Does Tariff Liberalization Kick the Good Apples Out?

Theory and Evidence

Alan C. Spearot∗

University of California - Santa Cruz

February 24, 2011

Abstract

This paper examines the liberalization of a common tariff when imported varieties vary

in quality and cost. Varieties of higher quality and/or lower cost (a) are imported at lower

absolute demand elasticities and (b) earn higher revenues. By virtue of larger demand elas-

ticities, low revenue varieties benefit the most from tariff liberalization. Further, if varieties

are substitutable, low revenue varieties may benefit at the expense of high revenue varieties.

These predictions are confirmed using a case study of US Uruguay Round tariff cuts, where

within narrowly defined products, low revenue exporters experienced large gains, and high rev-

enue exporters experienced negligible gains. Further, I find evidence suggesting that product

quality played a role in governing the effects of tariffs on trade flows.

JEL Classifications: F12, F13, F14

Keywords: Firm Heterogeneity, Trade Liberalization, Quality, Non-discrimination

∗Email: acspearot@gmail.com. Address: Economics Department, 1156 High Street, Santa Cruz, CA, 95064.I thank Pol Antras, David Hummels, Wolfgang Keller, Dirk Krueger, Volodymyr Lugovskyy, Phillip McCalman,Peter Morrow, Thanh Xuan Nguyen, Anson Soderbery, Robert Staiger and seminar participants at the 2009 OtagoWorkshop on International Trade, the London School of Economics, the 2009 Asia Pacific Economics AssociationAnnual Conference, the 2009 Empirical Investigations in Trade and Geography conference, the University of Colorado,the Stanford Institute for Theoretical Economics, and the University of Michigan for very helpful comments. I alsothank James Allen from the New Zealand Ministry of Foreign Affairs and Trade for HS8 trade data used for anearlier working paper. All remaining errors are my own.

1

1 Introduction

If one point has been made abundantly clear by the theory of international trade, it is that trade

is often driven by heterogeneity. This heterogeneity can result from classical Ricardian differences

in relative autarky prices, to the modern treatment of heterogeneity at the firm level. Regardless

of the source, heterogeneity almost always reflects in prices and/or market shares. And, as many

authors have documented (Schott 2004, in particular), trade data exhibits substantial variation in

unit values and trade volumes within precise product categories.1

However, despite the massive amount of heterogeneity in varieties across trading partners within

narrowly defined products, the rules of the GATT/WTO, on a basic level, seem designed for a more

homogenous environment. With regard to tariffs, this point is particularly salient, as outside of spe-

cial safeguards, retaliatory measures, and all-or-nothing regional agreements, members have very

little latitude regarding tariffs applied to different varieties of the same product. One particu-

lar guiding principle, “non-discrimination”, imposes that all GATT/WTO members receive equal

treatment, usually via a common, or “most favored nation” (MFN), tariff. This applies within any

product, across all export sources without preferential status, and (obviously) does not discrimi-

nate by quality or other characteristics. Further, this equal treatment rule extends to the process

of liberalization, and allows for no consideration of which exporters stand to gain more or less from

lower tariffs.

Overall, there is a clear friction between the precise intent of WTO rules regarding tariffs, and

the natural differentiation which occurs in trade flows. Critically, even though the WTO prefers

(and arguably promotes) the multilateral process over other preferential or regional schemes, it

remains unclear how the effects of MFN liberalization accrue within products that exhibit significant

within-product differentiation. In what way should MFN tariff reductions influence bilateral trade

flows? Are certain countries more likely to gain from MFN liberalization based on the fundamental

characteristics of the products they sell? Overall, how are the benefits of additional import market

access distributed across competing and differentiated exporters?

This paper answers these questions. Using a simple theoretical model based on Melitz and

Ottaviano (2008), I show that the liberalization of a common ad-valorem tariff need not increase

bilateral imports of all varieties. In particular, if varieties are differentiated by quality and/or

1As an example, consider the import of “Men’s or Boys’ Shirts, of Cotton” (HS6 code 620520) by New Zealand in1999. Overall, New Zealand imported varieties within this product category from 50 different countries. While theaverage pre-tariff free-on-board unit value was $20 exporter-specific unit values varied substantially. On one end ofthe spectrum, Belgium exported varieties at an average of $75 per unit. On the other end, Indonesia exported at anaverage of $3.60 per unit. Clearly, imports from Belgium and imports from Indonesia are in fact very different. Yet,the WTO mandates that these products are treated equally in setting tariffs, and when liberalizing tariffs.

2

production cost, I show that liberalization of a common tariff may increase imports of low revenue

varieties at the expense of high revenue varieties. In analyzing the effects of US Uruguay Round

tariff cuts, I find robust support for the model. Specifically, I find that the benefits of import

tariff liberalization are large and statistically significant for low revenue varieties and smaller and

statistically insignificant for high revenue varieties. Further, I find that this relationship breaks

down when characterizing varieties by their relative price within products, which as detailed below,

suggests that quality plays a role in governing the effects of tariffs on trade flows.

The key to the model is the degree to which demand elasticities vary across varieties. Specifically,

varieties will be sold at a lower (absolute) demand elasticity if they are of higher quality and/or

produced at a lower cost. In both cases, varieties earn larger revenues. Since demand is fairly

inelastic for these varieties, changes to tariffs have a fairly small direct effect on the value of bilateral

imports. This is in stark contrast with varieties of low quality and/or produced at a high cost. In

these cases, revenues are fairly small, and equilibrium demand elasticities are relatively large. Thus,

for these varieties, changes to tariffs have a relatively large effect on the value of bilateral imports.

In equilibrium, the effects of changes to a common tariff are aggregated across all varieties, and

competition is generally tougher when tariffs fall. That is, when tariffs fall, so does the residual

demand for each variety, all else equal. It is the resolution of the tension between this aggregate

effect and the effects related to demand elasticities which determine the varieties that benefit from

tariff liberalization and to what extent. In particular, I show that the aggregate effect may in

fact be larger than the direct effect for some varieties. As detailed above, these are the varieties

that earn relatively large revenues before tariffs are cut. Thus, as a novel result, I show that tariff

liberalization may in fact decrease imports of varieties that earn large revenues before the tariff

cut. In contrast, I show that varieties earning low revenues always benefit from tariff liberalization

by virtue of the high demand elasticity at which these varieties are sold. Overall, I show that the

traditional negative effects of tariffs are amplified for low revenue varieties, and smaller and/or of

opposite sign for high revenue varieties.

Empirically, the model is evaluated using a case study of tariff reductions by the United States

resulting from Uruguay Round GATT negotiations. Despite being a result of bargaining within

the GATT, this case study is sensible on a number of levels. For one, data are available at the

HS10 level, which provides a useful amount of detail within narrowly defined products.2 Second,

reductions to MFN tariffs were relatively quick, most of which occurring over the period 1995-

1999. Third, the evidence suggests that import growth rates before the Uruguay Round agreements

2For example, “Grand Pianos” is an HS8 product, and HS10 provides detail such as “Containing a case measuringless than 152.40 cm in length”.

3

were drafted were independent from whether or not an HS8 product received a tariff cut after the

Uruguay Round was completed. Finally, the US HS8 tariff data provides information on whether

non-tariff barriers were present. As these policy instruments are outside the scope of the model,

such information is crucial, and not available in other datasets such as TRAINS.

Using a number of different specifications, I find evidence which is broadly supportive of the

model. Regressing bilateral import growth rates on tariffs and an interaction with pre-Uruguay

Round import values, I find that within-products, bilateral import elasticities (with respect to

tariffs) were large, significant, and negative for the export sources with the smallest pre-Uruguay

Round import values. In contrast, estimated elasticities are small and, with few exceptions, insignif-

icantly different from zero for export sources with the largest pre-Uruguay Round import values.

Further, I find that this result is sharpened when focusing on products with below-median levels of

exporter concentration, and in products with a relatively large number of varieties. These results

are also fairly robust to a large set of supplemental regressions which control for an exporter-specific

response to US tariffs and potential outlier exporters.

Finally, to address a possible role for quality rather than only cost-heterogeneity, I expand on

the main theoretical result by showing that if quality plays a role, variety-specific revenue is the

only measure that captures the characteristics relevant for assessing the effects of tariffs. That is,

without precise assumptions over the relationship between quality and costs, there will not exist a

monotone relationship between other measures (prices and quantities) and demand elasticity when

quality is present. Empirically, this is confirmed in the data, where I show that ordering varieties

by their relative price within products rather than revenues results in a sizeable drop in power in

predicting the effects of tariff cuts. Further, I find that low price varieties benefited the most from

tariff cuts, suggesting that many low price varieties are also low quality varieties. Hence, the data

suggests that quality is playing a role in governing the effects of tariffs on bilateral trade flows.

Related literature

This paper adds to a number of different areas related to trade, firm heterogeneity, and trade policy.

Most notably, it is related to the burgeoning literature on the role quality and cost heterogeneity play

in trade flows. The results are reminiscent of Hummels and Skiba (2004), who evaluate the effect

of trade costs and tariffs on the average price of US exports. The allowance for both quality and

cost dimensions is similar to Baldwin and Harrigan (2011), and Johnson (2009). However, unlike

both papers, I do not assume a specific relationship between quality and costs, rather showing that

variety-specific revenue captures the characteristics necessary to evaluate a variety-specific effect

4

of tariffs. Indeed, the use of revenues as measure that captures characteristics of both costs and

quality is similar to recent work by Hallak and Sivadassan (2009).

The rather robust response of low revenue varieties is similar to Kehoe and Ruhl (2009). Indeed,

motivated in-part by the results in Kehoe and Ruhl (2009), Arkolakis (2009) presents a framework

based on endogenous marketing costs that generates a larger response to trade liberalization by low

revenue varieties. While similar to the results in my paper, the results are driven by a completely

different mechanism. Further, there is one crucial qualitative difference in equilibrium predictions,

where the Arkolakis framework guarantees that all firms gain from liberalization, which is not the

case in my paper.3 Indeed, in the forthcoming empirical work, the top-20% of exporters (within

HS8 products) rarely increase trade after tariff liberalization occurs.

There is a relatively recent literature examining the design of the WTO within the context of

heterogeneous countries and products. Saggi (2004) compares optimal tariffs set on an MFN basis

with those set via unconstrained discrimination in a n-country oligopoly model with heterogeneous

suppliers. Generally, Saggi’s model suggests that MFN tariff-setting benefits low cost suppliers

since the optimal tariffs applied to their exports would be lower. The model does not examine the

removal of tariffs on an MFN basis, and the corresponding effects on trade, which is the focus of

my work.4

On broader level, my paper is related to the work of Rose (2004) and Subramanian and Wei

(2007), who estimate the effects of GATT/WTO membership on bilateral trade flows using a stan-

dard gravity model. Critically, my model suggests that this particular class of empirical studies is

misspecified by failing to allow for differential effects of GATT/WTO membership as a function of

pre-GATT/WTO market penetration within products. In this way, the results detailed below may

also motivate modifications of the standard Anderson and Van Wincoop (2003) gravity model.

In terms of the US case study employed in this paper, a number of papers are relevant. Romalis

(2006) shows that MFN liberalization by the United States increased both the degree of openness

and the growth rates of developing countries. Ludema and Mayda (2009) examine the relationship

between exporter concentration and observed tariff concessions by the US. In particular, they show

that the US offered larger tariff concessions in products with a more concentrated group of export

sources. Finally, very recent work by Feenstra and Weinstein (2010) adapts the methods from

3Via a CES demand assumption, all firms in the Arkolakis (2009) framework receive the same percentage demandshock resulting from tariffs. Thus, all firms gain from a reduction in a common iceberg transport cost, but differ intheir supply response based on the marginal cost of reaching a larger fraction of consumers.

4In later work, Saggi and Sara (2007) uses a two country model to examine the National Treatment clause whenproducts may differ in quality. A similar result is reached as in Saggi (2004), where a national treatment clause tendsto help those exporters selling the most competitive goods (highest quality goods).

5

Feenstra (1994) to estimate the gains for trade using a non-CES (translog) demand system. Their

estimates imply pricing/mark-up behavior which is consistent with Melitz and Ottaviano (2008) -

the model at the heart of my paper.

The rest of the paper is organized as follows. Section two presents a simple theoretical model,

with an extension provided in the appendix. Section three details the dataset, and tests the pre-

dictions from section two. In section four, I briefly conclude.

2 Theory

Consumers

The key to the model is the way in which consumers value product variety and quality. Similar to

Melitz and Ottaviano (2008), I assume that consumer preferences are quasi-linear, non-homothetic,

and exhibit love-of-variety within a differentiated sector. However, I depart from Melitz and Otta-

viano by assuming that within the differentiated sector consumers earn utility based on “quality-

adjusted consumption” of each variety. Preferences of this type can be defined as follows:

U = x0 + θ

∫i∈Ω

λiqidi−1

2η

(∫i∈Ω

λiqidi

)2

− 1

2γ

∫i∈Ω

(λiqi)2 di (1)

Here, Ω is the measure of available varieties, qi is the consumption of variety i, and λi is its

associated quality. Thus, quality-adjusted consumption of variety i is defined as λiqi. The precise

role of quality will be discussed shortly. Further, x0 is the numeraire good, where θ (> 0) and η

(> 0) determine the substitution pattern between the differentiated industry and the numeraire.

Finally, γ (> 0) represents the degree to which consumers value product variety.

The budget constraint faced by consumers is written as,

x0 +

∫i∈Ω

pciqidi ≤ I

where I is income of the representative consumer, and pi is the price of variety i. Solving the

maximization problem of the representative consumer yields the following inverse demand function

for each variety,

pci = λi (A− γλiqi) (2)

where,

A = θ − η∫i∈Ω

λiqidi

6

Here, pci is the price at which consumers purchase varieties. Also, note that the demand intercept,

A, embodies the competitiveness of the differentiated market. All else equal, the market is more

competitive if the quality-adjusted consumption of all varieties,∫i∈Ω

λiqidi, is larger.

Before moving to the firm’s problem, it is instructive to examine the effects of variety-specific

quality, λi, on the demand for each variety. First, note that by differentiating (2), the willingness

to pay for a variety is increasing in quality if:

∂pci∂λi

= A− 2γλiq > 0

This will be satisfied in equilibrium, as A− 2γλiq > 0 when marginal revenue is positive. Second,

note the equation for demand elasticity as a function of variety specific quality and price:

εD,i =∂qiqi

pci∂pci

= − pciAλi − pci

Two results are apparent in the equation for εD,i. First, conditional on quality, varieties sold at

lower prices will be sold at lower absolute demand elasticities. Further, conditional on price, higher

quality varieties (higher λi) will be sold at lower absolute demand elasticities. Given that consumers

will tend to spend higher shares of their income on varieties with a lower quality-adjusted price,

in equilibrium, there will be a negative relationship between the absolute elasticity of demand and

variety-specific revenues. The relationship between revenues and elasticities is crucial for the main

theoretical results of the paper, and the link to the empirics.

