Product Heterogeneity, Cross-Country Taste Differences, and

European Economic Review 100 (2017) 1–27
Contents lists available at ScienceDirect
European Economic Review

journal homepage: www.elsevier.com/locate/euroecorev
Product heterogeneity, cross-country taste differences, and

the growth of world tradeR
Raphael A. Auer a,b,∗
a
Bank for International Settlements, Basel, Switzerland
b
CEPR, London, UK
a r t i c l e i n f o a b s t r a c t
Article history: This paper extends the analysis of the “home market” effect in Krugman (1980) to a flex-
Received 28 September 2015 ible demand structure and examines the dynamic effects of trade liberalization. I first de-
Accepted 9 May 2017
velop a model in which consumers are heterogeneous in their valuations of product at-
Available online 15 July 2017
tributes and firms offer goods of heterogeneous attribute levels. With international trade
in the presence of cross-country taste differences, consumption is home-biased in the im-
JEL classification:
F12 mediate aftermath of liberalization. Once industries specialize, the volume of trade grows
F15 and so do the gains from liberalization. In the long-run equilibrium with open markets, the
L15 volume of trade is diminished by the existence of cross-country taste differences only if
L16 countries specialize completely. I then show that the adaptation of industrial composition
to the demand structure of the European common market was associated with growing
Keywords:
within-European trade in the automotive industry.
Intra-industry trade
Monopolistic competition © 2017 Elsevier B.V. All rights reserved.
Increasing returns
Home-market effect
Product heterogeneity
Industrial structure
Firm dynamics
European economic integration
1. Introduction
Since it is the prediction that most clearly distinguishes the “new” trade theory from classical motives for trade, the
“home market” effect of Krugman (1980) has received ample attention in the empirical literature. Davis and Weinstein
(1999), Davis, 2003 and Keith and Ries (2001) show that production increases more than one-for-one with local demand
for a good, while Feenstra et al. (2001) show how increasing returns can be distinguished from other motives for trade in
a gravity framework. Hanson and Xiang (2004) extend the two-sector prediction of Krugman (1980) to a multi-industry
setup and then derive and test a relative home market prediction relating demand and sector characteristics to net trade
R
An earlier version of this paper was titled “Product Heterogeneity, Within-Industry Trade Patterns, and the Home Bias of Consumption.” I thank seminar
participants at the Institute for Advanced Studies, the MWIEG meetings at Pennsylvania State University, the Federal Reserve Bank of New York, Princeton
University, the University of Geneva, the University of Vienna, the SNB/CEPR Conference on Quality and Product Heterogeneity in International Trade, and
at the 2013 CEPR European Research Workshop in International Trade in Rotterdam for helpful comments and discussions. Part of this research project
was conducted while the author was on leave at Princeton University. I thank the Niehaus Center for Globalization and Governance for support. The views
expressed in this paper represent those of the author and not necessarily those of the Bank for International Settlements.
∗
Correspondence to: Bank for International Settlements, Centralbahnplatz 2, 4051 Basel, Switzerland.
E-mail address: raphael.auer@bis.org
http://dx.doi.org/10.1016/j.euroecorev.2017.05.010
0014-2921/© 2017 Elsevier B.V. All rights reserved.
2 R.A. Auer / European Economic Review 100 (2017) 1–27
flows. Aizenman and Brooks (2008), focusing on the case of beer and wine, present direct evidence for the existence of
pronounced cross-country differences in demand patterns (see also Atkin (2013)).1
The home market effect in Krugman (1980) is, however, derived under the restrictive assumption of demand being
unit-elastic. In his setting, expenditures thus do not react to the change in relative prices and the relative supply of varieties
that is brought about by trade liberalization. Against the backdrop of the widespread empirical evidence that expenditures
do strongly react to such changes (see, for example, Imbs and Mejean, 2015 for a recent study and survey of the literature),
it is worthwhile to extend the analysis of the home market effect relaxing the assumption of unit-elastic demand.
