Full text of Working Papers (Federal Reserve Bank of Chicago) : Trade Deflection and Trade Depression, Working Paper 2003-26

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Federal Reserve Bank of Chicago

Trade Deflection and Trade Depression
Chad P. Bown and Meredith A. Crowley

REVISED
September 2006
WP 2003-26

Trade Deﬂection and Trade Depression†
Chad P. Bown
Brandeis University

Meredith A. Crowley
Federal Reserve Bank of Chicago

This version: September 2006
Original version: November 2003
Abstract
This is the ﬁrst paper to empirically examine whether a country’s use of an import restricting
trade policy distorts a foreign country’s exports to third markets. We ﬁrst develop a theoretical model
of worldwide trade in which the imposition of antidumping and safeguard tariﬀs, or “trade remedies,” by
one country causes signiﬁcant distortions in world trade ﬂows. We then empirically test this model by
investigating the eﬀect of the United States’ use of such import restrictions on Japanese exports of roughly
4800 products into 37 countries between 1992 and 2001. Our estimation yields evidence that US restrictions
both deﬂect and depress Japanese export ﬂows to third countries. Imposition of a US antidumping measure
against Japan deﬂects trade, as the average antidumping duty on Japanese exports leads to a 5-7% increase
in Japanese exports of the same product to the average third country market. The imposition of a US
antidumping measure against a third country depresses trade, as the average US duty imposed on a third
country leads to a 5-19% decrease in Japanese exports of that same product to the average third country’s
market. We also document the substantial variation in trade deﬂection and trade depression across diﬀerent
importing countries and exported products.

JEL No. F10, F12, F13
Keywords: Deﬂected Trade, Depressed Trade, Antidumping, Safeguards, MFN

†Bown (corresponding author): Department of Economics and International Business School, MS 021,
Brandeis University, Waltham, MA 02454-9110 USA tel: 781-736-4823, fax: 781-736-2269,
email: cbown@brandeis.edu, web: http://www.brandeis.edu/cbown/
˜
Crowley: Department of Economic Research, Federal Reserve Bank of Chicago, 230 S. LaSalle St, Chicago,
IL 60604-1413 USA tel: 312-322-5856, fax: 312-322-2357, email: mcrowley@frbchi.org

Introduction

In March 2002, the United States imposed a “safeguard” - a broad-based set of tariﬀs and quotas - on imports
of steel to shield its domestic industry from foreign competition. Shortly thereafter, the European Union and a
number of other steel-importing countries responded by imposing their own import restrictions on steel.1 The
EU partially justiﬁed its trade policy by arguing that the change in US trade policy would re-route or “deﬂect”
Asian steel exports - initially destined for the newly closed US market - to what would have otherwise been a
relatively open EU market.2
Are the EU’s concerns in the steel case consistent with historical experience? When a large importing
country, such as the US, uses import restrictions such as a safeguard or antidumping duties to protect domestic
producers from imports, does this lead to the substantial deﬂection of exports to third country markets like the
EU? To our knowledge, this is the ﬁrst paper to address this question empirically. We begin by presenting a
simple theoretical model to illustrate the EU’s argument on deﬂected trade. This model embodies the potential
diﬀerential impact on world trade ﬂows of a country-speciﬁc antidumping duty (AD) versus a nondiscriminatory
safeguard measure (SG).3 We then test the model’s implications on a panel of Japanese product-level exports
from 1992-2001 that is matched at the product level to changes in US trade policy through the application of
antidumping duties and safeguard measures. We investigate whether there is evidence that the US use of such
AD and SG “trade remedies” has an impact on Japanese export patterns to third markets and then whether
1 In

addition to the EU, other WTO members that imposed at least preliminary safeguard protection on steel products between

March 2002 and October 2003 included Chile, China, Czech Republic, Hungary, Poland and Venezuela (WTO, 2002 and 2003).
Canada and Bulgaria had also initiated steel safeguard investigations after the US imposition of protection but did not apply
protection.
2 The

25 March 2002 EU press release announcing its steel safeguard response to the US steel safeguard of 5 March noted that

“[w]hilst US imports of steel have fallen by 33% since 1998, EU imports have risen by 18%. Given that worldwide there are 2
major steel markets (EU with 26.6m tonnes of imports in 2001 and US with 27.6m tonnes), this additional protection of the US
steel market will inevitably result in gravitation of steel from the rest of the world to the EU. This diversion is estimated to be as
much as 15m tonnes per year (56% of current import levels).” (European Union, 2002)
3A

‘nondiscriminatory’ trade policy (tariﬀ) is one that is applied equally to all importing countries, e.g., all imports into a

country face the same tariﬀ rate. The term most-favored-nation (MFN) will be synonymous with ‘nondiscriminatory’ for the
purpose of this paper. A ‘discriminatory’ or ‘preferential’ trade policy is one in which an importing country applies diﬀerent
tariﬀs to imports from diﬀerent exporting countries. For example, the import tariﬀ on goods from regional trade agreement
partners is usually lower than that imposed on goods from other countries. Although the WTO requires that all its members have
nondiscriminatory trade policies, there are numerous exceptions to this rule. Two of the most important exceptions in practice
are that the WTO allows for discriminatory tariﬀs when countries participate in preferential trade agreements and when countries
impose antidumping duties.

there is variation across importing countries and/or products of the size of any potential distortions.
In the empirical investigation, we use a dynamic panel data model to estimate the impact of US import
restrictions on Japanese exports to third countries. We construct a dataset of Japanese exports of roughly
4800 products into 37 countries between 1992 and 2001 to assess the eﬀect of US import restrictions, thus
exploiting the substantial variation across products and time of Japanese exports to third countries. Our
empirical approach allows us to estimate the impact of a US-imposed, Japan-speciﬁc antidumping duty on
Japanese exports, identifying whether trade is deﬂected to third markets. In addition we are able to identify
a second impact of US antidumping duties on Japanese exports; when a US duty is applied against a third
country’s exports, Japanese exports of the targeted product to the third country market are depressed.
Japan is a particularly useful starting point for such an investigation for a variety of reasons. First, Japanese
ﬁrms are frequently targeted by US acts of country-speciﬁc import protection; e.g., Japanese ﬁrms made up
10% of the US antidumping caseload that resulted in duties between 1992 and 2001.4 Over the period of our
sample, the US imposed antidumping duties on 157 unique 6-digit HS products from Japan. Furthermore, US
import restrictions targeting third country exports aﬀected an additional 167 products that Japan exports to
one or more third countries. Second, Japanese exporters are particularly prominent in world trade. Japanese
total exports as a share of world total exports was 7.5% in 1997, the midpoint of our sample.5 Third, as table 1
illustrates, Japanese exports to the US represent a substantial share of Japan’s total exports, i.e., roughly one
quarter of its total exports. This allows for the possibility of substantial trade being deﬂected to third country
markets after the imposition of a US trade restriction.
Figure 1 provides a preview of our empirical results on trade deﬂection and trade depression. The ﬁgure
presents the time path of Japanese exports to third country (i.e., non-US) markets of three diﬀerent categories
of Japanese products: 1) those for which only Japanese ﬁrms were hit with a US antidumping measure, 2) those
for which only non-Japanese ﬁrms were hit with a US antidumping measure, and 3) those for which no US
antidumping measures were applied. We use the mean growth rates of these three subsamples of observations
to plot the basic pattern in the data over a three year window: the year of the US antidumping investigation,
as well as the two preceding years. For the Japanese products that are the target of US AD cases, there is a
dramatic increase in Japanese exports to third country markets (i.e., “trade deﬂection”) in the year of the AD
4 Japan

was actually the second most targeted exporter, when measured as the number of petitions resulting in duties, as China

made up 16% of the caseload.
5 Japanese

exports as a share of world total exports peaked in 1986 at 10.3%. These calculations are based on the data provided

in Feenstra (2000) and include intra-EU trade in the calculation of world total exports.

investigation. Furthermore, for the Japanese exports of products to third countries where the third country
exporters were the target of a US AD investigation, there is a substantial reduction (i.e., “trade depression”)
of Japanese exports to that third country in the year of the AD investigation. In this paper we assess whether
the suggestive evidence presented in ﬁgure 1 is statistically and economically signicant when we control for
other factors aﬀecting Japan’s product-level export growth.
Our formal econometric results indicate that the imposition of US import restraints over the 1992-2001
period both deﬂect and depress Japanese export ﬂows to third country markets. Imposition of a US antidumping
duty against Japanese exporters is associated with substantial deﬂection of trade. For example, the median
antidumping duty against Japan leads to a 5-7% average increase in Japanese exports to a non-US trading
partner. Furthermore, there is also evidence of trade depression. When the median US antidumping duty
is imposed against a third country’s exporters, Japanese exports to the third country in the same product
category decrease by an average of 5-19%. Finally, when faced with a US safeguard measure, Japanese exports
to third countries fall by somewhere between 55% and 70%. Finally, in terms of the policy relevance, these
results provide evidence that the concerns voiced by the EU in their response to the March 2002 act of US
import protection may not be unfounded, given the historical experience with US trade remedies and the
associated Japanese export response.
Our empirical analysis, which examines how a discriminatory trade policy change aﬀects trade ﬂows, ﬁts
broadly into the literature on preferential trade agreements initiated by Viner (1950). Viner identiﬁed that
discriminatory trade policies associated with preferential trade agreements (PTAs) had both positive ‘trade
creation’ welfare eﬀects due to the enhanced trade between members (allowing members to exploit comparative advantage amongst themselves) and negative ‘trade diversion’ welfare eﬀects by potentially reducing trade
between members and non-members (and thus preventing the full exploitation of worldwide comparative advantage). Ultimately, Viner recognized that the overall welfare eﬀect of a PTA would have to be assessed
empirically. More recently, a substantial theoretical literature (including, but not limited to Bond and Syropoulos, 1996; Bagwell and Staiger, 1997, 1999, 2004; Levy, 1997; Ethier, 2004; and McLaren, 2002) examines
the role of preferential policy exceptions in multilateral trade agreements.6 These papers typically focus on
the import source diversion as the mechanism through which discriminatory trade policies aﬀect welfare; the
domestic welfare losses are derived from importing from someone who is not the lowest cost producer and
6 For

a survey of other recent papers focusing on diﬀerent theoretical elements of the interaction between preferential and

multilateral agreements, see Krishna (2004). For a literature survey on the nondiscrimination principle and the economic aspects
of the MFN clause in trade agreements, see Horn and Mavroidis (2001).

failing to (globally) exploit comparative advantage. In contrast, our analysis will focus on export diversion
where global welfare costs would arise because the low cost exporter is being shut out of a market for which it
would potentially be the most eﬃcient producer if there were non-discriminatory application of tariﬀs.
Our empirical approach is most similar to Romalis (2002) which investigates the import source diversion of
Mexican and Canadian exports to the US resulting from the North American Free Trade Agreement (NAFTA)
and the earlier Canada-US Free Trade Agreement (CUSFTA), respectively.7 Romalis uses a similarly disaggregated panel of product-level export data and ﬁnds that Mexican and Canadian shares of US imports have
increased most rapidly in the products facing the largest changes in trade policy; i.e., where the greatest PTA
tariﬀ preferences were conferred. While not the focus of his analysis, in presenting information on two of his
controls, Mexican and Canadian shares of EU imports, he documents evidence that is consistent with our results regarding the deﬂection of trade.8 Although related, our paper diﬀers from Romalis’ both in terms of the
question we ask and the empirical methodology we employ. Romalis uses a diﬀerence-in-diﬀerences approach
to examine if changes in US trade policy (NAFTA and CUSFTA) produced import source diversion. We use
dynamic panel data methods to examine if changes in US trade policy (AD and SG) are associated with the
deﬂection and also the depression of exports.
With respect to the economics literature on trade remedies, our results on the trade distorting eﬀects of
antidumping measures complement the work of Prusa (1997, 2001).9 Prusa (2001), for example, uses a panel
of US industry-level imports and data on US antidumping measures for 1980-1994 to investigate the import
source diversion that occurs for the United States when a discriminatory trade policy causes importers to
switch from a lower cost to a higher cost foreign supplier. He provides evidence that foreign exporters in an
industry subject to a US AD measure who are not-named in an antidumping petition increase exports to the
US in conjunction with the exports of the country targeted by the AD petition falling. Our paper can provide
insight as to where the products targeted by the US petition go, since they are no longer being exported to
the US market.10
7 Clausing

(2001) is another recent paper that looks at the trade creation and trade diversion eﬀects of the CUSFTA through an

analysis of a panel on product-level trade data and tariﬀ changes. Romalis (2002) presents a thorough discussion of the diﬀerences
in approaches and results of the two papers.
8 We

interpret Romalis’ Figures 1B and 1C for Mexico and Figures 2B and 2C for Canada as evidence of the deﬂection of

exports from the EU to the US. Starting from a non-discriminatory benchmark, the discriminatory removal of US trade barriers
lowers tariﬀs facing a Mexican or Canadian product, and leads to trade being deﬂected away from a third market like the EU.
9 For

a recent survey of the economics literature on antidumping, see Blonigen and Prusa (2004). For a survey on safeguards

protection, see Bown and Crowley (2005).
10 While

we do not investigate the issue here, our results suggest that there may also be substantial welfare distorting eﬀects

The rest of this paper proceeds as follows. Section 2 presents our simple economic model to ﬂush out
our empirical predictions. Section 3 presents our empirical model that will be used in the estimation and
a discussion of variable construction and data. In section 4 we discuss our empirical results, and section 5
concludes with a discussion of additional questions and further puzzles for future research.