Firms

For this particular model, each supplier will produce one variety, and thus suppliers are also indexed

by i. Further, supplier i will produce variety i at a constant marginal cost, ci. I will not assume

a specific relationship between the marginal cost of production ci and the quality of each variety,

λi (different from Baldwin and Harrigan, 2011, and Johnson, 2009). It will soon be clear that

the relationship between marginal costs and quality has no bearing on the results outside of their

relationship to the observed distribution of revenues within products.

For simplicity, I assume away the extensive margin of trade, focusing instead on a fixed measure

N of imported foreign varieties. An extended version of the model with an active extensive margin

is presented and discussed in the appendix. I also assume that there exists no domestic sector.5 To

5The results of the model go through with a domestic sector. The only difference is that a domestic sector tempersthe effect of tariffs on the aggregate measure, A. However, without reliable data that reports the size of the domestic

7

sell a variety in the import market, the supplier must pay an ad-valorem tariff τ for each unit sold.

Hence, the relationship between the consumer price detailed above and the price that producers

receive is pci = (1 + τ)psi . This yields the following inverse demand function that foreign suppliers

use to optimally set production for the import market.

psi =λit

(A− γλiqi)

Here, t = (1 + τ). Suppliers choose quantities to maximize profits:

π (λi, ci) = maxqi

λit

(A− γλiqi) · qi − ciqi

Optimal production of a given variety i, q (λi, ci), is written as,

q (λi, ci) =A− ci

λit

2γλi, (3)

where the price received for this variety is,

p (λi, ci) =λit

(A+ ci

λit

2

), (4)

and finally, the pre-tariff value of imports of variety i is written as:

v

(ciλi

)=

A2 −(ciλi

)2

t2

4γt

. (5)

As trade data reports the value of trade, and not the profits of each supplier, equation (5) is the

object of interest. The value of trade will be affected by tariffs through two channels. The first is

directly via the negative effect of tariffs on variety-specific revenue. This can be seen by the negative

impact of tariffs on the numerator of (5), and the positive impact of tariffs on the denominator of

(5). However, there are also indirect effects of tariffs through the competitiveness term, A:

A = θ − ηQ

= θ − ηNq

sector at the HS8 level, I cannot test this aspect of the model. Hence, I choose to develop the model without adomestic sector.

8

Here, q is the average quantity of imported varieties. Defining(cλ

)as average quality-adjusted costs

of imported varieties, and solving for A, we have,

A =

(2γ

2γ + ηN

)θ +

(ηN

2γ + ηN

)( cλ

)t (6)

In (6), higher tariffs increase A. Intuitively, higher tariffs decrease quality-adjusted production for

the import market, thus making the market less competitive. This yields a higher residual demand

for each variety, A. A useful way to characterize this result will be the elasticity of A with respect

to t:

εA ≡∂A

∂t

t

A≡

η(cλ

)t

2γθN

+ η(cλ

)t∈ (0, 1) (7)

While A is clearly increasing in t, on a percentage basis, A increases slower than t. Further, εA is

increasing in N and η. Respectively, these imply that in more competitive markets, in which there

is a larger degree of substitution between the differentiated sector and the numeraire, the response

of A is stronger relative to a change in t.

Trade Liberalization

As stated in the introduction, the goal of this paper is to evaluate how the effects of trade liber-

alization are distributed across export suppliers that may not sell varieties of the same quality, or

varieties produced at the same marginal cost. By taking logs and then fully differentiating (5) with

respect to all endogenous variables and t, I can write the following:

∂vivi

=2A∂A− 2t

(ciλi

)2

∂t

A2 −(ciλi

)2

t2− ∂t

t

To simplify this expression, I will exploit the monotone link between(ciλi

)2

and v(ciλi

). Solving

for(ciλi

)2

in (5) and substituting into the above equation yields a simple relationship between the

marginal effect of tariffs and the value of imports prior to the change in tariffs.

∂vi =A2

<0︷ ︸︸ ︷(εA − 1)

2γt

∂t

t+ vi

∂t

t(8)

9

Defined as a total elasticity of import value with respect to tariffs, I can write:

εv,i ≡∂vi∂t

t

vi=A2

<0︷ ︸︸ ︷(εA − 1)

2γtvi+ 1 (9)

In (8), the marginal effect of tariffs is a function of two terms. The first is a negative effect which

is common across all suppliers. The second is specific to the value of imports prior to the change

in tariffs, where the negative effect of tariffs is mitigated for higher value imports. This result is

summarized in the first proposition of the paper.

Proposition 1 The elasticity of trade value with respect to tariffs for variety i is increasing in

revenues earned by variety i.

Proof. Immediate from (8)

The result in Proposition 1 simply establishes that there is a disparate effect of tariffs across

imports within products, where specifically, these effects are governed by the elasticity at which each

variety is imported. In this case, high-quality and/or low-cost varieties are consumed at a lower

absolute elasticity of demand, and hence, experience a smaller effect from a percentage supply shock

(in this case a tariff cut).

There exist other papers, Arkolakis (2009) in particular, that yield a similar result to Proposition

1 but are not due to differences in demand elasticity. I will discuss this and other papers later in the

manuscript. However, in order to set-up a comparison of the existing literature with the demand-

side effects which are the focus of this paper, I will first go deeper into the quality-cost framework,

and in particular, the way in which tariffs induce shifts in demand for each variety.

Crucially, when noting that the tariff changes in this model are pervasive across all imported

varieties (MFN tariff cuts), competition is more fierce after the tariff cut. This is embodied in the

residual demand level, A, which falls as tariffs fall. In equilibrium, the fall in A may be larger or

smaller than the direct effects summarized in Proposition 1. I now examine when, and for which

varieties, changes to A are large enough to outweigh these direct effects. Precisely, lower tariffs

increase imports of variety i only if the value of imports prior to lower tariffs is sufficiently small.

That is, given ∂tt< 0, ∂vi > 0 only if:

vi <A2 (1− εA)

2γt

10

In contrast, when vi is sufficiently high, it is possible that lower tariffs decrease trade. To see this,

using (5), the highest possible value of trade is written as:

vmax = v (0) =A2

4γt

Substituting vmax into ∂vi, we have:

∂vmaxi =

A2 (2εA − 1)

4γt

∂t

t

Thus,∂vmaxi

∂t> 0 if εA > 1

2. In other words, the percentage change in residual demand A must be

sufficiently large. Using (7), as a function of model parameters, this occurs if:( cλ

)t >

2γθ

ηN(10)

This condition is summarized in the following proposition:

Proposition 2 If γθηN

is relatively small, then tariff liberalization increases imports only if pre-

tariff-cut revenues are sufficiently small.

As detailed in Proposition 2, within this adjusted Melitz-Ottaviano framework, it is possible for

the value imports of a given variety to fall when tariffs are cut. This is more likely to occur when love

of variety γ is small, η is large or N is large. Hence, in the forthcoming empirical work, along with

the differential effect of tariffs described above, I will focus on the absolute tariff-response of high

revenue varieties in order to examine whether variety-specific demand elasticities are empirically

relevant in the way the models suggests. Further, as N can be approximated in the empirical

framework, I will evaluate how these effects change when looking at products with relatively large

values of N .

The Role of Quality

A second way to distinguish this particular theory from alternate frameworks is to focus on the role

of quality, and in particular, the role quality plays in determining revenues, prices, and demand

elasticities.

When demand is of constant elasticity and consumers make decisions on quality-adjusted quan-

tity (utility defined over λiqi), within a CES demand system, quality and cost have no effect on

equilibrium demand elasticities. In contrast, within the quadratic preferences described above,

11

quality and cost have a very clear effect. Precisely, equilibrium absolute demand elasticities are

written as:

εD,i =A+ ci

λi

A− ciλi

(11)

In (11), equilibrium demand elasticities are increasing in costs and decreasing in quality, which

implies a monotone relationship between quality-adjusted costs and absolute demand elasticity.

In deriving the elasticity of trade value with respect to tariffs, this relationship is crucial, since

there is also a monotone, in this case negative, relationship between vi and ciλi

. Hence, initial trade

value captures exactly the variety-specific characteristic (demand elasticity) required for the main

propositions in the paper.

In contrast, when quality plays a role, variety-specific quantity and price do not necessarily

capture the necessary characteristics related to demand elasticities. To see this, first focus on

variety-specific price, p (λi, ci) = λit

(A+

ciλit

2

). Here, p (λi, ci) is increasing in both variety-specific

quality and cost. However, since absolute demand elasticities are falling in quality and rising in

cost, this implies that unless there is a specific relationship between quality and cost, there does

not exist a monotone relationship between prices and demand elasticities. Hence, prices do not

necessarily capture the variety-specific characteristics relevant to evaluate the effects of tariffs on

bilateral trade flows.

Next, consider variety-specific quantity, q (λi, ci) =A− ci

λit

2γλi. Clearly, q (λi, ci) is decreasing in ci.

The question is whether q (λi, ci) is increasing in λi. To see when this is the case, I can write the

elasticity of quantity with respect to λi as follows:

εD,λi =ciλit− 1

A− ciλit

(12)

In (12), quantity is increasing in quality only if ciλit > 1. Hence, if this condition is satisfied for all

i, then like revenues, quantity will have a monotone relationship with demand elasticities. If not,

then quantity does not have a monotone relationship with demand elasticities, and cannot be used

to evaluate any disparate effect of tariffs across varieties.

The discussion regarding demand elasticities is summarized in the following proposition.

Proposition 3 Without a quality dimension, variety-specific revenues, quantities, and prices all

have a monotone relationship with demand elasticities. With a quality dimension, only variety-

specific revenues are guaranteed to have a monotone relationship with demand elasticities.

The upshot of this discussion and the result in Proposition 3 is that when evaluating the effects

12

of tariffs, if quality plays a role at all, the only unconditional proxy for quality-adjusted cost is

revenue. Hence, as suggested by equation (8), when testing the model, I will construct measures

of within-product market penetration based on revenues, where the main predictions are that (1)

tariff liberalization disproportionately benefits the lowest-revenue varieties and (2) it is not necessary

that high revenue varieties gain at all from tariff liberalization. Further, to evaluate a potential

role for quality, I will also construct within-product rankings of quantities and prices. If the latter

two measures deliver results that are qualitatively similar to revenue-based rankings, then I will

conclude that support for the model is primarily driven by cost-heterogeneity, or a trivial link

between cost heterogeneity in quality heterogeneity. In contrast, if using rankings based on quantity,

and especially price, yield different results when compared with those based on revenues, then I will

conclude that quality is likely playing a role in demand elasticities, and hence, the effects of tariffs

on trade flows.

Discussion

The results detailed above are not the first to look at the relationship between product characteristics

and bilateral trade flows. For example, recent work by Baldwin and Harrigan (2011) and Johnson

(2009) have allowed for a precise impact of quality, via prices, on trade flows. Further, Hummels and

Skiba (2004) examine how trade costs and tariffs affect the price-composition of exports. However,

to my knowledge, the above framework is the first to demonstrate that the impact of import

liberalization may in fact be negative for some measure of imported varieties.

This main theoretical result is closely related to the discussion of tariffs in Hummels and Skiba

(2004). Along with per-unit trade costs, Hummels and Skiba allow for an ad-valorem component,

such as tariffs. Theoretically, they show that higher tariffs reduce the relative demand for high

quality (price) goods, thus lowering the average FOB price of exports. The opposite occurs with

transportation costs. Relative to Hummels and Skiba (2004), my model extends the literature in

a unique way, where I allow for elasticity differences that are arbitrarily correlated with prices. To

be more specific, consider an example in which all imported varieties are produced at the same

marginal cost, but yet differ in quality. In this case, higher quality varieties earn higher revenues,

and are sold at lower equilibrium demand elasticities. If raising tariffs, high quality varieties will

suffer less than low quality varieties, by virtue of the equilibrium demand elasticity at which each

are sold. Thus, higher tariffs increase the relative demand for high quality goods - the opposite of

what Hummels and Skiba (2004) predict. The critical difference is that the quality component in

my model may shift/skew the demand curve such that higher price goods are sold at lower demand

elasticities. In contrast, the model in Hummels and Skiba (2004) is consistent with a restricted

13

version of my model where quality and costs and correlated in such a way to ensure that demand

elasticities are always larger with higher prices. Indeed, in later empirical results, I will present

evidence that suggests that some low prices goods are actually subject to relatively high demand

elasticities.

As mentioned earlier, Arkolakis (2009) presents a model that delivers a differential response to

tariffs similar to Proposition 1, but that derives from features unrelated to demand elasticities. In

particular, Arkolakis details a model in which firms reach consumers by investing in marketing,

itself subject to a convex cost function. Via a CES demand assumption, all firms in the Arkolakis

framework receive the same percentage demand shock resulting from tariffs. Thus, all firms gain

from a reduction in a common iceberg transport cost, but differ in their supply response based on

the marginal cost of reaching a larger fraction of consumers. Indeed, small/less-productive firms

that charge higher prices, who are not as far up the marketing cost curve, experience a larger total

elasticity of import value with respect to tariffs. The difference in predictions between Arkolakis

(2009) and my work are subtle, though where they exist, they highlight the demand-side issues

at the heart of my model. Specifically, Proposition 2 highlights the possibility that high revenue

varieties do not increase imports after tariff liberalization. Indeed, in the forthcoming empirical

work, I show that the top-20% of exporters (within-products) rarely increase trade after tariff

liberalization occurs. Further, Proposition 4 implies that high-price varieties benefit the most from

liberalization only if quality has no role in demand elasticities. I will also show that if anything,

low-price varieties benefit the most, suggesting a role of quality that affects demand elasticities, and

hence, the effects of tariff liberalization on bilateral trade flows.

Finally, before moving to the empirics, a note about welfare. From a trade policy perspective,

the results in Proposition 1 seem in direct conflict with the intent of the central tenet of the WTO:

the principle of non-discrimination. That is, while non-discrimination requires that countries apply

and liberalize tariffs equally across MFN exporters, the benefits of that tariff reduction do indeed

discriminate along natural margins. The question then becomes, is non-discrimination a sensible

policy?6 Or should countries be allowed to discriminate along some margin?7

6As detailed in Jackson (1997), there are three traditional arguments which support the use of MFN. One is thatMFN prevents policies which distort bilateral patterns of comparative advantage. Further, some argue that MFNresults in more liberalization than under preferential systems. Finally, others argue that allowing for discriminatorytariffs would increase the costs of rule formation, where many tariff lines would be more costly to impose and enforcewhen compared to a single tariff line applied to all exporters.