This paper contributes to the literature on international trade by extending the analysis of the home market effect to
a flexible demand structure. With this extension, the model provides two novel insights regarding under what conditions
and to which extent cross-country taste differences affect the volume of and the gains from trade. First, compared to the
benchmark of countries having an identical distribution of preferences, cross-country taste differences lead to consumption
being home-biased in the immediate aftermath of liberalization. This is so as each country’s industry structure is optimized
for the preferences of domestic consumers and domestic output thus does not match well with preferences abroad.
Second, cross-country taste differences also have implications for the dynamic impact of a trade liberalization. This is so
as national industries adapt only gradually to the demand structure of a globalized market. With increasing specialization,
the aggregate volume of trade rises due to the extensive margin as firms with low export revenue are replaced by high
export revenue firms that produce goods better suited to foreign demand. This adaptation process of national industries
to the demand structure of international markets highlights that cross-country taste differences can explain the sluggish
response of trade volume to liberalization (documented for example by Yi, 2003; 2010, Ruhl, 2008, and Hummels, 2007).
I conclude by providing evidence that these considerations have indeed been significant determinants of trade flows in
the European automotive industry. I uncover taste differences along the product attribute dimension relying on external
estimates of the elasticity of demand and using the structure of the model. I then show that after the creation of the
European common market, the adaptation of industrial composition to the demand structure of the common market was
associated with growing within-European trade in the automotive industry.
In emphasizing the importance of shifts in the sectoral composition of the economy that are driven by demand patterns
rather than comparative advantage, this paper relates to the literature on structural transformation and international trade
(see Matsuyama, 1992 and more recently Uy et al., 2013) and also to the literature on dynamic Hecksher–Ohlin models
of trade. Regarding the latter literature, the mechanism of the model presented in this paper is closely related to Cunat
and Maffezzoli (2007) analysis of how the dynamic response of factor accumulation amplifies the initial growth effect of a
marginal trade liberalization in the presence of increasing returns.
In highlighting the importance of taste differences for the volume of trade, this paper connects the literature examining
home market effects in homothetic preference frameworks with the literature examining Linder’s (1961) hypothesis that in
the presence of non-homothetic preferences, income differences impede the volume of trade (see, among others, Markusen,
1986; Baier and Bergstrand, 2001; Bergstrand, 1990; Fajgelbaum et al., 2011; Föllmi et al., 2008; Hallak, 2010; Matsuyama,
20 0 0, and Fieler, 2011). In particular the analogy of the cases of complete and incomplete specialization in Fajgelbaum
et al. (2011) and in this study documents that a common effect is present in these two preference frameworks. In both,
the response of industry composition to cross-country differences in consumption decisions increases the volume of trade:
while for a fixed domestic industry composition, a lower fraction of consumers who value a certain good is associated with
a lower volume of imports of this good, the reverse holds true once the industry composition has adjusted to liberalization.
This is so as with trade, lower domestic valuation for an attribute is associated with an over-proportional reduction in
domestic production of goods embodying the attribute, and consequently, higher import volume of such goods.2
The structure of this paper is the following. Section 2 presents the model and also analyzes the equilibrium of the
closed economy. The immediate aftermath of liberalization is analyzed in Section 3, while Section 4 examines the long-run
equilibrium. Section 5 examines the role of cross-country taste differences for the growth of automobile trade during and
after the creation of the European common market. Section 6 concludes.
2. A model of demand for heterogeneous products
This section develops a model featuring consumers with homothetic preferences and heterogeneous tastes. The aim
is to extend the setting of Krugman (1980), who assumes that each consumer only consumes one type of good and that
expenditures for each type of good are thus fixed exogenously, to the case in which consumers differ in their propensity to
consume one type of good or the other, yet consumption decisions react to the availability of varieties and their respective
relative prices.
1
See also Davis (1998), Weder, 2003), Holmes and Stevens (2005), Crozet and Trionfetti (2008), Behrens et al. (2009), and Brülhart and Trionfetti (2009).