Theoretical Model

Assume there are three countries indexed i or j ∈ {A, B, C}, i = j. Each country has one ﬁrm, also indexed
i or j, which produces a single good for domestic consumption and for export. A good is denoted mij , where
the ﬁrst index, i, indicates the country of production, and the second index, j, indicates the country in which
the good is consumed. Thus, a good produced by ﬁrm i for export to country j is denoted mij . Output
produced for domestic consumption is denoted mii . Markets are segmented, ﬁrms compete on quantity, and
the good produced for domestic consumption and the imported goods are strategic substitutes (πmii mji < 0,
πmji m−ji < 0).
Production in each country employs the same technology. The marginal cost of production is increasing,
the cost function is c(xi ) where c (xi ) > 0 and c (xi ) > 0 and xi is ﬁrm i’s total output. Firm i’s total output
is the sum of domestic sales and sales in the two foreign markets, xi =

mij , j ∈ {A, B, C}.

Inverse demand in all countries is given by p(Qi , Yi ) where Qi is the total output sold in country i and Yi
is national income. Total output sold in i is the sum of domestic sales by the domestic ﬁrm and imports from
the other two countries, Qi =

mji , j ∈ {A, B, C}.

The objective of the ﬁrm in i is to chose a total output level and a level of sales for each market in order
to maximize proﬁts,

p(Qj )mij − τij mij − c(xi ),

max πi =
mij

(1)

where τij represents country j’s tariﬀ on imports from i and τii , the tariﬀ on consumption of the domestically
produced good, is equal to zero. The ﬁrms’ ﬁrst order conditions are given by the following:

∂πi
= p(Qj ) + p (Qj )mij − τij − c (xi ) = 0.
∂mij

(2)

of acts of US trade protection outside of the US, in addition to the sizable welfare distortions experienced inside the US and
documented, for example, by Gallaway, Blonigen and Flynn (1999).

Solving the ﬁrst order conditions for each j ∈ {A, B, C} yields ﬁrm i’s best responses to the sales decisions of
the other two ﬁrms. A best response function speciﬁes an amount to sell in each market, given the sales in that
market of the ﬁrm’s two rivals. Solving the nine best response functions simultaneously yields the Cournot
Nash equilibrium quantities sold by each ﬁrm in each country.

∀i, j ∈ {A, B, C}

mij = f p(Qi , Yi ), c(xi ), τij

(3)

In the Cournot Nash equilibrium, because the marginal cost of production is increasing, each ﬁrm will
choose to allocate its total output across the three countries so that its net marginal revenue (marginal revenue
less tariﬀ costs) is the same in all three markets.

2.1

Comparative statics for an antidumping duty

Without loss of generality, suppose that trade among the three countries is free, with the exception that country
A imposes a tariﬀ on imports from country B. How will an increase in A’s tariﬀ aﬀect trade among all three
countries? Figure 1 provides an illustration of proposition 1.
Proposition 1 For the three country Cournot model in which goods are strategic substitutes and ﬁrms face
increasing marginal costs in production, a tariﬀ by country A against country B causes, relative to the free
trade equilibrium:
ba
1. trade destruction, a decline in country B’s exports to country A ( dmba < 0),
dτ

ca
2. trade creation via import source diversion, an increase in country C’s exports to country A ( dmba > 0),
dτ

3. trade deﬂection, an increase in country B’s exports to country C( dmbc > 0), and
dτba
4. trade depression, a decrease in country C’s exports to country B ( dmcb < 0).
dτba
Proof: Totally diﬀerentiating the nine ﬁrst order conditions given by (2), dividing through by dτab , and
applying Cramer’s rule yields the signs of the comparative static eﬀects on the domestic output and exports
of all three ﬁrms of an increase in country A’s tariﬀ on imports from country B. For strategic substitutes and
an increasing marginal cost of production, without loss of generality, the following results are obtained for a
change in τba : for goods consumed in country A,
B,

dmbb
dτba

> 0,

dmab
dτba

< 0,

dmcb
dτba

dmaa
dτba

> 0,

dmba
dτba

< 0,

< 0, for goods consumed in country C,

dmca
dτba
dmcc
dτba

> 0, for goods consumed in country
< 0,

dmbc
dτba

> 0,

dmac
dτba

< 0. QED.

In this model, the existence of a deﬂected trade ﬂow relies critically on the assumption of an increasing
marginal cost of production. Because ﬁrms equate the net marginal revenue of producing for each market in
equilibrium, anything that raises the cost of selling in one market will cause ﬁrms to reallocate their sales across
markets.

2.2

Comparative statics for a safeguard tariﬀ

Without loss of generality, suppose that trade among the three countries is free, with the exception that country
A imposes a tariﬀ on imports from countries B and C. Assume that the magnitudes of the tariﬀs set against B
and C are identical (τba = τca ) and given by τ . How will an increase in country A’s tariﬀ aﬀect trade among
all three countries?
Proposition 2 For the three country Cournot model in which goods are strategic substitutes and ﬁrms face
increasing marginal costs in production, a tariﬀ by country A against all other countries (B and C) causes,
relative to the free trade equilibrium:
1. trade destruction, a decline in country C and B’s exports to country A ( dmba < 0,
dτ

dmca
dτ

< 0) and

2. two-way trade deﬂection, an increase in country B’s exports to country C( dmbc > 0) and an increase in
dτ
country C’s exports to country B( dmcb > 0).
dτ
Proof: Totally diﬀerentiating the nine ﬁrst order conditions given by (2), dividing through by dτ , and
applying Cramer’s rule yields the signs of the comparative static eﬀects of an increase in country A’s tariﬀ on
imports from all countries on the domestic output and exports of all three ﬁrms. For strategic substitutes and
an increasing marginal cost of production, without loss of generality, the following results are obtained for a

dmbb
dτ

> 0,

dmab
dτ

< 0,

dmcb
dτ

dmaa
dτ

dmba
dτ

< 0,

dmca
dτ

< 0, for goods consumed in country

> 0, for goods consumed in country C,

dmcc
dτ

> 0,

change in τ : for goods consumed in country A,

> 0,

dmbc
dτ

> 0,

dmac
dτ

< 0. QED.

Comparing a discriminatory antidumping policy and a nondiscriminatory safeguard, the theoretical model
predicts that two phenomena observed under an antidumping duty - trade creation via import source diversion
and trade depression - are absent under a safeguard. Because a safeguard creates an identical increase in
costs on products from both import sources, there is no incentive to favor one source over another. Thus, the
result that no trade is created through import source diversion is fairly obvious. With regard to the model’s
prediction of two-way trade deﬂection under a safeguard, the result is less obvious. For each country B and
C, the safeguard induces two conﬂicting forces of trade depression and trade deﬂection. Retained domestic
7

production that can no longer be sold in country A could “crowd out” imports and lead to trade depression,
but in the model this eﬀect is swamped by each ﬁrm’s strong desire to export so that it will not be competing
against itself in its domestic market.
In the next section, we test the model’s predictions about trade deﬂection, trade depression, and two-way
trade deﬂection on a panel of Japanese product exports. Our approach is thus diﬀerent from papers by Romalis
(2002) and Prusa (1997, 2001) who estimate empirical models of what we refer to as “trade destruction” and
“trade creation via import source diversion,”11 respectively.

Empirical Model and Estimation

3.1

The empirical investigation

The theoretical model presented in section 2 yields a number of predictions relating one country’s tariﬀs to trade
ﬂows between foreign countries. Our empirical analysis focuses on the predictions of deﬂected, depressed and
two-way deﬂected trade for Japanese exports to 37 non-US trading partners. For clarity of exposition, ignoring
Japan’s 36 other trading partners, what does our theoretical model predict when the country imposing tariﬀs
is the US and the foreign countries are Japan and the EU? First, if the US imposes a country-speciﬁc tariﬀ
against Japan in the form of an antidumping duty and imposes no tariﬀ against the EU, the model predicts
deﬂected trade, an increase in Japanese exports to the EU. Second, if the US imposes a country-speciﬁc tariﬀ
against the EU in the form of an antidumping duty, but not against Japan, the model predicts that Japanese
exports to the EU will fall, i.e., depressed trade. In this case, European exports that are diverted away from
the US market by the tariﬀ and sold domestically within the EU depress imports from Japan. Third, if the
US imposes tariﬀs against both Japan and the EU in the form of a broadly-applied safeguard measure or two
simultaneously-imposed antidumping duties, the model predicts two-way deﬂected trade, a rise in Japanese
exports to the EU and in EU exports to Japan.12

3.2

Basic empirical model

To investigate the questions identiﬁed by the theoretical model, we develop the following reduced-form speciﬁcation for the value of Japanese exports to country i based on equation (3):13
11 More
12 We

precisely, Romalis estimates the opposite eﬀect - trade creation arising from the removal of a tariﬀ.

do not test for the rise in EU exports to Japan here as our analysis focuses on the response of Japanese exports only.

13 Unfortunately,

only the value of imports is consistently available in the TRAINS data, so we cannot analyze the price and

quantity responses to a trade policy change separately.

ln(vmiht ) = αi + γh

+ β1 ln(Yt ) + β2 ln(Yit ) + β3 ln(eit ) + β4 τht + β5 τiht + β6 τt∗i
+ β7 ln(ckt ) + β8 ln(vmiht−1 ) +

(4)

iht ,

where i denotes an importing country, h denotes a 6-digit HS product, and t denotes time in years. The index
k denotes an industry aggregate at the 3-digit ISIC level, i.e. the products h = 1..h map into the industries
k = 1, h = h ..h map into k = 2, and so on until h = h∗ ..H map into k = K.
The variable vmiht denotes the value of imports from Japan of h into i at time t, Yt denotes Japan’s national
income (an export-supply shifter), Yit denotes the importing country i’s national income (an import-demand
shifter) and eit is the exchange rate between the yen and the importing country’s currency. The variable τht
designates US trade policy against Japan while τiht captures US trade policy against importing country i.
Japan’s industry k cost variables are denoted by ckt , while importing country i’s trade policy is denoted by
τt∗i .14 Finally, αi denotes country ﬁxed eﬀects, γh denotes product ﬁxed eﬀects, the β’s are the parameters to
be estimated, and

3.3

iht

is the error term.