7This debate has a very long history. Indeed, Irwin, Mavroidis, and Sykes (2007) describe how Keynes’, duringthe framing of the GATT, wished to keep open the option of discriminatory tariffs. While this was mainly arguedwithin the context of Great Britain’s imperial relationships, Keynes seems to express a notion that the post-wartrading system ought not to be constrained by MFN.

14

While I leave an in-depth analysis of welfare within a multi-country political economy model

for a follow-up paper, I can comment on how consumers in the importing country view non-

discrimination, and what would happen if they were allowed to discriminate. Specifically, I can

show that, similar to a Ramsey-type tax rule, the importing country would exacerbate the disad-

vantage to high revenue suppliers under discrimination.8 This is similar to the result in Saggi (2004)

that was discussed in the introduction. Hence, it seems that non-discrimination, while not neces-

sarily helpful to high revenue varieties, likely is preventing optimal discrimination that would make

their situation worse. Crucially, as this argument is based on different elasticities within products,

it begs the question whether there is evidence of variable elasticities in the data. To examine this

possibility, I now move to the empirical section of the paper.

3 Tariffs and bilateral trade: The US and the Uruguay

Round

In this section, I use detailed import and tariff data for the United States, over the period 1989 to

2004, to estimate the degree to which the effects of MFN tariffs differ within products. To do this,

I use HS10 bilateral import data from the UC Davis Center for International Data, as described in

Feenstra, Romalis, and Schott (2001). HS8 tariff data is obtained from the US International Trade

Commission and from Romalis (2004).9 I restrict attention to continuing HS10 products over the

entire sample, which abstracts from new and/or dying HS10 codes.10 More information regarding

the construction of the sample will be provided shortly.

The level of refinement in the import data will be the HS10-Exporter level. I will henceforth

refer to HS10-Exporter pairs as “varieties”. Since MFN tariffs in the US are set at the HS8 level,

I will refer to HS8 classifications as “products”. “Industries” will be defined either at the HS2 or

HS4 levels of aggregation.

The theory described in section two focused on the effects of a reduction in a common ad valorem

tariff that is pervasive across varieties. As such, I will restrict the sample to focus on the effects of

8I can prove this result in two separate ways, both of which are available upon request. First, starting from freetrade, I examine whether an importing country with no domestic sector can change tariffs in a discriminatory mannersuch that zero revenue is earned but consumer surplus increases. In this case, the optimal policy is to increase thetariff on high revenue varieties and subsidize imports of low revenue varieties. In the second example, also startingfrom free trade but allowing for revenues to change, I evaluate the direction of an optimal policy on a single variety.Precisely, I show that there exists a threshold of quality-adjusted costs above which import subsidies are optimal,and below which import tariffs are optimal.

9The Romalis data is also obtained from the UC Davis Center for International Data.10See Pierce and Schott (2009) for a discussion and concordance for new and dying product codes.

15

Figure 1: MFN Tariffs: 1989-2004

1990 1995 2000

02

46

810

Mean and Median MFN Tariffs

Year

MFN

Tar

iff

Mean TariffMedian Tariff

0 20 40 60 80 100

010

2030

40

MFN Tariff Distribution - 1992 and 2004

Percentile

MFN

Tar

iff

19922004

ad-valorem MFN tariffs on MFN imports. The issue, however, is that not all imports are subject

to MFN tariffs, and not all tariffs are ad-valorem. To deal with this, I will adopt the following

restrictions on the observations included in my dataset.

R1: HS10-Exporter pairs must always be imported under MFN over the entire sample period.

R2: HS8 products must not list any non-tariff barriers to trade over the entire sample period.

The first condition removes non-MFN trade from the dataset, and in particular, varieties for

which preferential access is an (relevant) option. This is particularly important for NAFTA trade,

and to test this restriction, I will later drop NAFTA countries from the sample to examine how the

results may change. In the second restriction, I remove products for which at some point in the

sample, non-tariff barriers, or specific tariffs, are listed in the HS8 tariff files.

Before presenting the precise regression framework (which will further restrict the sample), I first

detail the reductions in US import tariffs which resulted from the Uruguay Round. Two particular

characteristics of this round of tariffs cuts are presented in Figure 1. In the left panel, I have plotted

16

mean and median tariffs across HS8 products for each year, 1989-2004. The last year prior to the

enactment of the Uruguay Round (1994) is denoted by the vertical line. Notably, there is very

little movement in tariffs prior to the enactment of the Uruguay Round. Directly after, tariffs fell

steadily, flattening out between 2000 and 2004. Overall, mean and median tariffs fell by roughly

50%. Further, in the right-panel of Figure 1, I plot the empirical distribution of MFN tariffs in 1992

and 2004 (which will be the years over which I evaluate the effects of tariffs). Of note, US MFN

reductions in the Uruguay Round doubled the number of MFN tariffs that were zero.

Next, to give a broad picture of the effects of tariff reductions, I will impose two additional

restrictions on the sample:

R3: Tariffs must be stable over the period 1989-1994

R4: Tariffs must not rise over the period 1994-2004.

Together, the two restrictions remove 2.2% of HS8 products from the dataset, which removes

only 0.29% of the total number of HS10-Exporter pairs. The first restriction focuses the sample to

products with a stable “pre-period” during which initial market penetration can be measured. The

second restricts products to those where tariffs were cut or stayed constant after the formation of

the WTO. Since non-tariff barriers were allowed to be “tariffied” into ad-valorem rates upon their

removal, getting rid of the few products with increasing ad-valorem tariffs provides a secondary

check that these products are removed from the database.

In Figure 2, I provide some illustrative evidence on the effect of MFN tariff reductions, and a

differential effect of tariff reductions by market penetration before the tariff cut. To begin, focus on

the left panel of Figure 2, where I have plotted relative trade growth for products that received a

tariff cut and products that did not. To calculate relative trade growth, I first calculate the log of

the ratio of imports in year t to imports in 1992 for each HS10-Exporter pair. Then, for each year,

I de-mean these growth rates across Exporter-HS4 pairs. Thus, in each year, average trade growth

relative to 1992 is equal to zero. In the left panel of Figure 2, I have decomposed this “zero” into

HS8 products that received a tariff cut, and HS8 products that did not. Further, I have denoted the

first draft of the final act of the Uruguay round, which occurred at the end of 1991, and the year in

which the final act was signed, 1994, by vertical black lines. Clearly, products that received a tariff

cut experienced positive trade growth relative to those products that did not. More notably, trade

growth rates for each product group were very close to one another prior to 1994. In the appendix,

I present another chart showing that trade growth rates were similar all the way back to 1980.

This suggests that after controlling for factors unrelated to tariff cuts (the Exporter-HS4-Year fixed

17

Figure 2: Bilateral Trade Growth - 1989-2004

1990 1995 2000

-0.4

-0.2

0.0

0.2

0.4

Trade Growth and Tariff Cuts

Year

Trad

e G

row

th (1

992

Bas

e)

First Draft ->

HS8 Tariff CutNo HS8 Tariff Cut

<- Final Act Signed

1990 1995 2000

-0.4

-0.2

0.0

0.2

0.4

Trade growth by 1989/1990 Import Penetration (calculated within HS8 products)

Year

Trad

e G

row

th (1

992

Bas

e)

First Draft -> <- Final Act Signed

Bottom 80% + Zeros - HS8 Tariff CutBottom 80% + Zeros - No HS8 Tariff CutTop 20% - HS8 Tariff CutTop 20% - No HS8 Tariff Cut

effects), products which received a tariff cut were not growing differently than those that did not

receive a tariff cut.11 Given that good instruments for tariffs at the HS8 level do not exist for this

analysis (to the author’s knowledge), the fact that trade growth pre-WTO appears to be invariant

to future tariffs cuts is particularly helpful.

In the right panel of Figure 1, I have further decomposed these “zeros” by the relative posi-

tion of each HS10-Exporter observation within their respective HS8 product category. Precisely, I

decompose the averages in the left-panel by whether or not, within each HS8 product, an HS10-

Exporter was in the top 20% of positive trade values over the period 1989-1990. Clearly, the group

11This is a statistically robust statement. Regressing the log change in imports of each HS8 product between1989 and 1992 on a dummy variable identifying a tariff cut and HS4 fixed effects yields an insignificant relationshipbetween pre-Uruguay round trade growth rates and future tariff cuts. Precisely,

log(vi,j,92vi,j,89

) = −0.0647[0.054]

· I(ti,04ti,94

< 0) + HS4

where i represents HS10 varieties and j represents exporters. Standard errors are robust and clustered by Exporterand HS2 industry.

18

that benefited the most from a tariff cut relative to their no-tariff cut control group were those

HS10-Exporter pairs that either did not trade (“zeros”) or were in the bottom 80% of their given

product category. In contrast, it is unclear whether varieties in the top 20% of their given HS8 cat-

egory benefited from tariff cuts relative to similar varieties in the no-tariff cut HS8 control groups.

Overall, the evidence in Figure 2 details a differential impact of MFN tariffs which is consistent

with the theory developed in section two. I now test the robustness of these features using a broad

set of regressions.

Tariffs and Relative Market Penetration: Baseline regressions

Given that tariffs only changed once and over a relatively short period, the empirical strategy will

be similar to Trefler (2004), using long differences to evaluate the effects of tariffs on bilateral trade.

The object of interest will be the growth rate in trade over the period 1992-2004. This growth rate

will be labeled log(vi,j,04/vi,j,92), where vi,j,s measures the value of imports in nominal US dollars of

HS10 variety i from exporter j in year s. Hence, this growth rate requires the following (and final)

restriction on the sample used for the majority of the analysis.

R5: HS10-Exporter pairs must report imports in both 1992 and 2004.

The growth rate will be regressed on a number of factors within a number of different specifica-

tions, though the common regressor in each will be the log change in MFN tariffs over the period

1994-2004. Precisely, I will measure tariffs as in section two, where ti,s = 1 + τi,s. Here, τi,s is the

percentage point MFN tariff for HS10 variety i in year s (though understanding that tariffs are set

at the HS8 level). To calculate a measure of import elasticity, I will measure the change in tariffs

using log differences, log(ti,04ti,94

), similar to the way in which I measure the change in import value

for each HS10-Exporter. Finally, note that I measure the change in tariffs starting in 1994 because

tariffs were essentially frozen during WTO negotiations. Via (R3), those few products with tariffs

that move in the pre-period are dropped from the dataset.

Restriction (R3) is crucial in measuring pre-tariff-cut market penetration, as doing so requires

that policy measures were not moving. Initial market penetration will defined by six sub-groups,

five of which based on quintiles of positive imports over the period 1989-1990 (the earliest periods

such that HS10 classifications were used). After dropping observations of varieties that did not trade

in both 1992 and 2004, within each HS8 product, I first collect all varieties (HS10-Exporter pairs)

that did not report trade in both 1989 and 1990, and call this “group zero”. This group is identified

by a dummy variable D0i,j. Next, HS10-Exporter observations for which imports occurred during

1989-1990 are grouped into quintiles within each HS8 product. These quintiles will be identified in

19

Table 1: Mean and Median Import Value by within-product quintile

Group: 0 1 2 3 4 5Mean ($’s): 0 89,143 355,891 1,038,760 3,336,006 21,426,412

Median ($’s): 0 31,880 144,084 444,908 1,409,428 6,151,874Notes: This table presents mean and median import values over HS10-Exporter pairs broken downby import penetration group.

order by dummy variables D1i,j, D

2i,j, D

3i,j, D

4i,j, and D5

i,j, with the lowest 20% of positive imports

within each HS8 product identified by D1i,j, and the highest 20% of positive imports within each

HS8 product identified by D5i,j.

Table 1 presents the mean and median import values over 1989-1990 for each of these groups,

after dropping HS8 products without five HS10-Exporter observations with positive trade (i.e.

dropping HS8 products where quintiles are not defined). The point to be taken from Table 1 is that

the top 20% of HS10-Exporter observations are, on average, roughly 7X larger than those in the

60-80th percentile. A similar magnitude difference is evident at the median. Overall, the largest

20% of exporters are, in fact, quite large, which is consistent with similar results presented at the

firm level in Bernard, Jensen, Redding, and Schott (2007). Thus, given the theory in section two, I

hypothesize that the effects of MFN liberalization on this group are modest relative to groups 0-4.

I now seek evidence in support of this hypothesis.

To test for any differences across quintiles, I start by estimating the following equation.

log(vi,j,04

vi,j,92

) =(βUS D0D

0i,j + βUS + βUS D2D

2i,j + · · ·+ βUS D5D

5i,j

)log(

ti,04

ti,94

) (13)

+βD0D1i,j + · · ·+ βD5D

5i,j + Fixed+ εi,j

In (13), I allow for a differential impact of tariffs across groups, where in particular, I measure the

effects of group membership and a differential effect of tariff cuts relative to group one. In (13),

βUS represents the long-run elasticity of import value with respect to tariffs for the group of HS10-

Exporters that reported trade in 1989 and 1990, but were in the bottom 20% of reported import

values within their respective HS8 product. Thus, βUS Dk represents the difference in elasticity,

relative to group one, for group k. To be clear, βUS D5 represents the difference in import elasticity

of the largest exporters relative to the smallest exporters.

20

An alternate way of estimating the model in (13) is by directly estimating elasticities for each

group, and not differences relative to a base group. Equation (14) does exactly this:

log(vi,j,04

vi,j,92

) =(βUS D0D

0i,j + · · ·+ βUS D5D

5i,j

)log(

ti,04

ti,94

) (14)

+βD1D1i,j + · · ·+ βD5D

5i,j + Fixed+ εi,j

In both (13) and (14), I will include a robust set of fixed effects at the Exporter-Industry level

to control for trends unrelated to changes in MFN tariffs. For example, some exporter-industry

pairs may be declining or growing due to development or a long-run adjustment toward different

industries. This is particularly important in light of the Kehoe and Ruhl (2009) explanation for

extensive margin growth, which is that price-shocks in countries like China are one factor driving the

growth of the extensive margin. If price-shocks play a critical role, these Exporter-Industry trends

should help control for them. Further, these Exporter-Industry fixed effects may help control for

export sources which engage in liberalization of their own in certain industries, and hence, there

are additional issues of reallocation that may be absorbed in the fixed effects.