Keith et al. (2002) examine whether the home market effect arises under alternative preference frameworks that feature increasing returns, but do not
rely on the demand function of Dixit and Stiglitz (1977). Anderson and van Wincoop (2004) sketch a preference setup featuring heterogeneous tastes and
verbally analyze how differences in preferences might create implicit trade barriers, but they do not solve for the equilibrium of the sketched model.
2
It is noteworthy that whereas (Linder, 1961, p. 94) certainly stresses the importance of income heterogeneity as the main driver of taste differences, his
argument that “[t]he more similar is the demand structure of two countries, the more intensive, potentially, is the trade between these two countries” also
extends to the case of homothetic taste differences analyzed in this paper.
R.A. Auer / European Economic Review 100 (2017) 1–27 3
To this end, this section develops a model of consumer preferences combining two motives of consumption decisions.
The first motive is two-sided heterogeneity in consumer preferences and good attributes. These take the form av, where a is
the attribute value of a good and v is the consumer’s valuation for the attribute. Differences in attributes a reflect product
characteristics such as the good’s color (measured in wavelength). Similarly, differences in valuations v reflect differences
in people’s tastes for the attribute. For example, some consumers might have a preference for cars painted in Ferrari Red,
while others prefer British Racing Green.
The second motive is the love of variety motive of Dixit and Stiglitz (1977): consumers also value variety, i.e., they prefer
an economy featuring many different varieties of cars painted in British Racing Green to an economy featuring only one such
variety. In the model, each consumer is endowed with a consumer-firm specific utility draw x. Since having a larger number
of draws raises the expected maximum draw, consumer welfare rises with the number of available varieties. I next lay out
the functional forms used to model these intuitions, derive a firm’s demand, and describe the supply side of the economy.
2.1. Preferences
The world is composed of two countries named Home and Foreign, that are respectively populated by a mass of L and
L∗ consumers. Each consumer has preferences over a homogenous O (outside) good and over a finite set of differentiated
M (manufacturing) varieties. Each M firm produces exactly one differentiated variety that is characterized by its attribute
a. Each consumer has a valuation v for the attribute a and is also characterized by an idiosyncratic and consumer-firm
specific utility draw x.
Throughout the analysis, let i I index consumers (individuals) and j J index manufacturing firms. Each consumer i is
endowed with income θi = θ in terms of labor and a valuation draw vi {vL , vH }, where it holds that vL < vH . I denote the
fraction of the population that has a valuation draw of vi = vH by π H [0, 1]. The remainder fraction (1 − πH ) is endowed
with valuation draw vL .
Independently of her valuation draw vi , each consumer is also endowed with a consumer-firm specific draw xi, j for each
firm in j J.
Consumers care about the valuation- and idiosyncratic draw-adjusted effective quantity of the manufacturing M good
and the absolute quantity of the outside good O. Denoting the quantity consumer i consumes of the O good by oi and the
quantity she consumes from manufacturing firm j by qi, j , consumer i’s utility Ui is given by
α

Ui = (oi )1−α qi, j exi, j +a j vi . (1)
j J
Note that in (1), the taste shifter xi, j + a j vi – that could technically be subsumed in a single firm-customer specific
parameter – is split in two separate parts to emphasize the two distinct underlying economic intuitions: whereas xi, j is a
continuously distributed random variable giving rise to the “love of variety” motive and a well-defined demand elasticity for
each firm’s output, aj and vi are fixed for each firm or consumer respectively, thus giving rise to the systematic matching
of high aj firms with high vi consumers.
The consumer’s decision is subject to non-negativity for oi and for all qi, j . Her decision is also subject to her budget
constraint

oi pO + qi, j p j ≤ θi . (2)
i I
Differences in attributes a can be seen as differences product characteristics, while differences in valuations v reflect
differences in consumers’ tastes for the attribute. For example, while German consumers might prefer powerful cars made
for driving on the Autobahn, French consumers might have a preference for small agile cars that fit into a narrow Parisian
parking space. Consumers also value variety, i.e., they prefer an economy featuring many different fast car varieties to an
economy featuring only one such variety.