Estimation strategy

There are two problems to address in estimating equation (4). First, the autocorrelation of vmiht implies that
least squares estimation of (4) yields biased estimates. Second, in a short panel, the number of parameters to
be estimated (αi and γh ) increases with the number of countries and products. Thus, αi and γh cannot be
consistently estimated.
To address both of these problems, we estimate the ﬁrst diﬀerence of (4) using the optimal Generalized
Method of Moments (GMM) estimator proposed by Arellano and Bond (1991), in which multiple lags of the
level of the dependent variable are used as instruments for lags of the ﬁrst diﬀerence of the dependent variable.15
We thus use GMM to estimate

∆ln(vmiht )

= β1 ∆ln(Yt ) + β2 ∆ln(Yit ) + β3 ∆ln(eit ) + β4 ∆τht + β5 ∆τiht + β6 ∆τt∗i
+β7 ∆ln(ckt ) + β8 ∆ln(vmiht−1 ) + ∆

14 Ultimately

(5)

iht .

it would be preferable to also have product-level data for country i’s trade policy and Japan’s costs; unfortunately

neither of which is yet systematically available over this time period.
15 Direct

estimation of the ﬁrst diﬀerence of (4) by least squares would yield biased coeﬃcients because the lagged diﬀerence of

imports [ln(vmiht−1 ) − ln(vmiht−2 )] is correlated with the error term [

iht

−

iht−1 ].

3.4

Fixed eﬀects model

One potential criticism of our basic empirical model is that it does not adequately control for product-level
variation in production costs because our industry cost variables, ckt , are only available at a 3-digit industry
level whereas our trade data and policy changes are measured at a 6-digit product level. Therefore, as a
robustness check, we ﬁrst diﬀerence (5) and use 6-digit HS product ﬁxed eﬀects and country-year dummies to
control for detailed product-level variation as well as country-speciﬁc macroeconomic variation over time. Our
ﬁxed eﬀects model is given by the following:

∆ln(vmiht ) = µh + χit + η1 ∆τht + η2 ∆τiht + η3 ∆ln(vmiht−1 ) + ∆

iht ,

(6)

where µh are 6-digit HS product-speciﬁc ﬁxed eﬀects and χit represents a full set of country-year dummies.
Because of the large number of parameters to be estimated, Arellano and Bond’s GMM estimator is not
computationally feasible. Therefore, we estimate (6) using the two stage least squares/instrumental variables
(IV) approach of Anderson and Hsiao (1981, 1982) in which we instrument for ∆ln(vmiht−1 ) using the second
lag of the log level of the value of imports. Because the product ﬁxed eﬀects and the country-year dummies
absorb product-level and macroeconomic variation over time, this approach requires fewer control variables
(e.g. GDP growth, value-added per worker) than estimation of equation (5) and thus we are able to utilize a
much larger sample of trade and trade remedy data for many additional countries.
Nevertheless, because of the dynamic panel structure of our data, there are two potential problems with
the Anderson and Hsiao (1981, 1982) IV estimator; bias associated with the use of a weak instrument and
bias associated with correlation in measurement error. In appendix A we address both of these concerns. To
address the weak instrument problem, we test the quality of two instruments, ln(vmiht−2 ) and ln(vmiht−3 ).
We ﬁnd that both are strong instruments for ∆ln(vmiht−1 ) and conclude the IV approach is appropriate for
our problem. To address the issue of measurement error, we compare coeﬃcient estimates using the second
and third lags of the log level of imports. We ﬁnd that our coeﬃcient estimates are robust to the choice of
instrument, suggesting that measurement error is not a signiﬁcant problem and the use of ln(vmiht−2 ) as an
instrument is appropriate.

3.5

Variable construction and data

In this section we discuss the construction of variables used in the estimation of equations (5) and (6) as well
as the sources of our data. Table 2 summarizes variable descriptions and our predictions about the signs of the

estimated coeﬃcients, as well as providing summary statistics.

3.5.1

Trade variables

First consider the dependent variable in the estimation of equations (5) and (6), ∆ln(vmiht ), which is the
annual growth of third country i’s imports of product h from Japan. The detailed, highly disaggregated data
used in this paper represent a signiﬁcant improvement over many previous studies on US import restrictions
and trade remedies. Annual data on the nominal value of imports into 37 non-US countries for roughly 4800 6digit Harmonized System (HS) products for the years 1992 to 2001 come from UNCTAD’s TRAINS data base.
Import data for these 37 countries was reformatted into a dataset of Japanese exports to these countries.16 In
our basic speciﬁcation of equation (5), we are restricted to using a smaller set of 28 importing third countries due
to the limited availability of some of the macroeconomic data needed for the estimation. The alternative ﬁxed
eﬀects model (6) requires no macroeconomic data and utilizes a larger sample of 37 importing countries. The
countries included in the ﬁnal dataset include OECD members, many countries from Asia and Latin America,
and some former members of the USSR and Eastern European countries. Data on Africa is generally not
available in TRAINS, but as these countries are extremely small markets for Japanese exports, their omission
should not aﬀect our results. Table 1 also lists the countries used to estimate the diﬀerent speciﬁcations.
Because the TRAINS dataset does not include product-speciﬁc price deﬂators, we deﬂated the nominal import
data, which is reported in US dollars, using the US Bureau of Labor Statistic’s HS Import Price Indices, which
are available for the period 1992 through 2001.

3.5.2

US antidumping and safeguard policy variables

The main explanatory variables of interest in equations (5) and (6) are the changes to US import policy facing
a product h exported to the US from Japan (∆τht ) or from a third country (∆τiht ). Our estimates use data
on the country-speciﬁc, trade-weighted average of the antidumping duty in the year in which the antidumping
16 Because

this data is collected only on the import side, it is possible that discrepancies exist between a country’s imports from

Japan and Japan’s exports to that country. We checked the quality of our Japanese export dataset against Feenstra’s (2000)
NBER’s World Trade Database (WTDB). The WTDB includes data on worldwide import and export ﬂows at a 4-digit SITC level
and is thus too aggregated for our purposes, but is known to be of high quality because it matches import and export records
to resolve any discrepancies in the values of trade ﬂows between pairs of countries. Table 1 presents a comparison of Japan’s
aggregate export shares in 1996 calculated using our dataset and the WTDB. The shares from the two datasets are comparable
and we feel conﬁdent in using the TRAINS data in our analysis. Nevertheless there are some years for which trade data is missing
for certain countries.

measure was imposed.17 For a product h, we examine the eﬀect of (1) the imposition of a US antidumping duty
against Japan, (2) the imposition of a US antidumping duty against a third country i, and (3) the imposition of
a US safeguard policy. As discussed in section 2, the theoretical model predicts that the sign of the coeﬃcient
on (1) is positive, on (2) is negative, and on (3) is positive.18
We collected data on the US imposition of country-speciﬁc antidumping duties and safeguard measures
at the 6-digit HS level over the 1992 through 2001 period from a variety of US government publications,
most notably the Federal Register. For antidumping and safeguard cases ﬁled during the sample, we obtained
the names and 6-digit HS codes for the products involved, the outcome of the case (aﬃrmative, negative, or
terminated, as well as the type of measure for safeguards), the names of the countries that faced the import
restrictions, the trade-weighted average duty when duties were imposed, and, most importantly, the date a
case was initiated and the date a trade restriction began.19 For the antidumping policies, we interact a variable
indicating that the policy was imposed in year t with the level of the antidumping duty that is imposed, to
help control for the heterogeneity in duties imposed across exporters and across investigations. On the other
hand, we use a simple indicator variable to examine the safeguard policies, due to the fact that sometimes the
safeguard measure is imposed as a quantitative restriction or a tariﬀ-rate quota, as opposed to a simple ad
valorem duty.
17 The

duty

data

was

generously

provided

Bruce

Blonigen

and

his

website

http://darkwing.uoregon.edu/ bruceb/adpage.html .
18 We

do not investigate the impact on Japanese exports of US AD (or SG) investigations that do not result in duties, but

which are terminated or settled. While such an investigation could lend further insight into the overall impact of Staiger and
Wolak’s (1994) “investigation eﬀect,” “suspension eﬀect” and “withdrawal eﬀect,” of the non-duty impact of AD investigations, it
is beyond the scope of issues under investigation here and thus we leave it to future research. Furthermore, while the theoretical
model also generates predictions on the expected impact of the removal of trade remedies, we do not report estimates of those
impacts here. The primary issue is the quality of the antidumping policy removal data. We are primarily concerned with
measurement error as it is diﬃcult to “time” when an antidumping measure is removed, given that in many cases the removal is
applied retroactively (with the refunding of duties), but in a manner which could not have been reasonably anticipated by the
exporters. Nevertheless, we do note that the estimates (available from the authors upon request) for our policy variables of interest
are not signiﬁcantly altered with the inclusion of the policy removal variables.
19 To

clarify the timing of our diﬀerent variables, the variables ∆ AD Dutyjpn,ht and ∆ AD Dutyiht are nonzero in the period

in which the investigation into an antidumping case that results in a duty is begun. The ∆ SG Policyht variable is an indicator
that is equal to 1 in the period in which a safeguard measure goes into eﬀect. This reﬂects the fact that in AD cases the targeted
exporters begin to respond to provisional antidumping duties that are imposed shortly after the date the investigation is announced.
Safeguard cases, on the other hand, have a very uncertain outcome and almost never use temporary trade restrictions during the
investigation phase.

3.5.3

Macroeconomic variables

We include controls for growth in the exporting (∆ln(Yt )) and importing countries’ GDP (∆ln(Yit )), growth of
the real yen/country i currency exchange rate (∆ln(eit )), and proxies for changes in the importing countries’
trade policies (∆τt∗i ).
We expect an increase in the GDP growth of the exporting country (Japan) to lead to a fall in Japanese
export growth because domestic demand for the export goods will be higher. In other words, Japan is expected
to export domestic weakness. Second, in terms of currency changes, export growth should be higher when the
yen is weakening relative to the importing country’s currency. Thus, we expect a positive sign on the coeﬃcient
for growth of the real exchange rate.
For the importing country, an increase in GDP growth should be associated with higher Japanese export
growth. To proxy for changes to an importing country’s overall trade policy that we cannot observe, for example,
an across the board tariﬀ reduction or a reduction in the administrative cost of exporting to a particular
importing country, we control for changes in an importing country’s “openness.” Openness is deﬁned as the
sum of real aggregate imports and exports divided by real GDP. For some countries, real aggregate import
and export series were not available. For these countries, we calculated “openness” using the corresponding
nominal variables. We believe that an increase in this variable is associated with liberalization of country i’s
trade policy, and thus, expect a positive sign on its coeﬃcient.
Data on real GDP, real aggregate imports, and real aggregate exports come from two sources: the OECD
Main Economic Indicators and the IMF’s International Financial Statistics (IFS). Whenever possible, we used
the OECD data to construct the macroeconomic controls. When OECD data were not available or were only
available for a short timespan, we used data from the IFS. We construct real bilateral Japanese Yen to country
i currency rates for 20 countries using data supplied by the USDA Economic Research Service. An increase in
the value of the real exchange rate implies an appreciation of country i’s currency. For Norway and Switzerland,
bilateral rates were not available from USDA so we use real exchange rate indices from the IFS.20

3.5.4

Industry-level variables

Lastly, we use two measures of productivity changes for Japanese manufacturing industries: the growth of the
average wage and the growth of value-added per worker. This addresses a concern that our policy variables
20 For

the IFS series, an increase in the value of the real exchange rate index implies a real appreciation of the Norwegian and

Swiss currencies, respectively. We thank Matthew Shane of the ERS of the USDA for providing us with the data and answering
questions about the construction of the USDA’s real exchange rates.

may not be measuring true treatment eﬀects, but may be picking up the eﬀect of an omitted variable - like a
Japanese productivity improvement - that would be associated with the imposition of a US import barrier on
Japanese imports and an increase in Japanese export growth to other countries. We expect the sign on both
productivity measures to be positive.
Japanese manufacturing industry data at the 3-digit ISIC (Rev. 2) level for the years 1992-1999 came from
the UNIDO (2002). We used data on number of employees, value-added and average wages to construct two
productivity measures: the growth of value-added per worker and the growth of average wages.