I will occasionally include two additional controls, dV Ti,t and V Growi,t. In the first, I measure

the log of total trade growth between 1992 and 2004 for products in variety i′s industry (defined

either by HS2 or HS4 industries) but not variety i′s HS8 product. This is especially important in

light of tariff cuts in other HS8 products that are in the same HS2 or HS4 industry. I hypothesize that

the coefficient on this variable will be less than zero, where higher demand outside an HS8 industry

will force down demand within that particular HS8 industry.12 The second control, V Growi,t,

measures the log growth in imports for each HS10-Exporter variety over the period 1989-1990. I

will only be able to use this control when excluding group zero from the sample. The motivation

for adding in this control is to at least partially rule out a mean-reversion story. If there is mean

reversion, high import growth in 1989-1990 might suggest low import growth in later periods.

The results from estimating (13) and (14) are presented in Table 3. Within each quadrant in

Table 3, I present estimates for a restricted version of the model without tariff interactions in the

first two columns. In the second two columns, I add in tariff interactions from (14). Finally, in the

last two columns, I present the results for (13), which evaluates differential elasticities.

Table 3 reports strong support for the model developed in section two. To begin, I will discuss

the results for the full sample, which are in the top panels of Table 3. First, note columns 1

12For example, suppose that tariffs fall 3% in a given HS8 product, but that tariffs in their particular HS2 industryfall on average 10%. Thus, while the within-HS8 tariff cuts are the focus of the paper, disproportionally large outsidetariff cuts may shift down demand for all varieties in a given HS8 industry.

21

and 2, which report that the average elasticity of trade value with respect to tariffs is negative

and significantly different from zero. Depending on the assumptions over fixed effects, the point

estimates of the average elasticity of import value with respect to tariffs ranges from -3 to -6. As

foreshadowed in Figure 1, the tariff cuts which were implemented during the Uruguay Round had

a significant effect on imports into the US.

In terms of the predictions from section two, focusing on columns 4 and 10 in the top panel of

Table 3, it is clear that for varieties in group one, the smallest 20% of active varieties in 1989-90 as

measured by revenues, tariffs had a negative, large, and statistically significant effect on the growth

rate of imports between 1992 and 2004. As we move down rows from group one toward group

five, we see that the interaction with tariffs becomes less negative with each successive group. This

culminates with varieties in group five, where tariffs did not have a significant negative impact on

import values. Further, in columns 6 and 12, it is clear that when measuring a differential elasticity

between groups 1 and 5, the difference is positive and significantly different from zero. Focusing

on the last row in the top panel of Table 3, we see dV Ti,t has a negative and significant impact on

bilateral imports.

Next, in the bottom panels of Table 3, I estimate the model but while dropping varieties for

which trade was not reported in 1989 and 1990. Here, we find that the results are robust within

this smaller sample. Within this sample, we are also able to estimate the effect of both dV Ti,t

and V Growi,j,t. The former again has a significant and negative effect on trade. For the latter,

high trade growth in the pre-period tends to be associated with low growth in later periods. This

suggests that some sort of mean reversion is present in the data, though not enough such that the

effects of tariffs by groups are changed in a large way.

One last issue to discuss is the estimates for the group dummy variables that are not interacted

with tariffs. Here, we find that higher revenue varieties tend to have lower growth rates. This would

be supported by the model if I also included a cost shifter across varieties that was not related to

tariffs. For example, this might include a common rate of productivity growth, independent of

whether or not a product received a tariff cut.

Tariffs and Relative Market Penetration: Exporter Interactions

Next, I estimate a number of specifications which include interactions between US MFN tariffs and

exporter characteristics which are more general than variety-specific market penetration. First, I

estimate (13) allowing for a differential impact of tariffs as a function of exporter GDP per capita

(data taken from the Penn World Tables). The coefficient on this interaction is labeled βUS GDP .

This interaction is sensible on a number of levels. First, less-developed countries may have poor

22

institutions, and thus may respond slowly to changes in tariffs relative to more developed exporters.

In this case, βUS GDP would be negative. Or, perhaps developed countries are more likely to employ

unionized labor, which may yield a laggard response to price shocks. In this case, βUS GDP would

be positive. Another explanation for positive βUS GDP would be that developed countries are more

likely to sell high quality products, and that the effect of high quality products may be picked up

to some extent by this coefficient. Indeed, in the top panels of Table 4, though insignificant at

conventional levels in all regressions, this is exactly what I find, where there is an imprecise but

positive relationship between the effect of tariffs and GDP per capita of the exporting country.

However, despite this positive interaction between tariffs and exporter GDP per capita, we still find

that there is a positive and statistically significant difference in estimated elasticities between group

one and group five, as the theory predicts.

Moving beyond exporter development, I now allow for a full interaction between exporter dummy

variables and the log change in US MFN tariffs. Precisely, I estimate the following:

log(vi,j,04

vi,j,92

) =(βUS D0D

0i,j + βUS D2D

2i,j + · · ·+ βUS D5D

5i,j

)log(

ti,04

ti,94

) (15)

+ (βUS Exp · Exp) log(ti,04

ti,94

) + βD1D1i,j + · · ·+ βD5D

5i,j + Fixed+ εi,j

In (15), Exp is a vector of exporter dummies, and βUS Exp represents a vector of exporter-specific

import elasticities for group zero. The motivation for this interaction is as follows. In Appendix

A, when including a flexible extensive margin of trade, the relationship between quality-adjusted

costs (as defined by the Pareto shape parameter) and import value is negative only if the measure

of potential exporters is unresponsive to the Pareto shape parameter, and the pool of potential

exporting firms does not vary across exporters. Indeed, the interaction between exporter dummies

and US MFN tariff changes may account for both a larger difference in the pool of potential

exporters, or some differential effect of tariffs on this pool. Also, if financing issues vary substantially

across countries (as in Manova, 2008), then the exporter-tariff interaction is warranted.

The results from estimating (15) are presented in the bottom panel of Table 4. Again, βUS D0−βUS D5 measure the differential effect of MFN tariffs by quintile relative to an exporter-specific tariff

elasticity for group one. Relative to group one, higher revenue varieties fared worse with lower

tariffs than lower revenue varieties. Indeed, this effect is most pronounced in group five. Overall,

despite a huge number of additional interactions with the tariff variable, the results in Table 4 are

supportive of the theory from section two.

Next, I return to the specifications in (13) and (14), evaluating the effects of potential outlier

23

Table 2: Mean and Median Import Value by within-product quintile

Group: 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%HHI: 0.009 0.103 0.156 0.198 0.231 0.266 0.307 0.362 0.441 0.563 0.984

N : 5 9 11 14 17 21 27 37 55 97 429N (no D0): 5 6 8 10 13 16 20 26 36 70 345Notes: This table presents the empirical distribution for a number of HS8 product measures. The first, “HHI”, presents decilesof a share-based Herfindahl Index of Exporter concentration within HS8 products. The last two rows present the number ofHS10-Exporter pairs in each HS8 industry, with and without group zero.

exporters on the results presented in Table 3. In particular, I run regressions excluding NAFTA

countries, and also regressions excluding China. The motivation for the former is that while it is

possible that much trade from NAFTA countries may be accounted as MFN even if preferences

are available (since rules-of-origin can be onerous), it seems odd that either country (in particular

Canada) would enter the database significantly. Regarding the latter, China is obviously a large

and fast-growing export market, which enters the WTO during the sample and also engages in a

substantial amount of vertical trade with the US. Thus, I remove both NAFTA countries and China

separately to evaluate whether the results are driven by large regional trading partners and/or a

massive growing exporter. The results from excluding these countries are presented in Table 5.

Overall, high revenue varieties experience a muted effect of liberalization, and low revenue varieties

experience an amplified effect. Further, relative to group one, high revenue varieties fared worse

with lower tariffs than low revenue varieties, though when China is excluded, this difference is not

significant at conventional levels.

Tariffs and Relative Market Penetration: HS8 Characteristics

The key to the model in section two is that quality-cost heterogeneity yields heterogeneous demand

elasticities across varieties. To rank elasticities, import values are used, since they have a negative,

monotone relationship with absolute demand elasticities. While the results thus far tend to support

theory, I will now dig deeper to evaluate other explanations for the results, along with identifying

HS8 product “types” that are more likely to support the theory.

First, I will examine how the level of exporter concentration within a given product influences

the differential effect of tariffs within the same product. The effects of exporter concentration on

bargained tariff cuts have been examined by Ludema and Mayda (2009), where they show that

import markets with a higher concentration of export sources are forced to engage in larger tariff

concessions. This is confirmed empirically using a US case study over a similar time-period to

24

that in this paper. The intuition is that since tariff cuts in the GATT/WTO are determined via

extensive bargaining, agents in concentrated export sectors are less likely to free-ride off of tariff

concessions negotiated by other exporters (since the other exporters are small in comparison). Thus,

when exporter concentration is relatively large, free-riding is minimized and bargained tariff cuts

in the import market are larger. In the specifications above, the Exporter-Industry fixed effects

should control for mean differences exporter concentration. However, if exporter concentration has

unmodeled and unobserved effects on the differential response to tariffs, the above estimates may

be contaminated.

To examine these issues, within an HS8 product, I first aggregate trade from the HS10-Exporter

level to the Exporter level over the period 1989-1990. Then, I construct a Hirschman-Herfindahl

index (HHI) for each HS8 product based on the market share of exporters. The distribution of HHI

is presented in the first row of Table 2. After identifying the median value of HHI across the entire

sample, I group each product into “Low” and “High” measures of HHI. To examine the effects

of exporter concentration, I interact dummy variables identifying low and high HHI products with

all non-fixed effects in (14). The results are presented in the left panels of Table 6. Here, we

find distinct differences in import elasticities for products that likely have very different industry

structures, where only those products with below median HHI (low concentration) yield estimates

which are consistent with the theory. In particular, for low HHI products, the estimated effect of

tariffs is smallest for the top 20% of export sources in all regressions. In contrast, for products with

high HHI, there is no systematic difference in elasticities across groups. As high HHI products do

not seem to support the model developed in section two, this suggests that high HHI products may

have different underlying demand characteristics, or that supply conditions are more favorable for

suppliers which are highly successful relative to their peers.

Finally, recall that Proposition 2 details a condition on the relative values of γ, η, and N such

that the differential effect of tariff liberalization is strong enough to “kick out” high revenue varieties.

While γ and η are hard to back-out of the data, N can be approximated by the relative number of

varieties within an HS8 product category. In Table 2, the last rows detail the distribution of the

number of HS10-Exporter observations within each HS8 product. Remarkably, there is significant

variation in the number of varieties within HS8 products across the sample. The minimum is 5

(by construction since quintiles require at least 5 observations). The maximum is over 400. I also

report the number of varieties when restricting the sample to varieties with positive trade in 1989

and 1990.

Similar to HHI, I estimate the model with dummy variables identifying below median and above

median values of N , and an interaction of these dummy variables with all non-fixed effects in the

25

model. The estimates, presented in the right panels of Table 6, are supportive of the result in

Proposition 2. Here, when an HS8 product has a relatively large number of varieties, there is never

a significant effect of tariffs on high revenue varieties, and always a significant effect of tariffs on

low revenue varieties. In contrast, for HS8 products with a relatively low value of N , the results

are not supportive of the model.

Tariffs and Relative Market Penetration: Does Quality play a role?

As a final test of the model, I will address a potential role of variety-specific quality by constructing

alternate within-product rankings based on quantity and price, to be evaluated alongside the value

bins that have been used in the paper thus far. As discussed in the simple model in section 2,

pre-tariff-cut revenues map in a monotonic fashion with quality-adjusted costs, and hence, demand

elasticities. In contrast, it is not necessary that quantity or price map to quality-adjusted costs in a

monotonic fashion. Hence, it makes sense to define bins by quantity and price, and to see whether

the qualitative results are similar to value-based bins. If so, then it is likely that quality does not

matter, or that costs and quality are related in a systematic fashion. If not, then it is sufficient to

conclude that non-trivial correlations in quality and cost exist, both of which influencing demand

elasticities.

To define quantity and price bins, I must restrict the sample to products for which units are

consistent during the pre-period. Precisely, I restrict the sample to those HS8 products that use

only one quantity over 1989 and 1990. After restricting the sample in this fashion, I define quantity

and price bins according to a hypothetical ranking if quality played a minimal role. Precisely, when

constructing quantity bins, varieties that did not trade in 1989 or 1990 will be placed in group

Q0. For varieties that were imported in both 1989 and 1990, groups Q1 to Q5 are ranked as above

according to increasing quintiles of quantity within each HS8 product. For prices, P0 will represent

varieties that were not traded in 1989 or 1990. For groups P1 to P5, I assign varieties according

to decreasing unit-value quintiles within HS8 products. To be clear, the highest 20% of reported

prices will be in group P1, and the lowest 20% of reported prices will be in group P5.

To evaluate a potential role of quality, I run the same regression on the same sample, with the

exception that varieties may be placed in different bins depending on the ranking scheme that is

used. If quality plays a role, I hypothesize that value-based bins will capture more variation in

trade growth than quantity and price bins. Further, if quality plays a role, the response to tariffs

when bins are defined by price may be very different than with value bins. The results from these

regressions are presented in Table 7. Whether using the full sample, or the restricted sample,

the results when using value bins are similar to above, where specifically, low revenue varieties

26

benefit strongly from liberalization, and high revenue varieties do not benefit on any significant

level. This is again suggestive of elasticities that are diminishing in revenues. When using quantity

bins, the results are somewhat similar, though the ranking of coefficient estimates is not exactly

consistent with those from regressions using value-based bins. Further, the value of adjusted R2,

which measures the residual variation explained by the model after accounting for fixed effects, is

consistently one percentage point less than when using value bins. Since quality is the only factor

that might remove the positive relationship between quantity and revenues, this suggests a role for

quality in the model.

This point is even more stark when evaluating the differences in regression results when defining

bins according to HS10-Exporter prices. Here, lower price varieties are assigned to higher groups,

which adopts a convention that if quality does not play a role, lower price varieties should be

imported at lower absolute demand elasticities. When looking at these results, first, we see that the

amount of variation captured by the model is less than half of the variation captured when using

value bins. Thus, price bins do not seem to capture the necessary characteristics that govern the

effects of tariffs. Further, we see that of any group, the lowest price varieties (groups 4 and 5 when

defined by price bins), which would have low elasticities with no quality dimension, are the most

sensitive to tariffs. This suggests that there is a relevant role of quality in determining demand

elasticities, and hence, a role of quality in governing the effects of tariffs.

4 Conclusion

This paper has presented a simple model of import market liberalization and bilateral trade. Theo-

retically, I show that when quality/cost heterogeneity yields differences in demand elasticities, trade

liberalization necessarily increases imports only of varieties that are of high cost or low quality -

both of which sold at relatively high demand elasticities and earning relatively low revenues. In

evaluating the model using a case study of US tariff reductions resulting from the Uruguay Round,

I find robust support for the model.