The utility function (1) implies that for all consumers, all manufacturing goods are perfectly substitutable. However,
different consumers have different rates of substitution between different varieties; in equilibrium, therefore, certain types
of consumers are more or less likely to buy certain types of goods.
Consider first only the term ea j vi in (1). The key feature of this term in the preferences is that the rate at which
consumers value (or dislike) the attribute differs between consumers with different vi . Assume that two otherwise identical
consumers of valuations vL and vH > vL are offered to buy a certain good aL at price pL or a good aH at price pH where
aH > aL . What is the maximum price difference between pL and pH at which each consumer would prefer the high a good?
For the H-valuation consumer, this would be price ratio pH /pL = evH (aH −aL ) , while it would be pH /pL = evL (aH −aL ) for the
L-valuation consumer. Because higher valuation consumers value the attribute more, in equilibrium, they constitute the
relatively larger group of consumers of H-attribute goods.
Valuation shifters vi can be interpreted much broader than reflecting intrinsic differences in consumer preferences. For
example, they can be viewed as reflecting differences in infrastructures that shape the utility certain attributes deliver.
For example, it is obvious that on average, German consumers with access to highways without speed limits value a car’s
top speed more than do drivers in other nations who face strictly enforced limits far below the top speed of most cars.
Valuation shifters can also reflect national differences in attribute-specific regulation or taxation, such as for example tax
schemes that directly tax product attributes like engine capacity.
Next, consider only the term exi, j in (1).3 xi, j is a consumer-firm specific shock, reflecting the fact that some consumers
like or dislike the variety of a specific firm irrespective of the variety’s attribute. In (1), the idiosyncratic taste shock intro-
duces market power to the model: although firms cannot observe xi, j , they can engage in first-degree price discrimination by
charging a higher price and only attracting consumers with high xi, j draws. Throughout the analysis, I assume that xi, j is dis-
tributed maximum Gumbel (or Type I extreme value distribution) with scale and shape parameters 0 and 1/σ , respectively.

Gx xi, j = exp − exp −xi, j σ (3)
The consumer-firm specific shocks are orthogonal to firm attributes or consumer valuations and are independent across
firms and consumers: xi, j ⊥xi, n for n = j. Gabaix et al. (2006) demonstrate that these assumptions, in combination with a
utility function similar to (1) but without the attribute and valuation heterogeneity, yield an ideal-variety micro foundation
for the constant elasticity of substitution (CES) demand system of Dixit and Stiglitz (1977).4
2.2. Demand and consumer welfare
This subsection solves for a firm’s demand and consumer welfare. Consumer i consumes the agricultural O good and the

manufacturing composite Mi ≡ qi, j exi, j +a j vi . Before considering the choice among the single manufactured goods, consider
j J
first the decision of how much of the O good to consume. The first-order conditions of the utility function (1) with respect
to these two quantities and the budget constraint (2) imply that an agent with income 1 consumes
Mi = (1 − α )/pM,i and Oi = α /pO ,
where pM, i is the price of the manufacturing composite for consumer i (pM, i is heterogeneous across consumers).
Irrespective of this price, the consumer always spends a fraction α of her income on the O good.