Empirical Results

4.1

Estimation results using the GMM procedure

Table 3 presents our estimates of equation (5) using the Arellano and Bond (1991) GMM estimator. Speciﬁcations (1) and (2) present estimates on the full set of industries (agricultural and manufacturing) over the
1992-2001 period thus leaving out the industry-level controls. Speciﬁcations (3) through (5) present estimates
for all manufacturing industries from 1992-1999, all years for which the ISIC industry variables are available.
Consider ﬁrst speciﬁcation (1) and our estimates for the policy variables of interest, which provide evidence
in support of some of the key predictions of our theoretical model. US imposition of antidumping measures
against Japan is associated with statistically signiﬁcant deﬂection of Japanese exports to third country markets,
and US imposition of antidumping measures against third countries is associated with a statistically signiﬁcant depression of Japanese exports to those markets. Interestingly, the coeﬃcient estimate on the safeguard
indicator variable also implies that a safeguard policy is associated with a statistically signiﬁcant depression of
Japanese exports to third markets.
With respect to the size of the estimates, speciﬁcation (1) indicates that the imposition of a 1% antidumping
duty against Japanese exports of product h (but not exporters from country i) is associated with an 0.14%
increase in Japanese exports of h to country i. To understand the magnitude of the eﬀect, consider that the
median antidumping duty (conditional on a duty being imposed) facing a Japanese exporter in the sample is
37.13%, which implies a 5.24% average increase in Japanese exports of h to an importing country i. In the
next section we investigate whether the magnitude of the trade deﬂection eﬀect facing Japanese exports varies
substantially across some of its particularly important trading partners i as well important product categories
of h.

On the other hand, the imposition of a 1% antidumping duty against the third country i’s exporters, but
not against Japan, is associated with a 1.269% reduction in Japanese exports of that same product h to country
i. This is consistent with the idea that when the output produced by ﬁrms in country i cannot be sold in the
US, but is sold domestically, it depresses (crowds out) country i’s imports of the same product from Japan.
With the median duty facing a country i’s exports of h being 14.84% in this particular sample of data, this
translates into an 18.83% average reduction in Japanese exports to a third country, which is also a signiﬁcant
eﬀect.
Finally, and perhaps surprisingly, the US imposition of an MFN safeguard policy also has a strong trade
depressing eﬀect. After using the Kennedy (1981) formula to convert the coeﬃcient estimate for the dummy
variable to its marginal eﬀects interpretation, the imposition of a US SG on an HS product implies a 67%
reduction in Japanese exports of that product to country i.21
Next, the coeﬃcient estimates in speciﬁcation (1) for the macroeconomic and “openness” control variables
also have the predicted sign. Since they are not of particular interest to our investigation and are fairly robust
across speciﬁcations, we will omit a substantive discussion of them here.
In speciﬁcations (2) through (4) of table 3, we sequentially consider additional control variables as one
way to check the sensitivity of our results. Overall, the estimates on the policy variables of interest appear
robust to changes in model speciﬁcation. For example, the estimated impacts of the growth in the values
of the Yen/country i exchange rate, which we add in speciﬁcation (2), is positive as predicted by the theory.
Nevertheless, we do lose a substantial number of observations from the sample when we add in the real exchange
rate variable.
More importantly, in speciﬁcations (3) and (4) we add industry-level (ISIC 3-digit) control variables (available from 1992-1999). Because our results for the policy variables are robust to the inclusion of industry level
controls, we believe that the policy variables are likely capturing the true treatment eﬀect of the policy. Unfortunately, because the industry variables are only available for manufacturing industries between 1992 and 1999,
we also lose a number of observations in these speciﬁcations. Speciﬁcation (5) shows that the small changes
to the estimates for the policy variables of interest (the slight increase in the size of the AD duty imposition
21 While

safeguard measures often include exemptions for free-trade partners and developing countries, in our sample, safeguard

measures were applied quite broadly to almost all US import sources. The correlation coeﬃcient between changes in US safeguard
policy against Japan - the variable in our estimation - and against all other countries was 0.73. In contrast, there was considerably
more variation in the application of antidumping duties. The correlation coeﬃcient between changes in US antidumping policy
against Japan and against all other countries was only 0.24.

variables and decrease in the statistical signiﬁcance of the SG policy imposition variable) are most likely due to
the loss of observations from years 2000 and 2001. For the case of the safeguard variable in particular, this is
likely due to the variation generated by the observations surrounding the 2000 US safeguard policy on circular
welded pipe that particularly aﬀected global pipe trade from and into markets such as Korea, Japan and the
EU.22
At the bottom of table 3 we also report the z-statistic on the average autocovariance in residuals of order
2 for speciﬁcations (1) through (5). For all speciﬁcations, we are able to reject the hypothesis of second order
autocovariance in the residuals, which leads us to conclude that our Arellano and Bond GMM estimator yields
consistent parameter estimates. In all of the speciﬁcations reported in (1) through (5) we include two lags of
the dependent variable, as we found that inclusion of a second lag improved the ﬁt of the model and yet did
not signiﬁcantly change our parametric estimates.
Finally, speciﬁcation (6) presents a ﬁnal robustness check on the results in table 3. Using the sample from
speciﬁcation (1), we estimate the ﬁxed eﬀects speciﬁcation of equation (6) with the Anderson and Hsiao (1981,
1982) instrumental variables technique, as described in section 3.4. The results in speciﬁcation (6) are broadly
consistent with those in speciﬁcations (1) - (5). The estimated impact of a 0.127% increase in exports in
response to a 1% increase in the US antidumping duty against Japan falls within the 95% conﬁdence intervals
of the estimated coeﬃcients in speciﬁcations (1) - (5), but is not signiﬁcantly diﬀerent from zero. We will show
in the next section, however, this result appears to be driven by the particular sample of countries and years
available in the data set required for the GMM estimation.
To summarize the results of table 3, we ﬁnd ﬁrst that the US imposition of an AD duty against Japan
leads to a deﬂection of Japanese trade to third markets (row 1): Japanese exports to third markets increase
by estimates ranging from 0.127% to 0.168% for each 1% increase in the US duty. Second, the US imposition
of an AD duty against a third country is associated with the depression of Japanese exports to those third
markets (row 2): Japanese exports to third markets fall by 0.870% to 1.292% for each 1% increase in the duty.
Third, the US imposition of a broadly applied SG measure against Japan and other exporting countries leads
to a depression of Japanese trade to third markets (row 3): Japanese exports to third markets fall by 63% to
70%.
22 Korea

was the largest exporter adversely aﬀected by the US policy, so much so that it contested the measure through a formal

WTO trade dispute. It is therefore likely that because of a glut of Korean pipe being retained domestically, Japanese exports of
pipe to Korea were depressed in 2000, thus driving the signiﬁcance of the safeguard results in speciﬁcations in which the year 2000
data is included.

Even though the trade depressing eﬀect of a US safeguard measure is statistically signiﬁcant, we are concerned about the robustness of this particular result. While there were hundreds of US AD measures imposed
over the 1992-2001 period, there were only ﬁve US SG investigations which resulted in the imposition of deﬁnitive measures (tariﬀs, quotas or tariﬀ-rate quotas). Even though each of the SG measures may aﬀect more than
one 6-digit HS category, we are nevertheless concerned about the relatively few number of safeguard observations in the estimation. This concern is further driven by the fact that some US safeguard measures covered
products (e.g., brooms, lamb meat, wheat gluten) which were not of substantial importance to Japanese exporters. Nevertheless, our results with regard to the imposition of a safeguard measure do reject our theoretical
model’s prediction of two-way trade deﬂection.

4.2

IV estimates using ﬁxed eﬀects from an expanded sample of data

As described in section 3.4, an advantage to using the ﬁxed eﬀects model and instrumental variables estimation
procedure is that it does not require macroeconomic and industry controls. Thus, we can estimate the model
using a signiﬁcantly larger sample of trade data.23 Therefore, in table 4 we provide a set of estimation results
using data on Japanese exports to all of the countries listed in table 1 except the US. When compared to the
sample of speciﬁcation (6) of table 3, for example, this adds to the estimation sizable import markets such
as Taiwan, China, Singapore and the Philippines, in addition to requiring one fewer lag of the dependant
variable in the estimation, providing eﬀectively another year (1994) of trade remedy data. Together, these
elements add roughly 113,000 observations. Estimates of the coeﬀcients on the policy variables are consistent
with our ﬁndings of trade deﬂection and trade depression reported in table 3. In Appendix A we formally
describe the tests we perform on the strength of our instrumental variables. Because F-tests conﬁrm that all of
our instrumental variables are strong instruments, we conclude that our instrumental variables estimates are
unbiased.
Speciﬁcation (7) of table 4 presents the baseline IV speciﬁcation for comparison with the results of table 3.24
The estimates for trade deﬂection and trade depression are statistically signiﬁcant on this larger sample of
data. The primary change in results from table 3 relates to the size of the coeﬃcient estimate on the trade
depression eﬀect of an antidumping duty against country i, as it falls from -1.271 in speciﬁcation (1) to -0.281
in speciﬁcation (7).
Nevertheless, to better interpret and compare the magnitude of the estimates, consider table 5. The ﬁrst
23 The

results presented in table 4 use country-year dummies and 6-digit HS product ﬁxed eﬀects.

24 Note

that speciﬁcation (10) presents the biased OLS estimates as a benchmark for comparison.

column presents the median duty, conditional on a duty being applied, for the sample of 141,164 observations
used in speciﬁcations (1) - (6). The second column presents the median duty, conditional on a duty being
applied, for the sample of 254,074 observations used in speciﬁcations (7) -(10). The third column quantiﬁes
the eﬀect of imposing a typical duty, in this case, the conditional median25 using the coeﬃcient estimated from
speciﬁcation (1). The fourth and ﬁfth columns similarly quantify the eﬀect of imposing a typical duty using the
coeﬃcient estimates from speciﬁcations (7) and (8). First, comparing our estimates from speciﬁcations (1) and
(7), we see that the magnitudes of the trade deﬂection eﬀect in the two samples, using two diﬀerent econometric
speciﬁcations, is similar. In our GMM speciﬁcation (1), imposing the conditional median duty is associated
with a roughly 5% increase in Japanese exports of that product to third countries while for our IV speciﬁcation
(7), imposing the conditional median duty is associated with a roughly 7% increase in Japanese exports. There
is, however, a noticeable diﬀerence in the magnitude of the trade depression eﬀect between speciﬁcations (1)
and (7). The depression eﬀect associated with the conditional median in the GMM speciﬁcation (1) of a 19%
fall in Japanese exports to country i is considerably larger than that in the IV speciﬁcation (7) of a 5% fall in
Japanese exports. We believe this diﬀerence in the magnitude of trade depression is likely due to diﬀerences
in the underlying sample of data. In particular, the GMM sample requiring macroeconomic data does not
contain observations for a number of Japan’s sizable export markets, including China, Taiwan, Singapore and
the Philippines. Furthermore, the GMM procedure also requires an additional lag of the trade data, and the
result for those speciﬁcations is to lose all useful trade remedy variation taking place in 1994.
Thus, in speciﬁcation (8) of table 4 we explore the question of whether there are substantial diﬀerences
in trade deﬂection and trade depression eﬀects for Japanese exports across importing country markets. This
speciﬁcation is estimated on the identical sample of data as speciﬁcation (7), but in this column we present
estimates where we interact the antidumping variables of interest with a number of importing country indicators, examining the variation across some of Japan’s important import markets. This approach yields strong
evidence of trade deﬂection, for example, associated with Japan’s exports to both the EU and Korea, which
is quite intuitive, given that they are the two largest destination markets for Japanese exports in the sample
(see again table 1). These two countries are thus likely to be the “next best” alternative markets for Japanese
exports that get shut out of the US because of a US trade policy. Table 5 further illustrates the signiﬁcance of
the economic magnitude of Japan’s trade deﬂection to the EU and Korea, as the median duty imposed against
25 In

all our samples, the mean duty conditional on a duty being imposed is larger than the median. Thus, when we use the

conditional mean duty as a typical duty to quantify the magnitudes of trade deﬂection and trade depression, the results are slightly
(1-5 percentage points) larger.

products which the Japanese export to these countries results in a 16% increase in Japanese exports to the EU
and a 20% increase in Japanese exports to Korea. This is considerably larger than the roughly 7% increase in
Japanese exports to countries other than the EU, Korea, China, and India.
Turning to the country-speciﬁcity of trade depression, speciﬁcation (8) of table 4 indicates that US antidumping duties on third countries are associated with statistically signiﬁcant reduction in Japanese exports
to China and India. Table 5 illustrates the economic signiﬁcance of these results as well, quantifying an 81%
decrease in Japanese exports to India and a 30% decrease in Japanese exports to China associated with the imposition of median US antidumping duties against each of these two countries. We believe that the magnitude
and signiﬁcance of trade depression for India and China in particular may be related to three phenomena. First,
the second column of table 5 indicates that both of these countries face extremely high US antidumping duties:
the median duty imposed against Chinese exports to the US was 121% while the median duty against Indian
exports was 72%. It seems likely that duties facing these two countries are frequently prohibitive, which would
create a severe glut of the aﬀected products in the Chinese and Indian markets which could crowd out imports
from Japan and lead to a sizable amount of trade depression. Second, unlike a number of other importing
countries in the sample, China and India are also frequent targets of US antidumping activity. And ﬁnally,
even beyond a higher frequency of being targeted, both of these countries are also less frequently targeted
alongside Japan in a multi-country US AD investigation over the same product (i.e., relative to the EU and
Korea, whose exporters have a higher frequency of being alongside Japan in a multi-country US AD investigations over the same product). This variation likely allows for more precise estimation of the trade depressing
eﬀect associated with a US antidumping duty being applied on third countries such as China and India alone
(i.e., not simultaneously with Japan). This is also a potential explanation for the imprecisely estimated trade
depressing eﬀects of US antidumping toward the EU and Korea in speciﬁcation (8) of table 4.