As for future work, the most promising extension is a policy-equipped framework to evaluate

whether, in the presence of quality-cost heterogeneity, the current rules of the WTO are sufficient

to guarantee efficiency in trade negotiations. I plan to address this issue in a follow-up paper.

27

References

[1] Anderson, James E., and Eric van Wincoop (2003), “Gravity with Gravitas: A solution to theBorder Puzzle”, American Economic Review, Vol. 93-1, pp. 170-192

[2] Arkolakis, Costas, “Market Penetration Costs and the New Consumers Margin in InternationalTrade”, mimeo Yale University.

[3] Baldwin, Richard, and James Harrigan (2011), “Zeros, Quality, and Space: Trade Theory andTrade Evidence”, American Economic Journal: Microeconomics, Vol. 3, pp. 1-31

[4] Bernard, Andrew, Bradford Jenson, Stephen Redding, and Peter Schott (2007), “Firms inInternational Trade”, Journal of Economic Perspectives, Vol. 21-3, pp. 105-130.

[5] Brander, James A. and Paul Krugman, (1983) “A ’Reciprocal Dumping’ Model of InternationalTrade.” Journal of International Economics, Vol 15, pp. 313-321.

[6] Cameron, Colin, Jonah B. Gelbach, and Douglas L. Miller (2006), “Robust Inference withMulti-way Clustering”, mimeo UC Davis.

[7] Feenstra, Robert C. (1994), “New Product Varieties and the Measurement of InternationalPrices”, The American Economic Review, Vol. 84-1, pp. 157-177.

[8] Feenstra, Robert C. and David Weinstein (2010), “Globalization, Mark-ups, and the US PriceLevel”, NBER Working Paper No. 15749.

[9] Feenstra, Robert C., Robert E. Lipsey, Haiyan Deng, Alyson C. Ma, and Hengyong Mo (2005),“World Trade Flows: 1962-2000,” NBER Working Paper No. 11040

[10] Hallak, Juan Carlos (2005), “Product Quality and the Direction of Trade”, Journal of Inter-national Economics, 68-11, pp. 238-265.

[11] Hallak, Juan Carlos and Jagadeesh Sivadasan (2009), “Firms’ Exporting Behavior under Qual-ity Constraints”, mimeo University of Michigan.

[12] Helpman, Elhanan, Marc Melitz, and Yona Rubinstein (2008), “Estimating Trade Flows: Trad-ing partners and Trading Volumes”, The Quarterly Journal of Economics, Vol. 123- 2, pp.441-487

[13] Hummels, David and Alexandre Skiba (2004), “Shipping the Good Apples Out? An EmpricalConfirmation of the Alchian-Allen Conjecture”, Journal of Political Economy, Vol. 112-6, pp.1384-1401.

[14] Irwin, Douglas A., Petros Mavroidis and Alan O. Sykes (2007), “The Genesis of the GATT”,mimeo Stanford University

[15] Jackson, John H., The World Trading System, MIT Press, 1997

28

[16] Johnson, Robert, (2009) “Trade and Prices with Heterogeneous Firms”, mimeo PrincetonUniversity

[17] Kehoe, Timothy, and Kim Ruhl, (2009) “How important is the New Goods Margin in Interna-tional Trade”, mimeo University of Minnesota

[18] Ludema, Rodney and Anna Maria Mayda (2009), “Do Countries Free Ride on MFN”, Journalof International Economics, Vol. 77-2, pp. 137-150.

[19] Melitz, Marc and Gianmarco I.P. Ottaviano (2008), “Market Size, Trade, and Productivity”,Review of Economic Studies, Vol. 75-1, pp. 295-316.

[20] Manova, Kalina (2008), “Credit Constraints, Heterogeneous Firms, and International Trade”,mimeo Stanford University

[21] Pierce, Justin and Peter Schott (2009), “Concording U.S. Harmonized System Codes OverTime”, mimeo Yale University

[22] Romalis, John (2006), “Market Access, Openness and Growth”, mimeo University of Chicago

[23] Rose, Andrew (2004), “Do We Really Know that the WTO Increases Trade?”, AmericanEconomic Review, Vol. 94-1, pp. 98-114.

[24] Saggi, Kamal (2004), “Tariffs and the most favored nation clause”, Journal of InternationalEconomics, Vol 63, pp. 341-368.

[25] Saggi, Kamal, and Nese Sara (2008), “National Treatment at the WTO: The Role of Productand Country Heterogeneity”, International Economic Review, Vol. 49, pp. 1367-96.

[26] Schott, Peter (2004), “Across-product Versus Within-product Specialization in InternationalTrade”, Quarterly Journal of Economics, Vol. 119-2, pp. 647-678

[27] Subramanian, Arvind, and Shang-Jin Wei, (2007) “The WTO Promotes Trade, Strongly butUnevenly,” Journal of International Economics

[28] Trefler, Daniel (2004), “The Long and Short of the Canada-U.S. Free Trade Agreement”,American Economic Review, Vol. 94, pp. 870–895.

29

A Extensive Margin

An aspect absent from the above model is a treatment of the extensive margin. As bilateral trade

values, the object of interest in the empirical analysis, are recorded at the country level and not

the variety level, it is crucial to address the extent to which aggregation and entry of new varieties

matters for the above results. To examine these issues, I now derive an extended model in which L

exporters, indexed by l, supply to a common import market. Each exporter l is identical with the

exception of the distribution of quality-adjusted marginal costs, and the pool of potential exporting

firms, Nl. Precisely, I assume that, for exporter l, quality-adjusted costs are distributed from 0 to

cmax with Pareto shape parameter kl (> 1). The upper-bound on costs, cmax, is assumed to be non-

binding, and thus larger than the demand intercept, A. Thus, the distribution of quality-adjusted

costs for each country l is written as:

Gl

( cλ

)=

(cλ

)kl(cmax)kl

Here, higher values of kl yield a higher quality-adjusted cost, on average.

Focusing on an arbitrary exporter l, I now will derive the relationship between quality-adjusted

costs and import values. In (3), firms cannot sell unless their post-tariff price is less than or equal

to A. This will occur if(ciλi

)∈[0, A

t

]. Thus, assuming that quality-adjusted costs are distributed

according to Gl

(cλ

), the import value from country l is written as:

Vl =Nl

cklm

A2+kl

2γt1+kl(kl + 2)(16)

Similarly, total quality-adjusted quantity imported from importer l is written as:

Ql =Nl

cklm

A1+kl

2γtkl(kl + 1)(17)

Next, I will establish that Vl is decreasing in kl and then derive the effect of tariffs on import value

from exporter l. Taking logs of Vl and differentiating with respect to kl :

∂ log(Vl)

∂kl= log

(A

cmt

)− 1

kl + 2< 0

30

The inequality follows from the assumption that cm is non-binding.13 Holding the level of potential

exporters, Nl, constant, the value of imports from exporter l is decreasing in kl. Thus, unless

the number of potential exported varieties does not rise strongly with kl, those exporters with the

highest quality-adjusted cost earn the lowest revenues.14

With this relationship in mind, I now show that the results from the basic model described

above remain despite including a flexible response at the extensive margin. Precisely, taking logs

of Vl and differentiating with respect to t, the elasticity of import value with respect to tariffs for

country l is written as follows::

εv,l = kl(εA − 1)︸ ︷︷ ︸<0

+ (2εA − 1)︸ ︷︷ ︸?

(18)

where,

εA =1

1 + θη∑k∈LQlkl

∈ (0, 1)

Again εA > 0, as higher tariffs decrease competitiveness and increase the residual consumer demand

for each variety.

In (18), higher values of kl increase the negative impact of tariffs. Put differently, higher values

of kl yield a larger range of εA such that the impact of trade liberalization is negative. Precisely,

rearranging (18), εv,l < 0 if the following condition is satisfied:

εA <kl + 1

kl + 2(19)

As the right-hand side of (19) is increasing in kl, higher values of kl yield a larger range of εA such

that liberalization increasing imports from exporter l. In the limit, as kl approaches infinity, (19)

is always satisfied.

In summary, countries with a higher average quality-adjusted cost, who under reasonable con-

ditions earn lower revenues, benefit the most from the liberalization of MFN tariffs.

13cmt > cm > A implies that log(

Acmt

)< 0

14Allowing Nl to change with kl, there is still a negative relationship between kl and Vl if ∂Nlkl

∂klNl<

kl

(1

kl + 2− log

(A

cmt

))︸ ︷︷ ︸

>0

. This seems like a sensible assumption to make, since if firms need capital to enter the

foreign market, it is unlikely that more capital would be available in less efficient markets.

31

B Trade Growth Pre-1989

In Figure 2, I provided evidence suggesting that tariffs had a meaningful effect on trade, and that

trade growth before tariffs were cut was unrelated to future tariffs cuts. In Figure 3, I extend

the illustration of trade growth back to 1980 using TSUSA data from the UC Davis for Interna-

tional Data. The issue with using TSUSA is that there exists no nested concordance with the HS

classification system.

To begin, I calculate whether a TSUSA product receives a tariff cut during the Uruguay Round

according to the following procedure. Precisely, I adopt the convention that a TSUSA product

receives a tariff cut if at least one of the HS8 products that links to it receives a tariff cut. Once

I have identified which TSUSA products received a tariff cut, I merge this information with the

Exporter-TSUSA bilateral import data over the period 1980-1988.

To measure relative trade growth for observations occurring in 1989 or later, I first calculate

the log of the ratio of imports in year t to imports in 1992 for each HS10-Exporter pair. Then,

for each year, I de-mean these growth rates across Exporter-SITC4 pairs. SITC4 is used because

it has a nested link to both HS8 and TSUSA. Thus, in each year after 1989, average trade growth

relative to 1992 is equal to zero. For observations occurring before 1989, I first calculate the log of

the ratio of imports in year t to imports in 1988 for each TSUSA-Exporter pair. Then, for each

year, I de-mean these growth rates across Exporter-SITC4 pairs. Thus, in each year before 1988,

average trade growth relative to 1988 is equal to zero.

In Figure 3, I have decomposed this “zero” into HS8 products that received a tariff cut, and HS8

products that did not. Similar to Figure 2, we see that there is no obvious difference in pre-Uruguay

Round import growth rates between products that received a tariff cut over 1994-2004 and those

that didn’t.

32

Figure 3: Bilateral Trade Growth - 1980-2004

1980 1985 1990 1995 2000

-0.10

-0.05

0.00

0.05

0.10

Year

Trad

e G

row

th (1

992

Bas

e fo

r HS

, 198

8 B

ase

for T

SU

SA

)

TSUSA Data <<<>>> HS Data

HS8 Tariff Cut in 1994-2004No HS8 Tariff Cut in 1994-2004

<- Final Act Signed

33

Table 3: Effects of MFN tariffs by within-product quintile

Full Sample1 2 3 4 5 6 7 8 9 10 11 12

βD1 -0.017 -0.027 -0.017 -0.028 -0.017 -0.028 0.006 0.011 -0.024 -0.015 -0.024 -0.015

[0.058] [0.059] [0.069] [0.069] [0.069] [0.069] [0.062] [0.063] [0.079] [0.079] [0.079] [0.079]βD2 -0.336*** -0.349*** -0.314*** -0.33*** -0.314*** -0.33*** -0.275*** -0.27*** -0.275*** -0.265*** -0.275*** -0.265***

[0.064] [0.066] [0.074] [0.076] [0.074] [0.076] [0.074] [0.075] [0.081] [0.083] [0.081] [0.083]βD3 -0.612*** -0.626*** -0.594*** -0.61*** -0.594*** -0.61*** -0.558*** -0.553*** -0.541*** -0.527*** -0.541*** -0.527***

[0.07] [0.072] [0.075] [0.076] [0.075] [0.076] [0.079] [0.077] [0.078] [0.076] [0.078] [0.076]βD4 -0.925*** -0.939*** -0.884*** -0.901*** -0.884*** -0.901*** -0.852*** -0.848*** -0.828*** -0.817*** -0.828*** -0.817***

[0.095] [0.097] [0.093] [0.095] [0.093] [0.095] [0.093] [0.091] [0.09] [0.089] [0.09] [0.089]βD5 -1.288*** -1.303*** -1.223*** -1.24*** -1.223*** -1.24*** -1.243*** -1.242*** -1.207*** -1.195*** -1.207*** -1.195***

[0.104] [0.106] [0.109] [0.111] [0.109] [0.111] [0.102] [0.103] [0.123] [0.123] [0.123] [0.123]βUS -3.898** -3.294** -5.03** -4.399* -5.457*** -4.325*** -7.441*** -6.377***

[1.689] [1.637] [2.292] [2.272] [1.265] [1.17] [1.652] [1.465]βUS_D0 -5.092** -4.401** -0.062 -0.002 -5.887*** -5.065*** 1.555 1.313

[2.128] [2.111] [1.943] [1.945] [1.983] [1.952] [2.02] [1.977]βUS_D1 -5.03** -4.399* -7.441*** -6.377***

[2.292] [2.272] [1.652] [1.465]βUS_D2 -3.881** -3.387* 1.15 1.012 -5.874*** -4.777*** 1.568 1.601

[1.835] [1.73] [1.533] [1.532] [1.744] [1.743] [1.53] [1.514]βUS_D3 -4.11** -3.502** 0.921 0.897 -4.996*** -3.682** 2.446 2.695

[1.777] [1.715] [1.954] [1.972] [1.807] [1.739] [2.078] [2.047]βUS_D4 -2.911 -2.358 2.119 2.041 -4.592** -3.391 2.849 2.986

[2.04] [2.021] [1.728] [1.769] [2.114] [2.088] [2.269] [2.262]βUS_D5 -1.66 -1.054 3.37** 3.346** -3.962* -2.528 3.479 3.849*

[1.676] [1.609] [1.465] [1.475] [2.073] [1.933] [2.129] [2.108]βdVT -­‐3.751*** -3.75*** -3.75*** -1.153*** -1.154*** -1.154***

[0.771] [0.77] [0.77] [0.122] [0.122] [0.122]

Adj R2 0.038 0.044 0.038 0.044 0.038 0.044 0.036 0.05 0.036 0.051 0.036 0.051

N

Fixed

# Fixed

Clusters

Restricted Sample: Positive Trade in 1989 and 1990

1 2 3 4 5 6 7 8 9 10 11 12βD2 -0.333*** -0.334*** -0.313*** -0.32*** -0.313*** -0.32*** -0.293*** -0.296*** -0.262*** -0.268*** -0.262*** -0.268***