Thus, the consumer spends the remainder of her income on the manufacturing composite. Within the manufacturing
composite, since all goods are perfect substitutes, each consumer then chooses the variety that yields the highest ratio of
effective quantity per unit divided by the price of the variety. Since consumers with different valuation vi differ in the rate
at which they substitute goods of different attributes a, demand differs by v. It is given by the following proposition
Proposition 1 (Demand). The demand D(aj , pj ) of a firm with attribute aj and price pj is equal to

exp[σ vL a j ] exp[σ vH a j ]
D ( a j , p j ) = ( 1 − α )θ Lp−j (1+σ ) ( 1 − πH ) −σ + πH −σ (4)
P ( vL ) P ( vH )
where (...) is the gamma function and P (vi ) denotes the ideal price index for all consumers with vi = v, which is given by
−1/σ
pn
− σ
P ( vi ) ≡ . (5)
exp[vi an ]
n J
Proof. See Appendix
The proof of Proposition 1 builds on existing research demonstrating how the love of variety motive can arise in a
discrete choice setting: each consumer has a consumer-variety specific taste shock xi, j . For equal prices and good attributes,
the consumer chooses the maximum of all the realizations of the taste shocks xi, j , i.e., she chooses j = arg maxxi, j . Owing
j J
to the functional form assumption that the idiosyncratic taste shocks are distributed Gumbel with shape parameter 1/σ , all
firms face a constant elasticity of demand equal to −(1 + σ ).
Compared to the existing literature, the novel ingredient in the derivation of firm demand (4) is that the probability
of consumer i with valuation vi = v buying from firm j with attribute a j = a depends on the match of v and a, as well as
on how well the other goods in the economy match with the consumer’s taste, as summarized by the ideal price index of
consumers with vi = v, P (v ). Since the expected maximum draw is increasing in the number of draws, consumers prefer
having a larger number of varieties to choose from, i.e., they love variety. A key feature of the preferences developed here
is that consumer welfare is highly comparable to the one in Dixit and Stiglitz (1977).
3
The micro foundation of the love for variety motive in (1) is similar to the discrete choice settings of McFadden (1981), Anderson et al. (1987) and
Anderson et al. (1992), and Gabaix et al. (2010).
4
It is noteworthy that the closed-form assumption on the consumer-firm specific taste shocks (3) is not very restrictive, since in equilibrium consumers
buy only from the attribute-adjusted maximum realization of xi, j . Since the economy features a large number of firms, the distribution of this maxima
converges to the Type I extreme value distribution for a wide set of underlying distributions. Gabaix et al. (2010) analyze the conditions under which a
random distribution of idiosyncratic taste shocks converges to that specified in Eq. (3).
Corollary 1 (Expected consumer welfare). Denote the expected welfare of consumer i with vi = v and income θ i by E(Ui |v, θ i .).
If pO = 1,
E (Ui ) = (P (v ) )−α θi

where the ideal price index P (v ) is as defined in (5) and is a constant derived in the Appendix.
Proof. See Appendix
Corollary 1 is very convenient: the developed preference structure allows to directly map changes in the toughness of
competition for all consumers with vi = v into welfare changes for this group of consumers. As I document below, with
open markets, the interplay of the free entry conditions at home and abroad pins down the ideal relative price indices for
different v’s uniquely, hence leading to very sharp predictions regarding the welfare effects of trade.
Further, one can directly relate the findings regarding the gains from trade in this paper to the existing literature. In the
case where all firms produce the same good (an = a j = a), the valuation-attribute match in (4) cancels out and the demand
curve is the same as in Dixit and Stiglitz (1977).
The model at hand, therefore, is a generalization of the Dixit and Stiglitz (1977)framework and, consequently, the
international economy described below includes the Krugman (1980) model as a special case with extreme product
heterogeneity, which is convenient since it allows clearly highlighting the impact of such heterogeneity.5
2.3. Supply
In each country and at each moment in time, a large mass of potential entrepreneurs can enter the M industry by
paying a fixed cost of F labor units. When entering the industry, each entrepreneur can choose with what type of attribute
to enter the industry. After paying the entry cost F and deciding with what kind of good to enter the industry, the
entrepreneur j receives the blueprint to produce a new variety of the manufacturing good with attribute aj . While aj can
be chosen at the moment of entry, it cannot be changed thereafter. The entrepreneur has a perpetual monopoly over that
specific variety from the moment of entry onwards and faces an exogenous probability of firm death of δ > 0.
