tari¤s and NTBs could be complements within importer-exporterpair, comparing across products
Products that have low tari¤ barriers often have lower NTBs
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 22 / 22
Preliminary Conclusions
Tari¤s and NTBs could be substitutes within importer-HS6 products,comparing across exporters
Preferential tari¤s often come with rules of origin requirements
Tari¤s and NTBs could be complements within importer-exporterpair, comparing across products
Products that have low tari¤ barriers often have lower NTBs
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 22 / 22
Preliminary Conclusions
Tari¤s and NTBs could be substitutes within importer-HS6 products,comparing across exporters
Preferential tari¤s often come with rules of origin requirements
Tari¤s and NTBs could be complements within importer-exporterpair, comparing across products
Products that have low tari¤ barriers often have lower NTBs
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 22 / 22
Preliminary Conclusions
Tari¤s and NTBs could be substitutes within importer-HS6 products,comparing across exporters
Preferential tari¤s often come with rules of origin requirements
Tari¤s and NTBs could be complements within importer-exporterpair, comparing across products
Products that have low tari¤ barriers often have lower NTBs
Kee and Neagu (World Bank) Tari¤s versus Non-Tari¤ Barriers 10/11 22 / 22