[0.033] [0.033] [0.046] [0.047] [0.046] [0.047] [0.046] [0.047] [0.047] [0.05] [0.047] [0.05]βD3 -0.615*** -0.615*** -0.598*** -0.602*** -0.598*** -0.602*** -0.589*** -0.589*** -0.552*** -0.55*** -0.552*** -0.55***

[0.034] [0.032] [0.055] [0.055] [0.055] [0.055] [0.054] [0.052] [0.063] [0.063] [0.063] [0.063]βD4 -0.925*** -0.92*** -0.886*** -0.884*** -0.886*** -0.884*** -0.878*** -0.872*** -0.837*** -0.832*** -0.837*** -0.832***

[0.06] [0.059] [0.058] [0.059] [0.058] [0.059] [0.068] [0.065] [0.069] [0.069] [0.069] [0.069]βD5 -1.293*** -1.286*** -1.234*** -1.228*** -1.234*** -1.228*** -1.275*** -1.266*** -1.217*** -1.202*** -1.217*** -1.202***

[0.077] [0.075] [0.08] [0.079] [0.08] [0.079] [0.083] [0.081] [0.098] [0.097] [0.098] [0.097]βUS -3.914** -3.254** -5.31** -4.49* -5.791*** -4.607*** -7.574*** -6.443***

[1.698] [1.63] [2.331] [2.318] [1.326] [1.152] [1.763] [1.608]βUS_D1 -5.31** -4.49* -7.574*** -6.443***

[2.331] [2.318] [1.763] [1.608]βUS_D2 -4.267** -3.756** 1.043 0.734 -5.92*** -4.972*** 1.654 1.471

[1.898] [1.797] [1.449] [1.446] [1.901] [1.855] [1.092] [1.177]βUS_D3 -4.484*** -3.85** 0.826 0.639 -5.628*** -4.373** 1.946 2.07

[1.738] [1.659] [1.879] [1.931] [1.888] [1.753] [2.088] [2.086]βUS_D4 -3.281 -2.649 2.029 1.84 -5.478*** -4.371** 2.096 2.072

[2.024] [1.984] [1.755] [1.784] [1.905] [1.776] [2.222] [2.198]βUS_D5 -2.2 -1.474 3.11** 3.015** -4.539** -3.047* 3.035 3.396*

[1.743] [1.638] [1.443] [1.44] [1.928] [1.715] [2.085] [2.046]βGrow -0.105*** -0.105*** -0.105*** -0.114*** -0.114*** -0.114***

[0.011] [0.011] [0.011] [0.014] [0.014] [0.014]βdVT -3.93*** -3.932*** -3.932*** -1.215*** -1.217*** -1.217***

[0.829] [0.828] [0.828] [0.141] [0.141] [0.141]

Adj R2 0.039 0.051 0.039 0.051 0.039 0.051 0.04 0.063 0.04 0.063 0.04 0.063

N

Fixed

# Fixed

ClustersNotes: The dependent variable is the log ratio of import value in 2004 to import value in 1992 for exporter j in HS10 industry i. The excluded group is D0, which is thedummy variable identifying whether exporter j in HS10 product i was not imported to the US during the years 1989 or 1990. The "Full Sample" includes all observations,including those of varieties that were not imported in 1989 or 1990. The "Restricted Sample" excludes these varieties. Fixed effects are estimated using a withinprocedure according to the groups listed in the column "Fixed". Standard errors are heteroskedasticity robust and clustered by Exporter and Fixed effect industry(Exp_HS4 fixed effects implies Exporter and HS4 clusters used for standard errors) according to the multi-level procedure in Cameron, Gelbach, and Miller (2006). Thelabels *, **, and *** identify estimates which are significantly different from zero at the 10%, 5%, and 1% levels, respectively. "Adj R2" reports adjusted R2 values forthe regression after the data has been demeaned by the fixed effects.

13970

Exporter,HS4

TARIFFS

TARIFFS

48600

Exporter_HS2

3242

Exporter,HS2

48600

Exporter_HS4

36634

Exporter_HS2

2540

Exporter,HS2

36634

Exporter_HS4

11056

Exporter,HS4

34

Table 4: Effects of MFN tariffs by within-product quintile: Exporter Characteristics

GDP-Tariff Interaction

βD1 -0.022 -0.033 -0.03 -0.02[0.069] [0.069] [0.079] [0.079]

βD2 -0.319*** -0.335*** -0.28*** -0.27*** -0.313*** -0.32*** -0.262*** -0.268***[0.074] [0.076] [0.081] [0.083] [0.046] [0.047] [0.047] [0.05]

βD3 -0.6*** -0.615*** -0.546*** -0.533*** -0.599*** -0.604*** -0.551*** -0.549***[0.076] [0.077] [0.077] [0.076] [0.055] [0.054] [0.063] [0.063]

βD4 -0.89*** -0.906*** -0.834*** -0.823*** -0.887*** -0.884*** -0.838*** -0.833***[0.094] [0.096] [0.09] [0.09] [0.058] [0.059] [0.069] [0.07]

βD5 -1.227*** -1.244*** -1.209*** -1.198*** -1.233*** -1.228*** -1.214*** -1.199***[0.109] [0.111] [0.122] [0.123] [0.08] [0.079] [0.098] [0.096]

βUS -18.127* -16.689 -18.733 -17.762 -12.918 -11.081 -16.043* -15.628*[10.268] [10.47] [12.161] [12.072] [8.205] [7.949] [8.656] [7.982]

βUS_D0 0.211 0.254 1.778 1.543[1.902] [1.9] [2.032] [1.991]

βUS_D2 1.205 1.061 1.615 1.647 1.065 0.745 1.706 1.526[1.508] [1.508] [1.516] [1.499] [1.42] [1.42] [1.086] [1.166]

βUS_D3 0.952 0.924 2.503 2.752 0.834 0.637 2.005 2.134[1.923] [1.943] [2.061] [2.03] [1.855] [1.909] [2.07] [2.07]

βUS_D4 2.147 2.064 2.863 3 2.037 1.841 2.106 2.087[1.732] [1.772] [2.273] [2.267] [1.752] [1.781] [2.232] [2.205]

βUS_D5 3.407** 3.375** 3.543* 3.905* 3.137** 3.032** 3.107 3.466*[1.45] [1.464] [2.114] [2.096] [1.431] [1.433] [2.079] [2.042]

βUS_GDPPC 1.37 1.286 1.188 1.198 0.795 0.69 0.887 0.962[1.114] [1.13] [1.243] [1.25] [0.924] [0.895] [0.875] [0.826]

βGrow -0.106*** -0.114***[0.011] [0.014]

βdVT -3.761*** -1.155*** -3.947*** -1.217***[0.77] [0.123] [0.829] [0.141]

Adj R2 0.038 0.045 0.037 0.051 0.039 0.051 0.04 0.063N

Fixed# FixedClusters

Exporter-Tariff Interaction

βD1 -0.018 -0.029 -0.032 -0.022[0.069] [0.069] [0.08] [0.08]

βD2 -0.315*** -0.33*** -0.28*** -0.269*** -0.314*** -0.32*** -0.263*** -0.269***[0.075] [0.076] [0.082] [0.084] [0.047] [0.047] [0.048] [0.05]

βD3 -0.597*** -0.61*** -0.548*** -0.533*** -0.607*** -0.609*** -0.554*** -0.552***[0.078] [0.078] [0.079] [0.077] [0.055] [0.055] [0.064] [0.063]

βD4 -0.886*** -0.899*** -0.833*** -0.821*** -0.891*** -0.886*** -0.838*** -0.832***[0.095] [0.097] [0.091] [0.091] [0.057] [0.058] [0.07] [0.07]

βD5 -1.224*** -1.237*** -1.214*** -1.201*** -1.238*** -1.228*** -1.217*** -1.201***[0.109] [0.111] [0.124] [0.125] [0.077] [0.076] [0.097] [0.095]

βUS_D0 0.143 0.166 2.081 1.854[1.931] [1.921] [2.053] [1.996]

βUS_D2 1.292 1.201 1.818 1.848 0.915 0.695 1.766 1.588[1.609] [1.597] [1.552] [1.531] [1.498] [1.475] [1.209] [1.26]

βUS_D3 1.029 1.082 2.743 3.036 0.436 0.387 2 2.148[1.936] [1.95] [2.11] [2.083] [1.828] [1.875] [2.105] [2.094]

βUS_D4 2.26 2.277 3.056 3.241 1.72 1.702 2.142 2.14[1.825] [1.875] [2.362] [2.364] [1.876] [1.901] [2.309] [2.291]

βUS_D5 3.46** 3.58** 3.553 4.007* 2.645 2.81* 2.956 3.384[1.715] [1.743] [2.208] [2.186] [1.622] [1.648] [2.126] [2.082]

βGrow -0.106*** -0.114***[0.011] [0.014]

βdVT -3.745*** -1.159*** -3.938*** -1.22***[0.772] [0.123] [0.836] [0.142]

Adj R2 0.041 0.047 0.04 0.054 0.042 0.053 0.042 0.066N

Fixed# FixedClusters

TARIFFS

TARIFFS

Full Sample

Full Sample Restricted Sample: Positive Trade in 1989 and 1990

Restricted Sample: Positive Trade in 1989 and 1990

Exporter_HS2 Exporter_HS4 Exporter_HS2 Exporter_HS4

Exporter,HS4 Exporter,HS2 Exporter,HS4

48449 48449

138673172 2519

36584

13970 2540 11056

36584

11023

48600 48600 36634 36634

Exporter,HS2

Notes: The dependent variable is the log ratio of import value in 2004 to import value in 1992 for exporter j in HS10 industry i. The excluded group is D0, which is thedummy variable identifying whether exporter j in HS10 product i was not imported to the US during the years 1989 or 1990. The "Full Sample" includes all observations,including those of varieties that were not imported in 1989 or 1990. The "Restricted Sample" excludes these varieties. Fixed effects are estimated using a within procedureaccording to the groups listed in the column "Fixed". Standard errors are heteroskedasticity robust and clustered by Exporter and Fixed effect industry (Exp_HS4 fixedeffects implies Exporter and HS4 clusters used for standard errors) according to the multi-level procedure in Cameron, Gelbach, and Miller (2006). The labels *, **, and*** identify estimates which are significantly different from zero at the 10%, 5%, and 1% levels, respectively. "Adj R2" reports adjusted R2 values for the regression afterthe data has been demeaned by the fixed effects.

Exporter_HS2 Exporter_HS4 Exporter_HS2 Exporter_HS4

Exporter,HS2 Exporter,HS4 Exporter,HS2 Exporter,HS43242

35

Tab

le5:

Eff

ects

ofM

FN

tari

ffs

by

wit

hin

-pro

duct

quin

tile

:N

oN

AF

TA

,C

HIN

A

Full S

am

ple

β D1

-0.0

15-0

.015

-0.0

06-0

.006

-0.0

36-0

.036

-0.0

09-0

.009

-0.0

25-0

.025

-0.0

03-0

.003

[0.0

73]

[0.0

73]

[0.0

85]

[0.0

85]

[0.0

72]

[0.0

72]

[0.0

83]

[0.0

83]

[0.0

77]

[0.0

77]

[0.0

89]

[0.0

89]

β D2

-0.2

87**

*-0

.287

***

-0.2

36**

*-0

.236

***

-0.3

37**

*-0

.337

***

-0.2

58**

*-0

.258

***

-0.2

95**

*-0

.295

***

-0.2

25**

-0.2

25**

[0.0

76]

[0.0

76]

[0.0

88]

[0.0

88]

[0.0

79]

[0.0

79]

[0.0

86]

[0.0

86]

[0.0

79]

[0.0

79]

[0.0

91]

[0.0

91]

β D3

-0.5

94**

*-0

.594

***

-0.5

08**

*-0

.508

***

-0.6

14**

*-0

.614

***

-0.5

05**

*-0

.505

***

-0.6

02**

*-0

.602

***

-0.4

96**

*-0

.496

***

[0.0

78]

[0.0

78]

[0.0

85]

[0.0

85]

[0.0

8][0

.08]

[0.0

83]

[0.0

83]

[0.0

84]

[0.0

84]

[0.0

89]

[0.0

89]

β D4

-0.8

43**

*-0

.843

***

-0.7

62**

*-0

.762

***

-0.9

13**

*-0

.913

***

-0.8

19**

*-0

.819

***

-0.8

58**

*-0

.858

***

-0.7

65**

*-0

.765

***

[0.0

93]

[0.0

93]

[0.0

95]

[0.0

95]

[0.1

][0

.1]

[0.0

92]

[0.0

92]

[0.1

02]

[0.1

02]

[0.1

01]

[0.1

01]

β D5

-1.2

11**

*-1

.211

***

-1.1

53**

*-1

.153

***

-1.2

71**

*-1

.271

***

-1.2

03**

*-1

.203

***

-1.2

55**

*-1

.255

***

-1.1

87**

*-1

.187

***

[0.1

04]

[0.1

04]

[0.1

22]

[0.1

22]

[0.1

14]

[0.1

14]

[0.1

23]

[0.1

23]

[0.1

07]

[0.1

07]

[0.1

2][0

.12]

β US

-3.8

09-6

.796

***

-4.2

65*

-6.3

02**

*-3

.747

-6.7

66**

*[2

.7]

[1.9

91]

[2.3

2][1

.523

][2

.717

][2

.089

]β U

S_D0

-3.9

79**

-0.1

7-5

.565

***

1.23

1-4

.145

**0.

119

-5.1

94**

*1.

108

-3.7

69*

-0.0

22-5

.679

***

1.08

7[1

.993

][2

.243

][1

.965

][2

.045

][2

.098

][1

.95]

[1.8

94]

[1.9

91]

[1.9

93]

[2.2

91]

[1.9

08]

[2.1

57]

β US_D1

-3.8

09-6

.796

***

-4.2

65*

-6.3

02**

*-3

.747

-6.7

66**

*[2

.7]

[1.9

91]

[2.3

2][1

.523

][2

.717

][2

.089

]β U

S_D2

-3.1

25*

0.68

5-5

.711

***

1.08

4-3

.178

*1.

087

-4.6

94**

*1.

608

-2.8

87*

0.86

-5.3

3***

1.43

6[1

.705

][1

.924

][1

.949

][1

.669

][1

.682

][1

.532

][1

.639

][1

.445

][1

.735

][1

.893

][1

.908

][1

.635

]β U

S_D3

-3.9

3**

-0.1

21-4

.728

***

2.06

8-3

.253

*1.

012

-3.2

54*

3.04

8-3

.684

**0.

063

-4.3

39**

2.42

7[1

.657

][2

.245

][1

.631

][2

.464

][1

.719

][1

.991

][1

.815

][2

.179

][1

.705

][2

.316

][1

.805

][2

.703

]β U

S_D4

-2.0

191.