For expositional clarity, I restrict the universe of potential levels the attribute can take and assume that aj {aL , aH },
where 0 < aL < aH . I refer to the two attribute levels as the H-attribute or L-attribute “good”, “firm”, or “variety” in the
remainder of the paper. While alive, each firm can produce any quantity of its good at constant marginal costs (in units of
labor)6 equal to
c j = c. (6)
The outside good O is produced in a competitive sector at a marginal cost of one unit of labor. In total, the Home
economy thus has to satisfy the resource constraint that domestic production of the O and M sector and entry into the M
sector do not use more than θ L units of Home labor.
If markets are opened to trade, manufacturing firms can sell abroad at a cost c∗j = τ c j , where τ > 1. The analysis also
allows for export market access being subject to a one-time access costs T . I assume that T < τ −σ F so that in any long-run
equilibrium with open markets, there are firms that export. In contrast, if markets are open to trade, trade in the outside O
good is costless.
Throughout the analysis, I assume that the interest rate ρ ≥ 0 is given exogenously.
2.4. Equilibrium in the closed economy
This subsection solves the closed economy equilibrium. A starting observation is that demand (4) is such that firms face
a constant price elasticity of (1 + σ ) and thus charge a price of p j = 1+ σ 7
σ c j . The below analysis derives most of its results
insights based on the notion that consumers with different valuations are different enough so that they prefer, on average,
different types of attributes. Formally, this notion is equivalent to the following parameter restriction.
Assumption 1 (Separating valuations). The valuation pair vL and vH is said to be separating iff
vL < 0 < vH
Separating valuations imply that there exists both a group of consumers that prefers L-attribute goods as well as a group
that prefers H-attribute ones, which is a necessary condition for an equilibrium with positive entry of both type of firms.
5
Of course, compared to the framework of Dixit and Stiglitz (1977), a limitation of the proposed model is that it describes only average consumer
welfare, i.e., all predictions that follow below hold on average but not for each consumer. Depending on the realization of idiosyncratic draws, the welfare
of any single consumer may be higher or lower than predicted in Corollary 1.
6
The working paper version derives the equilibrium for the case in which marginal costs cj depend on the attribute of the produced good.
7
This statement only holds under the assumption that each firm is small and does not take into account its influence on the ideal price index. I assume
that this is the case throughout the analysis.
Letting N denote the total number of active firms in the industry and nH denote the fraction of these firms producing a
good with a j = aH . Domestic revenue (aj ) is equal to8
πH Leσ vH a j (1 − πH )Leσ vL a j
(a j ) = + (7)
N ( nH eσ vH aH + ( 1 − nH )eσ vH aL ) N ( nH eσ vL aH + ( 1 − nH )eσ vL aL )
Since valuations are separating, eσ vH aH > eσ vH aL , and H-attribute firms sell more to H-valuation consumers than do
L-valuation firms. Similarly, eσ vL aL > eσ vL aH and L firms sell more to L-valuation consumers. Sales to each group are
proportional to the number of consumers (there are Lπ H H-valuation consumers) and increasing in the ideal price indices
P(vH ) and P(vL ) (see Eq. (5)).
Given constant markup-pricing, firm profits are proportional to revenue. In the closed economy, this revenue depends
on the distribution of consumer valuations. For any given attribute, a higher proportion of H-valuation consumers implies
a larger market size for H-attribute firms.
In the existing literature that is based on Dixit and Stiglitz (1977), due to the constant elasticity demand structure, entry
of new competitors hurts the sales of all existing firms in the same proportion. In the preferences at hand, the effect of
such an increase in competition on a firm’s sales is different for different types of firms. The revenue (7) of a firm reacts
∂ (a ) ∂ (a )
more to entry of firms producing a similar good than to entry of firms producing a dissimilar good, i.e., | ∂ N H | > | ∂ N L |
H H
∂ (a ) ∂ (a )
and | ∂ N H | < | ∂ N L |. The latter feature implies that industries are partially segmented: for example, the sales of BMW
L L
depend much more on the product strategy of Mercedes rather than the one of Toyota, which caters to a slightly different
set of consumers. Similarly, Armani’s sales depend much more on the success of the latest collections by Prada than they
do depend on the success of the collections by Louis Vuitton or Hermes.