4.3

IV estimates for steel versus non-steel products

Another question to consider is whether the AD or SG measures associated with the US steel industry are
particularly important in our results, given that this industry is the most frequent user of US trade remedies.26
To address this issue, in speciﬁcation (9) of table 4 we separate out the estimated policy eﬀects for steel and
non-steel products by interacting each policy variable of interest with an indicator for whether the underlying
6-digit HS product was a steel (HS chapter 72 or 73) or non-steel product. With the exception of the estimates
26 For

example, for the 1992-2001 period, over 50% of US AD investigations that resulted in duties aﬀected steel imports.

for the SG policy (which again are tested on a relatively small number of policy actions), the estimates suggest
that the trade deﬂection and trade depression results may be even stronger for non-steel products than the steel
products that have traditionally been the most active targets of US trade remedy laws. This is particularly
important, given the likelihood that any future growth in use of US trade remedies is likely to come from
non-steel industries as they learn from the steel industry’s experience. Thus this table might suggest that
future use of trade remedies may lead to even more trade deﬂection and trade depression.

4.4

Speciﬁcation tests

As a ﬁnal check on our results, we conduct a speciﬁcation test on our GMM (5) and IV-ﬁxed eﬀects (6) econometric models. These results are reported in Table 6. The thought experiment is similar to that conducted in
the labor literature, beginning with Ashenfelter (1978), on the evaluation of training programs for unemployed
workers. As applied to our context, we investigate whether the imposed policy (i.e., an AD duty) can be used
to predict the change in the dependent variable that occurred before the policy was imposed. Speciﬁcally, we
might be concerned that a Japanese, product-level cost shock in t − 1 led to an increase in Japanese exports to
US and non-US markets in t − 1, and thus that the US policy response in t could actually be used to predict
the t − 1 export growth to importing countries i. If this were the case, our model could be misspeciﬁed and
what we claim to be trade deﬂection might just be an increase in Japanese exports associated with a favorable
cost shock to a particular product exported to many diﬀerent markets.
To investigate this question we therefore regress lagged (t − 1) product-level Japanese export growth to
country i on period t US policy changes and other explanatory variables.27 If the coeﬃcient on a US policy
change in t were positive and statistically signiﬁcant, the imposition of the US AD policy could be interpreted
as a predictor of higher than normal Japanese export growth of that product to all markets in t − 1 which, in
turn, might have been due to a cost shock. Similarly, if US imposition of an AD duty against country i in t
were a statistically signiﬁcant predictor of a reduction in Japanese exports to i in t − 1, this could suggest that
a product-level cost shock in the importing country was behind what we have described as “trade depression.”
Nevertheless, in all speciﬁcations in table 6, the coeﬃcients on changes in US AD policies against Japan and
country i are not statistically signiﬁcant. Our results indicate that the model is correctly speciﬁed and that
the policy variables are measuring the true treatment eﬀect of a policy change.
27 Due

to the small number of changes in safeguards policy, there were insuﬃcient observations to perform this speciﬁcation test

using safeguards.

Conclusion

This paper empirically examines whether a country’s use of an import-restricting trade policy distorts a foreign
country’s exports to third markets. To investigate this question we match data on US use of antidumping and
safeguard trade remedies over the 1992-2001 period to Japanese product-level exports to third countries. We
ﬁnd evidence that US trade remedies both deﬂect and depress Japanese exports. The median antidumping
duty against Japan leads to a 5-7% average increase in Japanese exports to a non-US trading partner. When
the median US antidumping duty is imposed against a third country’s exporters, Japanese exports in the same
product to that third country decrease by an average of 5-19%. Finally, when faced with a US safeguard
measure, Japanese exports to third countries fall by somewhere between 55% and 70%. Our results on the
“deﬂection” and “depression” of Japanese exports vary substantially (and in intuitively appealing ways) across
importing countries, and the estimated impact appears stronger for non-steel relative to steel products.
There are some limitations of our results and approach. First, we have focused on the export response
of only one US trading partner. An open research question is whether US trade policy similarly distorts the
exports of other trading partners, including developing countries. We speculate, for example, that the ability of
developing countries to deﬂect trade may be more limited than that of a country like Japan. Furthermore, we
are less conﬁdent in our results regarding the impact of safeguard policies, as there are relatively few safeguard
observations in our dataset.
Nevertheless, our results have implications for the empirical literature on the impact of trade policy decisions
made by “large” countries, deﬁned as those that are able to aﬀect exporters’ prices. For example, Chang and
Winters (2002) use similarly disaggregated, product level data on unit values and tariﬀs for Brazil and its
trading partners and ﬁnd that the creation of MERCOSUR was accompanied with a substantial decline in
the prices of non-member exporters to the region. While we do not test whether any of the countries in our
analysis are “large” in the sense of their ability to aﬀect the prices of foreign exporters, we provide evidence
that the US’s trade policy decisions do impact the export behavior of a particularly important trading partner.
Finally, we speculate that the results of this paper suggest an additional explanation for the proliferation
of antidumping laws around the world (Miranda et al., 1998; Prusa, 2001) that has not previously been investigated. Much of the prior literature commenting on this proliferation has focused on the retaliation argument:
countries adopt trade remedy laws in order to establish a credible retaliatory threat that will discourage foreign trade remedies targeted against their exporters (Prusa and Skeath, 2002; Blonigen and Bown, 2003). Our
results indicate that the imposition of a US trade remedy can lead to a substantial export surge to a third

country’s market. This third country may therefore face pressure of its own to respond with a trade remedy.
Therefore, US actions may induce trade policy actions by third country importers in addition to (and that is
separate from) retaliation-based trade policy actions. While we do not test here for the formal link between
US trade policy actions and responses by the governments of third countries facing deﬂected trade, our results
that associate substantial export surges with US trade policy changes suggests an additional explanatory factor
that should be an area of future research.

Acknowledgements

We thank Robert Staiger, Bruce Blonigen, Andrew Bernard, Matt Slaughter, Tom Prusa, Steve Redding,
James Tybout, Jeﬀ Campbell, Eric French, Dan Sullivan, Rachel McCulloch, Jay Shambaugh, Matthew Shane
and seminar participants at the 2004 NBER ITI Summer Institute, the 2003 MWIEG Spring meetings, the
2003 NASMES meetings, the 2003 ETSG meetings, the 2003 Federal Reserve SCIEA Fall meetings, Brandeis
University, Dartmouth College, the Federal Reserve Bank of Chicago, University of Maryland and American
University for helpful comments on an early version of the paper. Jaimie Lien provided outstanding research
assistance. Bown acknowledges ﬁnancial support from a Mazer Award and Perlmutter Fellowship at Brandeis
University, as well as the Okun-Model Fellowship at the Brookings Institution. The opinions expressed in this
paper are those of the authors and do not necessarily reﬂect those of the Federal Reserve Bank of Chicago or
the Federal Reserve System. All remaining errors are our own.

Appendix A: Instrument Tests
There are two potential problems with the IV estimator used in estimating (6); bias associated with the
use of a weak instrument and bias associated with correlation in measurement error.
First, in a dynamic panel model, if the autoregressive coeﬃcient on imports is suﬃciently large, then
the lagged level of imports, ln(vmiht−2 ) will be a weak instrument for the lagged diﬀerence, ∆ln(vmiht−1 )
(Blundell and Bond, 1998). In this case, the bias of the IV estimator in a small sample is large (Nelson and
Startz, 1990). To test the quality of two instruments, ln(vmiht−2 ) and ln(vmiht−3 ), the following ﬁrst-stage
model was estimated using each instrument for each of the speciﬁcations presented in table 4.

∆ln(vmiht−1 ) = µh + χit + η1 ∆τht + η2 ∆τiht + η3 ln(vmiht−2 ) + ∆

iht−1 ,

(7)

where ln(vmiht−3 ) was substituted for ln(vmiht−2 ) in some speciﬁcations. As a restricted regression, (7) was
estimated under the assumption that η3 is equal to zero. Table A-1 reports results using the parameters of
speciﬁcation (7) and speciﬁcation (9) in Table 4. Results for all other speciﬁcations are similar. For the model
based on the parameters in speciﬁcation (7), the F-statistic of 27,370 is far larger than the 99% critical χ2 (1)
of 6.63. Likewise, for the model based on the parameters in speciﬁcation (9), the F-statistic of 28,025 is larger
than the 99% critical χ2 (1) of 6.63. In all speciﬁcations, we ﬁnd that ln(vmiht−2 ) and ln(vmiht−3 ) are strong
instruments for ∆ln(vmiht−1 ) and conclude the IV approach is appropriate for our problem.
Second, consider the use of the logged second lag of the level, ln(vmiht−2 ), as an instrument for ∆ln(vmiht−1 ).
If there is measurement error in ln(vmiht ), then measurement error in the regressor, ∆ln(vmiht−1 ), will be
correlated with measurement error in the instrument, ln(vmiht−2 ), and the IV estimator will be biased. An
alternative IV, the third lag of the logged level, ln(vmiht−3 ), has the advantage that its measurement error
will not be correlated with measurement error in the regressor. The disadvantage of this instrument is that
it further shortens an already short panel. Our approach is to estimate (6) using each of these instruments
for every IV speciﬁcation reported in Table 4. By necessity, this requires using the small sample that obtains
when we use the third lag of the level as the instrument. We ﬁnd that the coeﬃcient estimates are robust to
the choice of instrument, suggesting that measurement error in ln(vmiht ) is not a signiﬁcant problem and the
use of ln(vmiht−2 ) as an instrument for ∆ln(vmiht−1 ) is appropriate.