79-3

.492

3.30

4-2

.245

2.02

-3.4

81*

2.82

1-1

.814

1.93

3-3

.298

3.46

8[2

.065

][2

.27]

[2.3

47]

[2.6

95]

[2.0

15]

[1.8

3][2

.071

][2

.239

][2

.166

][2

.39]

[2.4

1][2

.723

]β U

S_D5

-1.9

731.

836

-3.5

863.

21-1

.363

2.90

1*-2

.751

3.55

1*-2

.397

1.35

-3.9

48*

2.81

8[1

.908

][1

.831

][2

.186

][2

.505

][1

.51]

[1.5

][1

.858

][2

.06]

[1.8

54]

[1.9

23]

[2.1

53]

[2.5

15]

β dVT

-3.6

14**

*-3

.614

***

-1.1

17**

*-1

.117

***

-3.8

21**

*-3

.821

***

-1.1

71**

*-1

.171

***

-3.6

67**

*-3

.667

***

-1.1

35**

*-1

.135

***

[0.7

59]

[0.7

59]

[0.1

21]

[0.1

21]

[0.7

82]

[0.7

82]

[0.1

22]

[0.1

22]

[0.7

66]

[0.7

66]

[0.1

21]

[0.1

21]

Ad

j R

20.

044

0.04

40.

052

0.05

20.

046

0.04

60.

051

0.05

10.

042

0.04

20.

049

0.04

9N

Fixed

# F

ixed

Clu

sters

Rest

rict

ed

Sam

ple

: P

osi

tive

Tra

de in

19

89

an

d 1

99

0

β D2

-0.2

91**

*-0

.291

***

-0.2

44**

*-0

.244

***

-0.3

16**

*-0

.316

***

-0.2

62**

*-0

.262

***

-0.2

88**

*-0

.288

***

-0.2

28**

*-0

.228

***

[0.0

64]

[0.0

64]

[0.0

62]

[0.0

62]

[0.0

45]

[0.0

45]

[0.0

5][0

.05]

[0.0

64]

[0.0

64]

[0.0

67]

[0.0

67]

β D3

-0.6

03**

*-0

.603

***

-0.5

52**

*-0

.552

***

-0.5

89**

*-0

.589

***

-0.5

14**

*-0

.514

***

-0.5

95**

*-0

.595

***

-0.5

24**

*-0

.524

***

[0.0

58]

[0.0

58]

[0.0

7][0

.07]

[0.0

57]

[0.0

57]

[0.0

71]

[0.0

71]

[0.0

6][0

.06]

[0.0

76]

[0.0

76]

β D4

-0.8

48**

*-0

.848

***

-0.8

09**

*-0

.809

***

-0.8

84**

*-0

.884

***

-0.8

28**

*-0

.828

***

-0.8

44**

*-0

.844

***

-0.7

93**

*-0

.793

***

[0.0

66]

[0.0

66]

[0.0

71]

[0.0

71]

[0.0

68]

[0.0

68]

[0.0

73]

[0.0

73]

[0.0

77]

[0.0

77]

[0.0

77]

[0.0

77]

β D5

-1.2

24**

*-1

.224

***

-1.1

97**

*-1

.197

***

-1.2

38**

*-1

.238

***

-1.2

02**

*-1

.202

***

-1.2

44**

*-1

.244

***

-1.2

08**

*-1

.208

***

[0.0

88]

[0.0

88]

[0.1

06]

[0.1

06]

[0.0

86]

[0.0

86]

[0.0

97]

[0.0

97]

[0.0

98]

[0.0

98]

[0.1

1][0

.11]

β US

-3.9

14-6

.604

***

-4.3

83*

-6.5

64**

*-3

.776

-6.7

55**

*[2

.763

][2

.192

][2

.371

][1

.682

][2

.791

][2

.251

]β U

S_D1

-3.9

14-6

.604

***

-4.3

83*

-6.5

64**

*-3

.776

-6.7

55**

*[2

.763

][2

.192

][2

.371

][1

.682

][2

.791

][2

.251

]β U

S_D2

-3.4

24*

0.49

-5.2

12**

*1.

392

-3.5

52**

0.83

1-4

.96*

**1.

604

-3.2

06*

0.57

-4.9

6**

1.79

5[1

.778

][1

.903

][2

][1

.541

][1

.762

][1

.411

][1

.811

][1

.092

][1

.802

][1

.89]

[1.9

75]

[1.5

33]

β US_D3

-4.3

04**

*-0

.39

-5.0

84**

*1.

521

-3.4

53**

0.93

-3.7

02**

2.86

2-3

.968

**-0

.192

-4.6

06**

2.14

9[1

.663

][2

.227

][1

.745

][2

.547

][1

.666

][1

.94]

[1.7

95]

[2.2

35]

[1.6

55]

[2.2

9][1

.847

][2

.75]

β US_D4

-2.2

921.

621

-4.2

14**

2.39

-2.4

961.

888

-4.4

11**

2.15

3-2

.044

1.73

3-4

.064

*2.

69[2

.045

][2

.308

][2

.075

][2

.706

][1

.98]

[1.8

87]

[1.7

62]

[2.2

08]

[2.1

04]

[2.4

28]

[2.0

78]

[2.7

43]

β US_D5

-2.4

211.

493

-3.9

83**

2.62

1-1

.573

2.81

1*-3

.119

*3.

445*

-2.7

221.

054

-4.3

22**

2.43

3[1

.967

][1

.8]

[1.9

84]

[2.4

92]

[1.5

34]

[1.5

][1

.693

][2

.054

][1

.886

][1

.917

][1

.975

][2

.52]

β Grow

-0.1

04**

*-0

.104

***

-0.1

13**

*-0

.113

***

-0.1

02**

*-0

.102

***

-0.1

11**

*-0

.111

***

-0.1

***

-0.1

***

-0.1

1***

-0.1

1***

[0.0

11]

[0.0

11]

[0.0

15]

[0.0

15]

[0.0

11]

[0.0

11]

[0.0

14]

[0.0

14]

[0.0

12]

[0.0

12]

[0.0

16]

[0.0

16]

β dVT

-3.7

92**

*-3

.792

***

-1.1

83**

*-1

.183

***

-3.9

81**

*-3

.981

***

-1.2

36**

*-1

.236

***

-3.8

21**

*-3

.821

***

-1.2

02**

*-1

.202

***

[0.8

25]

[0.8

25]

[0.1

39]

[0.1

39]

[0.8

34]

[0.8

34]

[0.1

4][0

.14]

[0.8

31]

[0.8

31]

[0.1

38]

[0.1

38]

Ad

j R

20.

048

0.04

80.

061

0.06

10.

051

0.05

10.

064

0.06

40.

048

0.04

80.

062

0.06

2N

Fixed

# F

ixed

Clu

sters

T A R I F F S T A R I F F S

4528

8

Expo

rter

,HS2

Expo

rter

,HS4

Expo

rter

,HS4

4528

847

081

4708

143

760

4376

0

3079

2978

Expo

rter

_HS2

Expo

rter

_HS4

Expo

rter

,HS2

Expo

rter

,HS4

Expo

rter

,HS2

Expo

rter

,HS4

2454

Expo

rter

,HS4

3153

1323

2

1046

424

0710

575

Expo

rter

_HS2

Expo

rter

_HS4

3425

134

251

3554

6

Expo

rter

_HS4

Expo

rter

,HS2

Expo

rter

_HS4

Expo

rter

_HS2

Expo

rter

,HS2

Expo

rter

,HS4

Expo

rter

,HS2

Expo

rter

_HS2

3554

633

157

1334

9

2312

9979

3315

7Ex

port

er_H

S2

Expo

rter

_HS4

No

tes:

The

depe

nden

tva

riab

leis

the

log

ratio

ofim

port

valu

ein

2004

toim

port

valu

ein

1992

for

expo

rter

jin

HS10

indu

stry

i.Th

eex

clud

edgr

oup

isD

0,w

hich

isth

edu

mm

yva

riab

leid

entify

ing

whe

ther

expo

rter

jin

HS10

prod

uct

iwas

not

impo

rted

toth

eU

Sdu

ring

the

year

s19

89or

1990

.Th

e"F

ullS

ampl

e"in

clud

esal

lobs

erva

tion

s,in

clud

ing

thos

eof

variet

ies

that

wer

eno

tim

port

edin

1989

or19

90.

The

"Res

tric

ted

Sam

ple"

excl

udes

thes

eva

riet

ies.

Fixe

def

fect

sar

ees

tim

ated

usin

ga

withi

npr

oced

ure

acco

rdin

gto

the

grou

pslis

ted

inth

eco

lum

n"F

ixed

".Sta

ndar

der

rors

are

hete

rosk

edas

tici

tyro

bust

and

clus

tere

dby

Expo

rter

and

Fixe

def

fect

indu

stry

(Exp

_HS4

fixed

effe

cts

impl

ies

Expo

rter

and

HS4

clus

ters

used

for

stan

dard

erro

rs)

acco

rdin

gto

the

mul

ti-l

evel

proc

edur

ein

Cam

eron

,G

elba

ch,

and

Mill

er(2

006)

.Th

ela

bels

*,**

,an

d**

*id

entify

estim

ates

whi

char

esi

gnifi

cant

lydi

ffer

ent

from

zero

atth

e10

%,

5%,

and

1%le

vels

,re

spec

tive

ly.

"Adj

R2"

repo

rts

adju

sted

R2

valu

es for

the

reg

ress

ion

afte

r th

e da

ta h

as b

een

dem

eane

d by

the

fix

ed e

ffec

ts.

No

Ch

ina

No

NA

FTA

No

Ch

ina,

NA

FTA

No

Ch

ina

No

NA

FTA

No

Ch

ina,

NA

FTA

1259

9Ex

port

er_H

S2

Expo

rter

_HS4

36

Table 6: Effects of MFN tariffs by within-product quintile: Exporter HHI, and Number of Varieties

HS8 Exporter Concentration HS8 Variety

Low HHI High HHI Low HHI High HHI Low N High N Low N High NβLow -0.087 -0.105 0.076 0.037

[0.083] [0.101] [0.093] [0.109]βD1 -0.018 -0.068 -0.011 -0.051 -0.233** 0.028 -0.291* 0.05

[0.093] [0.092] [0.087] [0.122] [0.113] [0.087] [0.166] [0.087]βD2 -0.281*** -0.449*** -0.217** -0.404*** -0.594*** -0.248*** -0.546*** -0.194**

[0.083] [0.101] [0.094] [0.112] [0.115] [0.079] [0.112] [0.093]βD3 -0.621*** -0.604*** -0.529*** -0.553*** -0.724*** -0.583*** -0.671*** -0.5***

[0.091] [0.093] [0.085] [0.11] [0.111] [0.087] [0.114] [0.086]βD4 -0.846*** -1.039*** -0.778*** -0.956*** -1.132*** -0.836*** -1.076*** -0.763***

[0.106] [0.119] [0.093] [0.123] [0.138] [0.102] [0.155] [0.093]βD5 -1.144*** -1.488*** -1.097*** -1.506*** -1.363*** -1.222*** -1.422*** -1.156***

[0.126] [0.11] [0.137] [0.13] [0.115] [0.12] [0.153] [0.133]βUS_D0 -4.684** -3.631 -5.077** -4.499* -2.715 -5.127** -3.017 -5.876**

[2.379] [2.763] [2.449] [2.649] [2.505] [2.485] [2.581] [2.785]βUS_D1 -4.866* -3.536 -6.905*** -5.09** -3.744 -5.182* -5.222* -7.436***

[2.712] [2.704] [2.086] [2.292] [2.903] [2.78] [3.049] [2.213]βUS_D2 -2.193 -5.488** -3.285 -7.558*** -7.131*** -1.89 -7.904*** -3.737

[2.157] [2.769] [2.698] [2.794] [2.337] [2.148] [2.917] [2.632]βUS_D3 -4.523*** -1.603 -4.088*** -2.498 -3.729 -3.553* -5.021* -3.097

[1.645] [2.375] [1.566] [3.053] [2.627] [1.813] [2.742] [2.319]βUS_D4 -1.807 -3.441 -3.312 -3.421 -5.024** -1.342 -5.546** -2.666

[2.355] [2.266] [2.22] [2.791] [2.492] [1.968] [2.686] [2.488]βUS_D5 0.125 -3.549 -1.006 -5.826** -2.014 -0.734 -5.903*** -1.199

[2.399] [2.31] [3.063] [2.388] [2.425] [2.315] [2.209] [2.988]βdVT -3.741*** -3.769*** -1.153*** -1.163*** -3.842*** -3.739*** -1.195*** -1.142***

[0.771] [0.789] [0.126] [0.124] [0.774] [0.763] [0.128] [0.126]Adj R2

NFixed

# FixedClusters

Low HHI High HHI Low HHI High HHI Low N High N Low N High NβLow -0.006 -0.08 -0.112 -0.283

[0.111] [0.119] [0.134] [0.178]βD2 -0.283*** -0.393*** -0.216*** -0.386*** -0.383*** -0.295*** -0.251* -0.269***

[0.065] [0.078] [0.064] [0.101] [0.103] [0.062] [0.13] [0.065]βD3 -0.613*** -0.578*** -0.542*** -0.572*** -0.523*** -0.632*** -0.426** -0.589***

[0.056] [0.091] [0.066] [0.11] [0.135] [0.065] [0.178] [0.076]βD4 -0.834*** -0.99*** -0.793*** -0.936*** -0.93*** -0.872*** -0.825*** -0.844***

[0.072] [0.081] [0.085] [0.101] [0.105] [0.079] [0.154] [0.095]βD5 -1.136*** -1.447*** -1.102*** -1.483*** -1.16*** -1.266*** -1.148*** -1.234***

[0.098] [0.108] [0.123] [0.13] [0.138] [0.089] [0.145] [0.12]βUS_D1 -5.377* -2.914 -7.428*** -4.319 -3.395 -5.444* -5.03 -7.599***

[2.778] [2.768] [2.19] [2.65] [2.637] [2.962] [3.189] [2.59]βUS_D2 -3.052 -5.02* -3.814 -7.165** -6.814*** -2.568 -7.269** -4.224

[2.319] [2.772] [2.702] [3.065] [2.248] [2.318] [3.023] [2.692]βUS_D3 -4.916*** -1.939 -5.068*** -2.794 -3.526 -4.162** -5.006 -4.101*

[1.591] [2.299] [1.643] [3.091] [2.6] [1.779] [3.154] [2.331]βUS_D4 -2.187 -3.615 -4.33** -4.351* -5.496** -1.494 -6.843*** -3.455

[2.376] [2.213] [1.941] [2.637] [2.41] [1.951] [2.617] [2.143]βUS_D5 -0.544 -3.564 -1.805 -5.815** -2.19 -1.23 -6.043** -1.787

[2.474] [2.331] [2.878] [2.524] [2.457] [2.429] [2.394] [2.747]βGrow -0.102*** -0.114*** -0.109*** -0.13*** -0.105*** -0.107*** -0.127*** -0.111***

[0.011] [0.023] [0.017] [0.028] [0.03] [0.01] [0.027] [0.016]βdVT -3.941*** -3.91*** -1.219*** -1.215*** -4.088*** -3.912*** -1.284*** -1.196***

[0.829] [0.846] [0.145] [0.141] [0.834] [0.822] [0.141] [0.148]Adj R2

NFixed

# FixedClusters

0.046 0.053

0.053 0.066

13970Exporter,HS4

Restricted Sample

0.051

Exporter,HS2 Exporter,HS4

48600Exporter_HS2

3242Exporter,HS2

48600Exporter_HS4

2540 11056

Full Sample

0.051

TARIFFS

36634 36634Exporter_HS2 Exporter_HS4

Exporter,HS2

36634

0.045

Exporter_HS4366340.064

Exporter_HS2

11056

TARIFFS

48600

Exporter_HS22540

Notes: The dependent variable is the log ratio of import value in 2004 to import value in 1992 for exporter j in HS10 industry i. The excluded groups are D0,which is the dummy variable identifying whether exporter j in HS10 product i was not imported to the US during the years 1989 or 1990, and the dummyvariable identifying above median HHI or N. The "Full Sample" includes all observations, including those of varieties that were not imported in 1989 or 1990.The "Restricted Sample" excludes these varieties. Fixed effects are estimated using a within procedure according to the groups listed in the column "Fixed".Standard errors are heteroskedasticity robust and clustered by Exporter and Fixed effect industry (Exp_HS4 fixed effects implies Exporter and HS4 clusters usedfor standard errors) according to the multi-level procedure in Cameron, Gelbach, and Miller (2006). The labels *, **, and *** identify estimates which aresignificantly different from zero at the 10%, 5%, and 1% levels, respectively. "Adj R2" reports adjusted R2 values for the regression after the data has beendemeaned by the fixed effects.