With demand being pinned down, it is straightforward to derive entry in the closed economy. Denoting the value that
a variable takes in the autarky steady state by an A superscript, the following holds.
Proposition 2 (Autarky equilibrium). Denote by NA the autarky equilibrium number of firms in and by nAH [0, 1] the autarky
equilibrium fraction of entrepreneurs producing the H-attribute good. There exists a unique autarky equilibrium featuring N =
L
σ (δ +ρ )F and
⎧ σv a σv a
⎪
⎨0 if πH < eσ vL aH eσ vH aHeeσ vHL aHL −e
−e H L
σ vH aL eσ vL aH
eσ vL aL σv a
nAH = σ vL aL −eσ vL aH πH − (1 − πH ) eσ vHeaH −e
H L
σ vH aL otherwise . (8)
⎪
⎩
e
σv a σ vL aH
1 if πH > eσ vH aH eσ vH aH eeσ vLL aLL −e
−eσ vH aL eσ vL aH
Proof. Since firms are free to enter with an H- or L-attribute good, an equilibrium with positive entry of both types of firms
requires that the flow of revenues are equal for both H-attribute and L-attribute firms, i.e.

πH eσ vH aH (1 − πH )eσ vL aH πH eσ vH aL (1 − πH )eσ vL aL
L + =L + , (9)
P (vH )−σ P (vL )−σ P (vH )−σ P (vL )−σ
where P(vH ) and P(vL ) are the ideal price indices of H- and L-valuation consumers, equal to P (vH )−σ =
N (nH eσ vH aH + (1 − nH )eσ vH aL ) and P (vL )−σ = N (nH eσ vH aH + (1 − nH )eσ vH aL ). Thus, reformulating (9) as the difference in sales
to H-valuation and L-valuation consumers yields
πH eσ vH aH − eσ vH aL eσ vL aL − eσ vL aH
= . (10)
1 − πH nH eσ vH aH + (1 − nH )eσ vH aL nH eσ vL aH + (1 − nH )eσ vL aL
Since eσ vH aH > eσ vH aL , the LHS of (10) is increasing in relative entry of H firms nH . Since eσ vL aL > eσ vL aH the RHS is decreas-
ing in nH . Thus, nH is uniquely determined. NA depends on the flow of instantaneous profits which have to be discounted
at rate (δ + ρ ) and pin down the number of firms by the free entry condition F = L
A.
σ (δ +ρ )N
It is noteworthy that in general equilibrium, as long as nAH [0, 1], nAH is increasing in the number of H-valuation
∂ nA ∂ nA ∂ nA
consumers ( ∂πH > 0) and also that nAH is increasing in both valuations vL and vH ( ∂vH > 0, ∂vH > 0). Furthermore, denoting
H L H
the autarky equilibrium ideal price indices by PA (vj ), it is true that whenever nAH is interior
φπH NA φ (1 − πH )NA
P A (vH )−σ = σ v a and P A (vL )−σ = σ v a (11)
e L L − eσ vL aH e H H − eσ vH aL
where φ ≡ eσ vH aH eσ vL aL − eσ vH aL eσ vL aH > 0. The fact that P A (vH )−σ and P A (vL )−σ are linear in π H and (1 − πH ), respectively,
is a direct consequence of the fact that firms can decide with what kind of product to enter the industry. Therefore, a
higher π H has to be offset exactly by an increase in nH so that firms with different attributes operate at the same level of
8
In all of the below analysis, to keep notation compact for the ease of shorter notation, I assume that θ (1 − α ) = 1. While this assumption is not a mere
normalization, it is innocuous in the sense that defining θ (1 − α ) = and carrying through the analysis is technically identical to assuming that labor
supply at home and abroad is multiplied by the same factor, which will proportionally shift the number of firms in all countries.
profits, i.e. in the closed economy, the level of competition for H- and L-valuation consumers is proportional to the number
of customers π H and (1 − πH ), respectively.