References
[1] Anderson, T.W., Hsiao, C., 1981. Estimation of Dynamic Models with Error Components. Journal of the
American Statistical Association 76, 598-606.
[2] Anderson, T.W., Hsiao, C., 1982. Formulation and Estimation of Dynamic Models Using Panel Data.
Journal of Econometrics 18, 47-82.
[3] Arellano, M., Bond, S., 1991. Some Tests of Speciﬁcation for Panel Data: Monte Carlo Evidence and an
Application to Employment Equations. Review of Economic Studies 58, 277-297.
[4] Ashenfelter, O., 1978. Estimating the Eﬀect of Training Programs on Earnings. Review of Economics and
Statistics 60, 47-57.
[5] Bagwell, K., Staiger, R.W., 1997. Multilateral Tariﬀ Cooperation during the Formation of Free Trade
Areas. International Economic Review 38, 291-319.
[6] Bagwell, K., Staiger, R.W., 1999. An Economic Theory of GATT. American Economic Review 89, 215-248.
[7] Bagwell, K., Staiger, R.W., 2004. Multilateral Trade Negotiations, Bilateral Opportunism and the Rules
of GATT/WTO. Journal of International Economics 63, 1-29.
[8] Blundell, R., Bond, S., 1998. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.
Journal of Econometrics 87, 115-143.
[9] Blonigen, B.A., Prusa, T.J., 2004. Antidumping. in: Harrigan, J., Choi, E.K., (Eds.), Handbook of
International Trade, Vol. I. Blackwell, Oxford, pp. 251-284.
[10] Blonigen, B.A., Bown, C.P., 2003. Antidumping and Retaliation Threats. Journal of International Economics 60, 249-273.
[11] Bown, C.P., Crowley, M.A., 2005. Safeguards. in: Macrory, P., Appleton, A., Plummer, M. (Eds.), The
World Trade Organization, Legal, Economic and Political Analysis, Vol. II. Springer, Dordrecht, pp. 43-66.
[12] Bond, E.W., Syropolous, C., 1996. The Size of Trading Blocs: Market Power and World Welfare Eﬀects.
Journal of International Economics 40, 411-437.
[13] Chang, W., Winters, L.A., 2002. How Regional Blocs Aﬀect Excluded Countries: The Price Eﬀects of
MERCOSUR. American Economic Review 92, 889-904.
24

[14] Clausing, K.A., 2001. Trade Creation and Trade Diversion in the Canada - United States Free Trade
Agreement. Canadian Journal of Economics 34, 677-696.
[15] Ethier, W.J., 2004. Political Externalites, Nondiscrimination, and a Multilateral World. Review of International Economics 12, 303-320.
[16] European Union., 2002. Proposed EU Steel Safeguard Measures. Press release MEMO/02/67, 25 March.
[17] Feenstra, R.C., 2000. World Trade Flows, 1980-1997, with Production and Tariﬀ Data. UC-Davis, mimeo
(and accompanying CD-Rom).
[18] Gallaway, M.P., Blonigen, B.A., Flynn, J.E., 1999. Welfare Costs of the US Antidumping and Countervailing Duty Laws. Journal of International Economics 49, 211-244.
[19] Horn, H., Mavroidis, P., 2001. Economic and Legal Aspects of the Most-Favored Nation Clause. European
Journal of Political Economy 17, 233-279.
[20] International Monetary Fund, 2003. International Financial Statistics, Database and Browser CD-Rom.
[21] Kennedy, P.E., 1981. Estimation with Correctly Interpreted Dummy Variables in Semilogarithmic Equations. American Economic Review 71, 801.
[22] Krishna, P., 2004. The Economics of Preferential Trade Agreements. in: Choi, E.K., Hartigan, J. (Eds.),
Handbook of International Trade, Vol. II: Economic and Legal Analyses of Trade Policy and Institutions,
Blackwell, Oxford.
[23] Levy, P., 1997. A Political-Economic Analysis of Free-Trade Agreements. American Economic Review. 87,
506-519.
[24] McLaren, J.E., 2002. A Theory of Insidious Regionalism. Quarterly Journal of Economics CXVII, 571-608.
[25] Miranda, J., Torres, R.A., Ruiz, M., 1998. The International Use of Antidumping: 1987-1997. Journal of
World Trade 32, 5-71.
[26] Nelson, C.R., Startz, R., 1990. Some Further Results on the Exact Small Sample Properties of the Instrumental Variable Estimator. Econometrica 58, 967-976.
[27] Organization for Economic Cooperation and Development, 2002. OECD Main Economic Indicators.

[28] Prusa, T.J., 1997. The Trade Eﬀects of US Antidumping Actions. in: Feenstra, R.C. (Ed.), The Eﬀects of
US Trade Protection and Promotion Policies. University of Chicago Press, Chicago.
[29] Prusa, T.J., 2001. On the Spread and Impact of Anti-Dumping. Canadian Journal of Economics 34,
591-611.
[30] Prusa, T.J., Skeath, S., 2002. The Economic and Strategic Motives for Antidumping Filings.
Weltwirtschaftliches Archiv 138, 389-413.
[31] Romalis, J., 2002. NAFTA’s and CUSFTA’s Impact on North American Trade. University of Chicago GSB mimeo, July.
[32] Staiger, R.W., Wolak, F.A., 1994. Measuring Industry Speciﬁc Protection: Antidumping in the United
States. Brookings Papers on Economic Activity: Microeconomics, 51-118.
[33] United Nations Industrial Development Organization, 2002. Industrial Statistics Database: 3-digit level
of ISIC Code (Rev. 2) CD-Rom.
[34] United Nations Conference on Trade and Development, various issues. Trade Analysis and Information
System (TRAINS) CD-Rom.
[35] United States Department of Agriculture (Economic Research Service), 2005. International Macroeconomic
Dataset. available on-line at http://www.ers.usda.gov/Data/macroeconomics/.
[36] Viner, J., 1950. The Customs Union Issue. Carnegie Endowment for International Peace, New York.
[37] WTO, 2002. Report (2002) of the Committee on Safeguards to the Council for Trade in Goods. available
on-line at http://www.wto.org/, document number G/L/583, 4 November.
[38] WTO, 2003. Report (2003) of the Committee on Safeguards to the Council for Trade in Goods. available
on-line at http://www.wto.org/, document number G/L/651, 24 October.

Figure 1: Trade Deﬂection and Trade Depression Associated with US Antidumping

115
Mean value of Japanese
product-level exports
110
to third countries...
(indexed so year t-3 =100)

trade deflection

105
100
…where only Japanese
products were hit with US AD

95
90

…where only third-country
products were hit with US AD

85
…where no products were hit
with US AD

80
75
70

trade depression

t-3

t-2

t-1

t
(year of US AD investigation)

Figure 2: Trade Flows under an Antidumping Duty

Firm A

trade destruction
Country A

tariff barrier
trade creation via import
source diversion

Country C
Country B

Firm C
Firm B

trade depression

trade deflection

Trade flow increases relative to free trade
Trade flow decreases relative to free trade

Table 1: Japan’s Major Export Markets - 1996

Country

Rank

TRAINS
Export share

Rank

US
1
26.4%
EU a, b, c
2
13.7
Korea a, b, c
3
7.7
4
7.1
Taiwan c
China c
5
7.1
Hong Kong a, b, c
6
6.4
Thailand a, b, c
7
5.7
Singapore c
8
5.4
9
5.0
Malaysia a, b, c
Canada a, b, c
10
2.5
Indonesia a, b, c
11
2.2
Philippines c
12
2.0
Australia a, b, c
13
2.0
Brazil a, b, c
14
0.8
Mexico a, b, c
15
0.8
New Zealand a, b, c
16
0.6
Saudi Arabia c
17
0.6
Switzerland a, b, c
18
0.5
India a, b, c
19
0.5
Turkey a, b, c
20
0.4
Norway a, b, c
21
0.3
South Africa a, b, c
22
0.3
Israel c
23
0.3
Russia a, b, c
24
0.3
Chile c
25
0.2
Argentina a, b, c
26
0.2
27
0.2
Colombia a, c
Peru a, c
28
0.1
Bangladesh a, b, c
29
0.1
Poland a, b, c
30
0.1
Czech Republic a, b, c
31
0.1
32
<0.1
Egypt a, b, c
Hungary a, b, c
33
<0.1
34
<0.1
Venezuela a, c
Nigeria c
35
<0.1
Panama a, c
36
<0.1
Guatemala c
37
<0.1
Romania a, c
38
<0.1
Notes: † World Trade Data Base source is Feenstra (2000).
a
Japanese trading partner used in estimating specifications (1) and (6)
b
Japanese trading partner used in estimating specifications (2), (3), (4), and (5)
c
Japanese trading partner used in estimating specifications (7), (8), (9), and (10)

1
2
3
4
6
5
8
7
9
13
10
12
11
17
15
22
16
19
20
24
28
21
26
23
29
36
32
41
42
40
39
37
45
44
43
14
76
85

WTDB†
Export Share
27.5%
15.3
7.2
6.3
5.5
6.2
4.6
5.3
3.9
1.4
1.9
1.7
1.8
0.6
0.7
0.4
0.7
0.5
0.5
0.3
0.3
0.5
0.3
0.3
0.2
0.1
0.2
<0.1
<0.1
<0.1
<0.1
0.1
<0.1
<0.1
<0.1
1.3
<0.1
<0.1

Table 2: Data Description and Summary Statistics

Variable

Vector

Predicted
Sign

Description

Mean

Standard
Deviation

Observations

-0.0250

0.9532

254074

Dependent Variable
∆ln(vmiht)

∆ln(vmiht)

Growth of country i's real imports of
product h from Japan†

Explanatory Variables
∆AD Duty jpn,ht

∆τht

US AD duty against Japan on product h

(+)

0.0014

0.0324

254074

∆AD Duty iht

∆τiht

US AD duty against country i on product h

(-)

0.0003

0.0187

254074

∆SG Policy ht

∆τht

US SG policy on product h in year t

(+)

0.0001

0.0095

254074

∆ln(realGDP) it

∆ln(Yit)

Growth of country i's GDP

(+)

0.0273

0.0451

141164

∆ln(realGDP) jpn,t

∆ln(Yt)

Growth of Japan’s GDP

(-)

0.0111

0.0148

141164

Growth of country i's openness to world
[∆ln((Xit+Mit)/GDPit)]

(+)

0.0311

0.0849

141164

∆ln(open) it

∆τ*it

∆ln(yen/curri)t

∆ln(eit)

Growth of bilateral real Japanese
Yen/country i's currency rate*

(+)

0.0011

0.1465

136583

∆ln(avg.wage) jpn,kt

∆ln(ckt)

Growth of Japan’s industry k average wage

(+)

-0.0297

0.1182

113393

∆ln(v.add/worker) jpn,kt

∆ln(ckt)

Growth of Japan’s industry k average
value-added per worker

(+)

-0.0247

0.1419

113393

Notes: Subscript i is an importing country, h is a 6-digit HS product, k is a 3-digit ISIC industry, and t is a year. † The trade-weighted mean growth of
country i's real, product-level imports from Japan is 0.1743 with a trade-weighted standard deviation of 0.6222 in the sample of 254074 observations,
and 0.1684 with a trade-weighted standard deviation of 0.6174 in the sample of 141164 observations. *For Norway and Switzerland, this is the
growth of the real exchange rate index from the IFS.

Table 3: Arellano-Bond GMM Estimation: Results for Japanese Exports, 1995-2001

Dependent Variable: Δln(vmiht)
Baseline
specification

Add currency
depreciation

Add industry
controls

Substitute
growth in
value-added per
worker

(1)

(2)

(3)

(4)

(5)

(6)

0.141*
(0.086)

0.148*
(0.086)

0.156*
(0.091)

0.168*
(0.091)

0.167*
(0.090)

0.127
(0.081)

ΔAD Duty iht

-1.269***
(0.411)

-1.207***
(0.408)

-1.292***
(0.418)

-1.262***
(0.417)

-1.265***
(0.415)

-0.870***
(0.243)

ΔSG Policy ht

-1.012*
(0.536)

-0.981*
(0.530)

-1.023
(2.111)

-1.102
(2.105)

-1.018
(2.099)

-1.189***
(0.275)

IV for Δln(vm) iht-1

0.276***
(0.012)

0.245***
(0.012)

0.328***
(0.012)

0.319***
(0.012)

0.306***
(0.012)

0.323***
(0.009)

IV for Δln(vm) iht-2

0.008
(0.006)

-0.003
(0.006)

0.024***
(0.006)

0.021***
(0.006)

0.017***
(0.006)

Δln(realGDP) it

3.428***
(0.072)

3.315***
(0.075)

3.014***
(0.083)

3.062***
(0.082)

3.439***
(0.079)

Δln(realGDP) jpn,t

-2.434***
(0.185)

-3.238***
(0.198)

-2.640***
(0.221)

-3.239***
(0.217)

-3.501***
(0.217)

Δln(open) it

0.391***
(0.045)

0.741***
(0.047)

0.813***
(0.051)

0.790***
(0.051)

0.638***
(0.051)

Δln(yen/curri)t

0.265***
(0.025)

0.437***
(0.030)

0.416***
(0.029)

0.268***
(0.027)

Δln(avg.wage) jpn,kt

0.479***
(0.032)

Δln(v.add/worker) jpn,kt

0.362***
(0.025)

-0.117***
(0.003)

-0.120***
(0.003)

-0.103***
(0.004)

-0.122***
(0.004)

6-digit HS product fixed effects

Yes

Country i-year dummies

Yes

141164

136583

113393

141164

-1.36

-0.73

-0.99

-0.95

-0.99

0.081

Explanatory Variables

Remove
industry
controls

IV/ fixed effects
specification
instead of GMM

Policy Variables
ΔAD Duty jpn,ht

Other Control Variables

Constant

Observations
Average autocovariance in
residuals of order 2 [z-stat]
R2

Notes: Subscript i is an importing country, h is a 6-digit HS product, k is a 3-digit ISIC industry, and t is a year. In parentheses are White’s heteroskedasticityconsistent standard errors corrected for clusters defined on the variable defined as the 6-digit HS product and year combination. ***, **, and * denote
variables statistically significant at the 1, 5, and 10 percent levels, respectively.