Full Sample

Restricted SampleExporter,HS2 Exporter,HS4

48600Exporter_HS4

3242 13970

Exporter,HS4

37

Tab

le7:

Com

par

ison

ofV

alue,

Quan

tity

,an

dP

rice

bin

s

Full S

am

ple

Bin

Defi

nit

ion

:β D

1-0

.09

-0.0

76-0

.051

-0.0

23-0

.426

***

-0.3

93**

*-0

.083

-0.0

64-0

.008

0.00

4-0

.368

**-0

.348

**[0

.133

][0

.135

][0

.132

][0

.135

][0

.144

][0

.147

][0

.132

][0

.132

][0

.133

][0

.131

][0

.153

][0

.151

]β D

2-0

.393

***

-0.3

83**

*-0

.348

***

-0.3

23**

-0.5

11**

*-0

.486

***

-0.3

31**

-0.3

09**

-0.2

92**

-0.2

76**

-0.4

07**

*-0

.387

***

[0.1

36]

[0.1

37]

[0.1

25]

[0.1

27]

[0.1

28]

[0.1

31]

[0.1

35]

[0.1

36]

[0.1

29]

[0.1

27]

[0.1

35]

[0.1

31]

β D3

-0.6

64**

*-0

.653

***

-0.5

18**

*-0

.493

***

-0.5

95**

*-0

.571

***

-0.5

89**

*-0

.566

***

-0.4

63**

*-0

.448

***

-0.5

22**

*-0

.509

***

[0.1

49]

[0.1

5][0

.119

][0

.121

][0

.143

][0

.145

][0

.147

][0

.145

][0

.123

][0

.121

][0

.151

][0

.148

]β D

4-0

.977

***

-0.9

65**

*-0

.862

***

-0.8

37**

*-0

.561

***

-0.5

37**

*-0

.922

***

-0.8

99**

*-0

.775

***

-0.7

61**

*-0

.507

***

-0.4

95**

*[0

.16]

[0.1

62]

[0.1

33]

[0.1

36]

[0.1

43]

[0.1

46]

[0.1

51]

[0.1

5][0

.137

][0

.134

][0

.142

][0

.139

]β D

5-1

.345

***

-1.3

31**

*-1

.178

***

-1.1

52**

*-0

.566

***

-0.5

43**

*-1

.329

***

-1.3

04**

*-1

.173

***

-1.1

58**

*-0

.532

***

-0.5

24**

*[0

.172

][0

.173

][0

.168

][0

.172

][0

.139

][0

.142

][0

.183

][0

.183

][0

.163

][0

.16]

[0.1

31]

[0.1

3]β U

S_D0

-3.1

84-2

.994

-2.9

87-2

.89

-2.9

02-2

.805

-5.1

42-4

.051

-5.5

95-4

.275

-5.4

23-4

.101

[2.3

16]

[2.3

57]

[2.6

56]

[2.6

22]

[2.6

91]

[2.6

62]

[3.5

11]

[3.4

36]

[3.4

49]

[3.3

69]

[3.5

44]

[3.4

57]

β US_D1

-3.7

87-3

.135

-2.6

2-1

.945

-3.2

12*

-2.4

53-7

.192

***

-6.2

34**

*-5

.746

***

-4.7

22**

*-4

.934

*-3

.599

[2.3

24]

[2.3

12]

[2.2

12]

[2.1

86]

[1.9

43]

[1.8

93]

[2.0

13]

[1.8

23]

[1.9

81]

[1.8

31]

[2.8

44]

[2.6

52]

β US_D2

-2.8

45-2

.336

-2.5

67-2

.003

-1.6

27-1

.054

-4.7

5**

-3.5

64-4

.541

-3.3

25-3

.425

*-2

.217

[2.1

59]

[2.0

71]

[2.1

78]

[2.1

1][2

.242

][2

.205

][2

.396

][2

.363

][2

.878

][2

.675

][2

.056

][1

.917

]β U

S_D3

-3.0

3*-2

.4-2

.946

*-2

.311

-1.2

71-0

.681

-4.0

43*

-2.6

52-5

.493

***

-4.2

32**

-3.7

11*

-2.4

95[1

.783

][1

.743

][1

.772

][1

.689

][1

.719

][1

.666

][2

.212

][2

.035

][2

.01]

[1.8

68]

[2.1

76]

[2.0

18]

β US_D4

-2.0

03-1

.394

-3.4

44*

-2.8

59-2

.573

-1.9

63-4

.194

*-2

.878

-4.4

89*

-3.2

27-5

.666

**-4

.402

**[2

.073

][2

.058

][2

.024

][2

.009

][1

.705

][1

.677

][2

.391

][2

.321

][2

.437

][2

.349

][2

.318

][2

.233

]β U

S_D5

-1.1

86-0

.482

-1.4

4-0

.748

-3.4

69-2

.864

-4.0

38*

-2.5

05-4

.212

*-2

.683

-4.8

61*

-3.6

36[1

.558

][1

.447

][1

.751

][1

.67]

[2.2

5][2

.259

][2

.412

][2

.2]

[2.2

11]

[2.0

1][2

.556

][2

.489

]β d

VT

-3.7

83**

*-3

.717

***

-3.7

07**

*-1

.265

***

-1.2

69**

*-1

.259

***

[0.8

31]

[0.8

31]

[0.8

18]

[0.1

57]

[0.1

59]

[0.1

61]

Ad

j R

20.

040.

046

0.03

0.03

60.

008

0.01

40.

040.

055

0.03

10.

046

0.00

70.

021

NFi

xed

# F

ixed

Clu

sters

Rest

rict

ed

Sam

ple

: P

osi

tive

Tra

de in

19

89

an

d 1

99

0B

in D

efi

nit

ion

:β D

2-0

.309

***

-0.3

13**

*-0

.243

***

-0.2

4***

-0.0

7-0

.071

-0.2

48**

*-0

.251

***

-0.2

49**

*-0

.253

***

-0.0

26-0

.03

[0.0

51]

[0.0

53]

[0.0

3][0

.037

][0

.053

][0

.053

][0

.05]

[0.0

55]

[0.0

75]

[0.0

74]

[0.0

7][0

.071

]β D

3-0

.578

***

-0.5

82**

*-0

.417

***

-0.4

11**

*-0

.149

***

-0.1

53**

*-0

.517

***

-0.5

17**

*-0

.425

***

-0.4

25**

*-0

.14*

*-0

.154

**[0

.064

][0

.065

][0

.06]

[0.0

63]

[0.0

55]

[0.0

56]

[0.0

67]

[0.0

7][0

.075

][0

.075

][0

.066

][0

.068

]β D

4-0

.887

***

-0.8

84**

*-0

.752

***

-0.7

43**

*-0

.099

-0.1

09-0

.83*

**-0

.821

***

-0.7

28**

*-0

.722

***

-0.1

11-0

.136

*[0

.069

][0

.071

][0

.075

][0

.078

][0

.065

][0

.067

][0

.083

][0

.084

][0

.088

][0

.087

][0

.074

][0

.076

]β D

5-1

.254

***

-1.2

44**

*-1

.065

***

-1.0

51**

*-0

.1-0

.11

-1.2

26**

*-1

.208

***

-1.1

06**

*-1

.095

***

-0.1

57-0

.181

*[0

.086

][0

.084

][0

.103

][0

.105

][0

.09]

[0.0

93]

[0.1

13]

[0.1

13]

[0.1

1][0

.11]

[0.0

99]

[0.1

02]

β US_D1

-4.0

31*

-3.2

78-3

.939

*-3

.497

-3.2

4*-2

.727

-7.5

02**

*-6

.434

***

-6.9

17**

*-5

.911

***

-4.8

24-3

.537

[2.3

95]

[2.3

85]

[2.2

05]

[2.1

96]

[1.9

53]

[1.9

1][2

.021

][1

.857

][2

.215

][2

.002

][2

.95]

[2.6

59]

β US_D2

-2.9

27-2

.457

-2.6

33-2

.07

-1.5

21-0

.931

-4.7

64*

-3.7

56-4

.349

-3.2

98-2

.876

-1.7

72[2

.148

][2

.037

][2

.206

][2

.105

][2

.285

][2

.249

][2

.487

][2

.41]

[3.0

53]

[2.7

89]

[2.0

01]

[1.8

43]

β US_D3

-3.1

53*

-2.6

21-3

.225

*-2

.733

-1.3

91-0

.852

-4.6

21**

-3.4

08*

-6.0

41**

*-5

.03*

**-3

.641

*-2

.57

[1.8

29]

[1.7

64]

[1.7

51]

[1.6

66]

[1.6

33]

[1.5

87]

[2.2

62]

[2.0

46]

[1.9

19]

[1.7

19]

[2.0

44]

[1.8

66]

β US_D4

-2.1

86-1

.633

-3.5

76*

-3.0

22-2

.451

-1.9

77-4

.535

**-3

.306

-4.8

18**

-3.6

49-5

.566

**-4

.603

**[2

.007

][1

.978

][1

.985

][1

.966

][1

.678

][1

.637

][2

.258

][2

.102

][2

.397

][2

.291

][2

.382

][2

.242

]β U

S_D5

-1.2

32-0

.497

-1.5

24-0

.834

-3.8

95*

-3.3

06-4

.111

*-2

.524

-4.4

67**

-2.9

66-5

.696

**-4

.576

*[1

.583

][1

.458

][1

.741

][1

.631

][2

.233

][2

.185

][2

.454

][2

.19]

[2.1

74]

[1.8

83]

[2.7

][2

.612

]β G

row

-3.7

83**

*-3

.713

***

-3.7

64**

*-1

.275

***

-1.2

75**

*-1

.26*

**[0

.888

][0

.889

][0

.874

][0

.163

][0

.162

][0

.163

]β d

VT

-0.1

05**

*-0

.11*

**-0

.113

***

-0.1

17**

*-0

.122

***

-0.1

29**

*[0

.011

][0

.011

][0

.011

][0

.015

][0

.015

][0

.017

]A

dj

R2

0.03

80.

050.

027

0.03

80.

001

0.01

30.

040.

062

0.03

0.05

20.

002

0.02

5N

Fixed

# F

ixed

Clu

sters

Quantity

3341

1Ex

port

er_H

S2

T A R I F F S T A R I F F S

Not

es:

The

depe

nden

tva

riab

leis

the

log

ratio

ofim

port

valu

ein

2004

toim

port

valu

ein

1992

for

expo

rter

jin

HS10

indu

stry

i.Th

eex

clud

edgr

oup

isD

0,w

hich

isth

edu

mm

yva

riab

leid

entify

ing

whe

ther

expo

rter

jin

HS10

prod

uct

iwas

not

impo

rted

toth

eU

Sdu

ring

the

year

s19

89or

1990

.Th

e"F

ullSam

ple"

incl

udes

allob

serv

atio

ns,

incl

udin

gth

ose

ofva

riet

ies

that

wer

eno

tim

port

edin

1989

or19

90.

The

"Res

tric

ted

Sam

ple"

excl

udes

thes

eva

riet

ies.

Fixe

def

fect

sar

ees

tim

ated

usin

ga

withi

npr

oced

ure

acco

rdin

gto

the

grou

pslis

ted

inth

eco

lum

n"F

ixed

".Sta

ndar

der

rors

are

hete

rosk

edas

tici

tyro

bust

and

clus

tere

dby

Expo

rter

and

Fixe

def

fect

indu

stry

(Exp

_HS4

fixed

effe

cts

impl

ies

Expo

rter

and

HS4

clus

ters

used

for

stan

dard

erro

rs)

acco

rdin

gto

the

mul

ti-l

evel

proc

edur

ein

Cam

eron

,G

elba

ch,

and

Mill

er(2

006)

.Th

ela

bels

*,**

,an

d**

*id

entify

estim

ates

whi

char

esi

gnifi

cant

lydi

ffer

ent

from

zero

atth

e10

%,

5%,

and

1%le

vels

,re

spec

tive

ly.

"Adj

R2"

rep

orts

adj

uste

d R2

valu

es for

the

reg

ress

ion

afte

r th

e da

ta h

as b

een

dem

eane

d by

the

fix

ed e

ffec

ts.

Value

Quantity

Price

Expo

rter

,HS2

3341

1Ex

port

er_H

S4

1028

9Ex

port

er,H

S4

2739

Value

2904

9Ex

port

er_H

S2

2298

Expo

rter

_HS2

8844

Expo

rter

_HS4

Value

Quantity

Price

Value

Quantity

Price

Price

2904

9

Expo

rter

_HS4

38