Table 4: IV Estimation Using Fixed Eﬀects: Results for Japanese Exports, 1994-2001

Dependent Variable: ∆ln(vmiht)
IV
by country

IV
by commodity

Full sample
OLS specification

(7)

Explanatory Variables

Full sample
IV specification

(8)

(9)

(10)

0.190***
(0.061)

0.105*
(0.059)

AD Duty Changes Against Japan
∆AD Duty jpn,ht
∆AD Duty jpn,ht for exports sent to EU
∆AD Duty jpn,ht for exports sent to Korea

0.390*
(0.222)
0.578**
(0.242)
-0.326
(0.203)
0.238
(0.454)
0.198***
(0.070)

---

∆AD Duty jpn,ht for exports sent to China

∆AD Duty jpn,ht for exports sent to India

∆AD Duty pn,ht for exports sent to Other

∆AD Duty jpn,ht for exports of Steel

∆AD Duty jpn,ht for exports of Non-steel

0.157
(0.100)
0.204***
(0.079)

---

AD Duty Changes Against Country i
∆AD Duty iht
∆AD Duty i,ht for exports sent to EU

-0.281***
(0.101)

--0.812
(0.915)
-1.000
(1.830)
-0.249**
(0.120)
-1.118***
(0.308)
0.145
(0.243)

∆AD Duty i,ht for exports sent to Korea

∆AD Duty i,ht for exports sent to China

∆AD Duty i,ht for exports sent to India

-0.242**
(0.096)

∆AD Duty pn,ht for exports sent to Other

∆AD Duty i,ht for exports of Steel

∆AD Duty i,ht for exports of Non-steel

-0.807***
(0.208)

-0.805***
(0.208)

∆SG Policy ht for exports of Steel

∆SG Policy ht for exports of Non-steel

0.306***
(0.006)

0.339***
(0.035)

0.338***
(0.035)

0.339***
(0.035)

-0.308***
(0.002)
-0.682
(0.923)

Yes

--0.077
(0.160)
-0.413***
(0.130)

----

SG Policy Changes
∆SG Policy ht

--0.868***
(0.240)
-0.621
(0.415)

-0.547***
(0.198)
---

Other Control Variables
IV for ∆ln(vm) iht-1
∆ln(vm) iht-1
Constant
6-digit HS product fixed effects
Country i-year dummies

R2
F-statistic

Yes

254074

0.065

Observations

Notes:

0.065

0.147

27370.6

27497.1

28025.0

Subscript i is an importing country, h is a 6-digit HS product, and t is a year. In parentheses are White’s heteroskedasticity-consistent
standard errors corrected for clusters defined on the variable defined as the 6-digit HS product and year combination. ***, **, and * denote
variables statistically significant at the 1, 5, and 10 percent levels, respectively. F-statistic is for the weak instrument test described in
appendix A.

Table 5: Magnitude of Trade Deﬂection and Trade Depression Estimates

Percent Increase in Affected Japanese Product-Level Exports
Associated with Subsample’s Median Duty,
Conditional on a Duty Being Imposed
Subsample’s
Median
Conditional Duty
[141164 obs.]

Subsample’s
Median
Conditional Duty
[254074 obs.]

GMM
Specification

IV
Specification

(1)

(7)

(8)

∆AD Duty jpn,ht

37.13

5.24*

∆AD Duty jpn,ht

37.13

7.05***

∆AD Duty jpn,ht for exports sent to EU

41.72

16.27*

∆AD Duty jpn,ht for exports sent to Korea

34.83

20.13**

∆AD Duty jpn,ht for exports sent to China

37.13

-12.10

∆AD Duty jpn,ht for exports sent to India

32.52

7.74

∆AD Duty jpn,ht for exports sent to Other

37.13

7.35***

∆AD Duty iht

14.84

-18.83***

∆AD Duty iht

16.26

-4.57***

∆AD Duty iht for exports sent to EU

10.43

-8.45

∆AD Duty iht for exports sent to Korea

3.08

-3.08

∆AD Duty iht for exports sent to China

121.46

-30.24**

∆AD Duty iht for exports sent to India

72.49

-81.04***

∆AD Duty iht for exports sent to Other

20.28

2.94

Trade Deflection

Trade Depression

Notes:

Subscript i is an importing country, h is a 6-digit HS product, and t is a year. ***, **, and * denotes underlying coefficient estimate was statistically
significant at the 1, 5, and 10 percent levels, respectively. Specification (1) uses the coefficient estimates from table 3 and specifications (7) and (8) use
the coefficient estimates from table 4.

Table 6: Speciﬁcation Test: Do US Trade Remedies Predict Japanese Export Growth to Third Countries
Before They’re Imposed?

Dependent Variable: Δln(vmiht-1)
[lagged Japanese export growth of product h to importer i]
Test of specification

Test of specification

(1)

(2)

(7)

-0.124
(0.109)

-0.152
(0.108)

-0.053
(0.072)

0.391
(0.349)

0.396
(0.345)

0.148
(0.144)

-0.016
(0.118)

-0.006
(0.117)

-0.099*
(0.055)

0.277***
(0.012)
0.009
(0.006)
3.425***
(0.072)
-2.412***
(0.185)
0.395**
(0.045)

0.306***
(0.006)

-0.117***
(0.003)

0.245***
(0.012)
-0.002
(0.006)
3.310***
(0.075)
-3.222***
(0.198)
0.747***
(0.047)
0.267***
(0.045)
-0.120***
(0.003)

0.339***
(0.035)

6-digit HS product fixed effects

Yes

Country i-year dummies

Yes

141162

136581

254070

-1.35

-0.72

0.065

Explanatory Variables
AD Duty Changes Against Japan
ΔAD Duty jpn,ht
AD Duty Changes Against Country i
ΔAD Duty iht
SG Policy Changes
ΔSG Policy ht
Other Control Variables
IV for Δln(vm) iht-2
Δln(vm) iht-3
Δln(realGDP) it-1
Δln(realGDP) jpn,t-1
Δln(open) it-1
Δln(yen/curri)t-1
Constant

Observations
Average autocovariance in residuals
of order 2 [z-stat]
R

------

Notes: subscript i is an importing country, h is a 6-digit HS product, and t is a year. In parentheses are White’s
heteroskedasticity-consistent standard errors corrected for clusters defined on the variable defined as the 6-digit HS
product and year combination. ***, **, and * denote variables statistically significant at the 1, 5, and 10 percent levels,
respectively.

Table A-1: Testing Instrument Quality

Dependent Variable: Δln(vmiht)
First-stage
restricted
regression based
on specification

First-stage
unrestricted
regression based
on specification

First-stage
restricted
regression based
on specification

(7)

Explanatory Variables

First-stage
unrestricted
regression based
on specification

(7)

(9)

-0.131**
(0.060)

-0.137**
(0.063)

AD Duty Changes Against Japan
ΔAD Duty jpn,ht
ΔAD Duty jpn,ht for exports of Steel

ΔAD Duty jpn,ht for exports of Non-steel

0.034
(0.097)

0.069
(0.102)

-0.053
(0.096)
-0.175**
(0.076)

-0.043
(0.101)
-0.189**
(0.080)

AD Duty Changes Against Country i
ΔAD Duty iht

-0.151
(0.154)
0.150
(0.125)

-0.198
(0.162)
0.237*
(0.131)

0.225
(0.231)
0.909**
(0.400)

0.211
(0.243)
1.065**
(0.421)

ΔAD Duty i,ht for exports of Steel

ΔAD Duty i,ht for exports of Non-steel

0.397**
(0.200)

0.423**
(0.211)

ΔSG Policy ht for exports of Steel

ΔSG Policy ht for exports of Non-steel

-0.210***
(0.001)

6-digit HS product fixed effects

Yes

Country i-year dummies

Yes

254598

0.152

0.059

0.154

0.059

SG Policy Changes
ΔSG Policy ht

Other Control Variables
Δln(vm) iht-1

Observations
R2

Notes: Subscript i is an importing country, h is a 6-digit HS product, and t is a year. In parentheses are White’s
heteroskedasticity-consistent standard errors corrected for clusters defined on the variable defined as the 6-digit HS
product and year combination. ***, **, and * denote variables statistically significant at the 1, 5, and 10 percent levels,
respectively.

Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
Dynamic Monetary Equilibrium in a Random-Matching Economy
Edward J. Green and Ruilin Zhou

WP-00-1

The Effects of Health, Wealth, and Wages on Labor Supply and Retirement Behavior
Eric French

WP-00-2

Market Discipline in the Governance of U.S. Bank Holding Companies:
Monitoring vs. Influencing
Robert R. Bliss and Mark J. Flannery

WP-00-3

Using Market Valuation to Assess the Importance and Efficiency
of Public School Spending
Lisa Barrow and Cecilia Elena Rouse
Employment Flows, Capital Mobility, and Policy Analysis
Marcelo Veracierto
Does the Community Reinvestment Act Influence Lending? An Analysis
of Changes in Bank Low-Income Mortgage Activity
Drew Dahl, Douglas D. Evanoff and Michael F. Spivey

WP-00-4

WP-00-5

WP-00-6

Subordinated Debt and Bank Capital Reform
Douglas D. Evanoff and Larry D. Wall

WP-00-7

The Labor Supply Response To (Mismeasured But) Predictable Wage Changes
Eric French

WP-00-8

For How Long Are Newly Chartered Banks Financially Fragile?
Robert DeYoung

WP-00-9

Bank Capital Regulation With and Without State-Contingent Penalties
David A. Marshall and Edward S. Prescott

WP-00-10

Why Is Productivity Procyclical? Why Do We Care?
Susanto Basu and John Fernald

WP-00-11

Oligopoly Banking and Capital Accumulation
Nicola Cetorelli and Pietro F. Peretto

WP-00-12

Puzzles in the Chinese Stock Market
John Fernald and John H. Rogers

WP-00-13

The Effects of Geographic Expansion on Bank Efficiency
Allen N. Berger and Robert DeYoung

WP-00-14

Idiosyncratic Risk and Aggregate Employment Dynamics
Jeffrey R. Campbell and Jonas D.M. Fisher

WP-00-15

Working Paper Series (continued)
Post-Resolution Treatment of Depositors at Failed Banks: Implications for the Severity
of Banking Crises, Systemic Risk, and Too-Big-To-Fail
George G. Kaufman and Steven A. Seelig

WP-00-16

The Double Play: Simultaneous Speculative Attacks on Currency and Equity Markets
Sujit Chakravorti and Subir Lall

WP-00-17

Capital Requirements and Competition in the Banking Industry
Peter J.G. Vlaar

WP-00-18

Financial-Intermediation Regime and Efficiency in a Boyd-Prescott Economy
Yeong-Yuh Chiang and Edward J. Green

WP-00-19

How Do Retail Prices React to Minimum Wage Increases?
James M. MacDonald and Daniel Aaronson

WP-00-20

Financial Signal Processing: A Self Calibrating Model
Robert J. Elliott, William C. Hunter and Barbara M. Jamieson

WP-00-21

An Empirical Examination of the Price-Dividend Relation with Dividend Management
Lucy F. Ackert and William C. Hunter

WP-00-22

Savings of Young Parents
Annamaria Lusardi, Ricardo Cossa, and Erin L. Krupka

WP-00-23

The Pitfalls in Inferring Risk from Financial Market Data
Robert R. Bliss

WP-00-24

What Can Account for Fluctuations in the Terms of Trade?
Marianne Baxter and Michael A. Kouparitsas

WP-00-25

Data Revisions and the Identification of Monetary Policy Shocks
Dean Croushore and Charles L. Evans

WP-00-26

Recent Evidence on the Relationship Between Unemployment and Wage Growth
Daniel Aaronson and Daniel Sullivan

WP-00-27

Supplier Relationships and Small Business Use of Trade Credit
Daniel Aaronson, Raphael Bostic, Paul Huck and Robert Townsend

WP-00-28

What are the Short-Run Effects of Increasing Labor Market Flexibility?
Marcelo Veracierto

WP-00-29

Equilibrium Lending Mechanism and Aggregate Activity
Cheng Wang and Ruilin Zhou

WP-00-30

Impact of Independent Directors and the Regulatory Environment on Bank Merger Prices:
Evidence from Takeover Activity in the 1990s
Elijah Brewer III, William E. Jackson III, and Julapa A. Jagtiani
Does Bank Concentration Lead to Concentration in Industrial Sectors?
Nicola Cetorelli

WP-00-31

WP-01-01

Working Paper Series (continued)
On the Fiscal Implications of Twin Crises
Craig Burnside, Martin Eichenbaum and Sergio Rebelo

WP-01-02

Sub-Debt Yield Spreads as Bank Risk Measures
Douglas D. Evanoff and Larry D. Wall

WP-01-03

Productivity Growth in the 1990s: Technology, Utilization, or Adjustment?
Susanto Basu, John G. Fernald and Matthew D. Shapiro

WP-01-04

Do Regulators Search for the Quiet Life? The Relationship Between Regulators and
The Regulated in Banking
Richard J. Rosen
Learning-by-Doing, Scale Efficiencies, and Financial Performance at Internet-Only Banks
Robert DeYoung
The Role of Real Wages, Productivity, and Fiscal Policy in Germany’s
Great Depression 1928-37
Jonas D. M. Fisher and Andreas Hornstein

WP-01-05

WP-01-06

WP-01-07

Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy
Lawrence J. Christiano, Martin Eichenbaum and Charles L. Evans

WP-01-08

Outsourcing Business Service and the Scope of Local Markets
Yukako Ono

WP-01-09

The Effect of Market Size Structure on Competition: The Case of Small Business Lending
Allen N. Berger, Richard J. Rosen and Gregory F. Udell

WP-01-10

Deregulation, the Internet, and the Competitive Viability of Large Banks
and Community Banks
Robert DeYoung and William C. Hunter

WP-01-11

Price Ceilings as Focal Points for Tacit Collusion: Evidence from Credit Cards
Christopher R. Knittel and Victor Stango

WP-01-12

Gaps and Triangles
Bernardino Adão, Isabel Correia and Pedro Teles

WP-01-13

A Real Explanation for Heterogeneous Investment Dynamics
Jonas D.M. Fisher

WP-01-14

Recovering Risk Aversion from Options
Robert R. Bliss and Nikolaos Panigirtzoglou

WP-01-15

Economic Determinants of the Nominal Treasury Yield Curve
Charles L. Evans and David Marshall

WP-01-16

Price Level Uniformity in a Random Matching Model with Perfectly Patient Traders
Edward J. Green and Ruilin Zhou

WP-01-17

Earnings Mobility in the US: A New Look at Intergenerational Inequality
Bhashkar Mazumder

WP-01-18

Working Paper Series (continued)
The Effects of Health Insurance and Self-Insurance on Retirement Behavior
Eric French and John Bailey Jones

WP-01-19

The Effect of Part-Time Work on Wages: Evidence from the Social Security Rules
Daniel Aaronson and Eric French

WP-01-20

Antidumping Policy Under Imperfect Competition
Meredith A. Crowley

WP-01-21

Is the United States an Optimum Currency Area?
An Empirical Analysis of Regional Business Cycles
Michael A. Kouparitsas

WP-01-22

A Note on the Estimation of Linear Regression Models with Heteroskedastic
Measurement Errors
Daniel G. Sullivan

WP-01-23

The Mis-Measurement of Permanent Earnings: New Evidence from Social
Security Earnings Data
Bhashkar Mazumder

WP-01-24

Pricing IPOs of Mutual Thrift Conversions: The Joint Effect of Regulation
and Market Discipline
Elijah Brewer III, Douglas D. Evanoff and Jacky So

WP-01-25

Opportunity Cost and Prudentiality: An Analysis of Collateral Decisions in
Bilateral and Multilateral Settings
Herbert L. Baer, Virginia G. France and James T. Moser

WP-01-26

Outsourcing Business Services and the Role of Central Administrative Offices
Yukako Ono

WP-02-01

Strategic Responses to Regulatory Threat in the Credit Card Market*
Victor Stango

WP-02-02

The Optimal Mix of Taxes on Money, Consumption and Income
Fiorella De Fiore and Pedro Teles

WP-02-03

Expectation Traps and Monetary Policy
Stefania Albanesi, V. V. Chari and Lawrence J. Christiano

WP-02-04

Monetary Policy in a Financial Crisis
Lawrence J. Christiano, Christopher Gust and Jorge Roldos

WP-02-05

Regulatory Incentives and Consolidation: The Case of Commercial Bank Mergers
and the Community Reinvestment Act
Raphael Bostic, Hamid Mehran, Anna Paulson and Marc Saidenberg
Technological Progress and the Geographic Expansion of the Banking Industry
Allen N. Berger and Robert DeYoung

WP-02-06

WP-02-07

Working Paper Series (continued)
Choosing the Right Parents: Changes in the Intergenerational Transmission
of Inequality  Between 1980 and the Early 1990s
David I. Levine and Bhashkar Mazumder

WP-02-08

The Immediacy Implications of Exchange Organization
James T. Moser

WP-02-09

Maternal Employment and Overweight Children
Patricia M. Anderson, Kristin F. Butcher and Phillip B. Levine

WP-02-10

The Costs and Benefits of Moral Suasion: Evidence from the Rescue of
Long-Term Capital Management
Craig Furfine

WP-02-11

On the Cyclical Behavior of Employment, Unemployment and Labor Force Participation
Marcelo Veracierto

WP-02-12

Do Safeguard Tariffs and Antidumping Duties Open or Close Technology Gaps?
Meredith A. Crowley

WP-02-13

Technology Shocks Matter
Jonas D. M. Fisher

WP-02-14

Money as a Mechanism in a Bewley Economy
Edward J. Green and Ruilin Zhou

WP-02-15

Optimal Fiscal and Monetary Policy: Equivalence Results
Isabel Correia, Juan Pablo Nicolini and Pedro Teles

WP-02-16

Real Exchange Rate Fluctuations and the Dynamics of Retail Trade Industries
on the U.S.-Canada Border
Jeffrey R. Campbell and Beverly Lapham

WP-02-17

Bank Procyclicality, Credit Crunches, and Asymmetric Monetary Policy Effects:
A Unifying Model
Robert R. Bliss and George G. Kaufman

WP-02-18

Location of Headquarter Growth During the 90s
Thomas H. Klier

WP-02-19

The Value of Banking Relationships During a Financial Crisis:
Evidence from Failures of Japanese Banks
Elijah Brewer III, Hesna Genay, William Curt Hunter and George G. Kaufman

WP-02-20

On the Distribution and Dynamics of Health Costs
Eric French and John Bailey Jones

WP-02-21

The Effects of Progressive Taxation on Labor Supply when Hours and Wages are
Jointly Determined
Daniel Aaronson and Eric French

WP-02-22

Working Paper Series (continued)
Inter-industry Contagion and the Competitive Effects of Financial Distress Announcements:
Evidence from Commercial Banks and Life Insurance Companies
Elijah Brewer III and William E. Jackson III

WP-02-23

State-Contingent Bank Regulation With Unobserved Action and
Unobserved Characteristics
David A. Marshall and Edward Simpson Prescott

WP-02-24

Local Market Consolidation and Bank Productive Efficiency
Douglas D. Evanoff and Evren Örs

WP-02-25

Life-Cycle Dynamics in Industrial Sectors. The Role of Banking Market Structure
Nicola Cetorelli

WP-02-26

Private School Location and Neighborhood Characteristics
Lisa Barrow

WP-02-27

Teachers and Student Achievement in the Chicago Public High Schools
Daniel Aaronson, Lisa Barrow and William Sander

WP-02-28

The Crime of 1873: Back to the Scene
François R. Velde

WP-02-29

Trade Structure, Industrial Structure, and International Business Cycles
Marianne Baxter and Michael A. Kouparitsas

WP-02-30

Estimating the Returns to Community College Schooling for Displaced Workers
Louis Jacobson, Robert LaLonde and Daniel G. Sullivan

WP-02-31

A Proposal for Efficiently Resolving Out-of-the-Money Swap Positions
at Large Insolvent Banks
George G. Kaufman

WP-03-01

Depositor Liquidity and Loss-Sharing in Bank Failure Resolutions
George G. Kaufman

WP-03-02

Subordinated Debt and Prompt Corrective Regulatory Action
Douglas D. Evanoff and Larry D. Wall

WP-03-03

When is Inter-Transaction Time Informative?
Craig Furfine

WP-03-04

Tenure Choice with Location Selection: The Case of Hispanic Neighborhoods
in Chicago
Maude Toussaint-Comeau and Sherrie L.W. Rhine

WP-03-05

Distinguishing Limited Commitment from Moral Hazard in Models of
Growth with Inequality*
Anna L. Paulson and Robert Townsend

WP-03-06

Resolving Large Complex Financial Organizations
Robert R. Bliss

WP-03-07

Working Paper Series (continued)
The Case of the Missing Productivity Growth:
Or, Does information technology explain why productivity accelerated in the United States
but not the United Kingdom?
Susanto Basu, John G. Fernald, Nicholas Oulton and Sylaja Srinivasan

WP-03-08

Inside-Outside Money Competition
Ramon Marimon, Juan Pablo Nicolini and Pedro Teles

WP-03-09

The Importance of Check-Cashing Businesses to the Unbanked: Racial/Ethnic Differences
William H. Greene, Sherrie L.W. Rhine and Maude Toussaint-Comeau

WP-03-10

A Structural Empirical Model of Firm Growth, Learning, and Survival
Jaap H. Abbring and Jeffrey R. Campbell

WP-03-11

Market Size Matters
Jeffrey R. Campbell and Hugo A. Hopenhayn

WP-03-12

The Cost of Business Cycles under Endogenous Growth
Gadi Barlevy

WP-03-13

The Past, Present, and Probable Future for Community Banks
Robert DeYoung, William C. Hunter and Gregory F. Udell

WP-03-14

Measuring Productivity Growth in Asia: Do Market Imperfections Matter?
John Fernald and Brent Neiman

WP-03-15

Revised Estimates of Intergenerational Income Mobility in the United States
Bhashkar Mazumder

WP-03-16

Product Market Evidence on the Employment Effects of the Minimum Wage
Daniel Aaronson and Eric French

WP-03-17

Estimating Models of On-the-Job Search using Record Statistics
Gadi Barlevy

WP-03-18

Banking Market Conditions and Deposit Interest Rates
Richard J. Rosen

WP-03-19

Creating a National State Rainy Day Fund: A Modest Proposal to Improve Future
State Fiscal Performance
Richard Mattoon

WP-03-20

Managerial Incentive and Financial Contagion
Sujit Chakravorti, Anna Llyina and Subir Lall

WP-03-21

Women and the Phillips Curve: Do Women’s and Men’s Labor Market Outcomes
Differentially Affect Real Wage Growth and Inflation?
Katharine Anderson, Lisa Barrow and Kristin F. Butcher

WP-03-22

Evaluating the Calvo Model of Sticky Prices
Martin Eichenbaum and Jonas D.M. Fisher

WP-03-23

Working Paper Series (continued)
The Growing Importance of Family and Community: An Analysis of Changes in the
Sibling Correlation in Earnings
Bhashkar Mazumder and David I. Levine

WP-03-24

Should We Teach Old Dogs New Tricks? The Impact of Community College Retraining
on Older Displaced Workers
Louis Jacobson, Robert J. LaLonde and Daniel Sullivan

WP-03-25

Trade Deflection and Trade Depression
Chad P. Bown and Meredith A. Crowley

WP-03-26

Full text of Working Papers (Federal Reserve Bank of Chicago) : Trade Deflection and Trade Depression, Working Paper 2003-26

FRASER