A trivial (yet crucial for explaining the volume of trade in Section 3) observation is that all firms have equal revenue
((aH ) = (aH ) = LA ) for any level of π H : in equilibrium, the sales of a firm do not depend on the relative distribution of
N
consumer tastes.
Summarizing, the equilibrium in the closed economy has the following properties. First, a necessary condition for an
equilibrium featuring both kinds of firms is that there exists both a group of consumers that prefers L goods as well as
a group that prefers H goods. Second, in an equilibrium featuring positive entry of both types of firms, the fraction of
H-attribute firms is increasing in the number of H-valuation consumers. The fraction of such firms is also increasing in vH
and vL , since an increase in either valuation leads to higher relative expenditures on H-attribute goods. Third, in equilibrium,
owing to the free entry condition, all firms have the same revenue and profit flows.
3. The immediate aftermath of liberalization
Cross-country taste differences create implicit trade barriers because industries are adapted to the local taste. If, for
example, German cars are designed to satisfy the German consumer’s preference for driving fast on the Autobahn, while
French cars are designed to fit into a narrow Parisian parking space, the volume of trade is low because the typical foreign
car is not “appropriate” for the taste of the typical domestic consumer. This section analyzes the immediate aftermath of
an unanticipated liberalization during which the industry structure is still determined by autarky demand conditions.
If markets are opened to trade, exporting the manufactured good is subject to a one-time export access cost T. The
manufacturing firms that have paid this cost can export their good at iceberg trade cost τ > 1. In contrast, trade in the
outside O good is costless.
In the immediate aftermath of an unanticipated trade liberalization, the industry structure is still determined by autarky
demand conditions. I here analyze how a liberalization impacts the economy in the short run if the two countries differ
in the fraction of H- and L-valuation consumers and contrast this to the static effects of liberalization without such taste
heterogeneity.
At the instant of opening markets to trade, the number of firms is at its autarky level (8). Since accessing the export
market is subject only to small fixed costs (T τ −σ < F ), all firms export9 and there are N A nAH producers of the H-attribute
good exporting from Home to Foreign and N A∗ nAH∗ producers of the H-attribute good exporting from Foreign to Home.
Denoting the values that variables take immediately at the moment of opening to trade by S and S∗ superscripts, the
aggregate volume of H-attribute exports (denoted by XHS ) is equal to

τ −σ e σ v H a H ∗ ∗τ
−σ σ v L a H
e
XHS = N A nAH πH∗ L∗ + ( 1 − π )L .
P S∗ (vH ) H
P S∗ (vL )
A symmetric condition holds for Foreign’s exports to Home. In each country, the price indices are affected by import
competition. Since all firms export, Home’s exports are more H-attribute intensive than is the domestic production in
Foreign. Trade, therefore, intensifies competition relatively more in the sector with relatively fewer foreign consumers.
Lemma 1 (Liberalization and short run relative competition). Assume that πH > πH∗ and nAH , nAH∗ [0, 1]. When opening markets
to trade, competition in Home intensifies more in the L-attribute segment of the industry than in the H-attribute segment, while
competition in Foreign intensifies more in the H-attribute segment of the industry than the L-attribute segment. It holds that
P S ( vH ) P A ( vH ) P S∗ (vH ) P A∗ ( vH )
> and < .
P S ( vL ) P A ( vL ) P S∗ (vL ) P A∗ ( vL )
Proof. See Appendix
Lemma 1 is indicative of why differences in the distribution of tastes across Home and Foreign reduce the short-run
aggregate volume of trade. The aggregate volume of Home’s exports is equal to the number of H-attribute firms times
exports per such firm plus the number of L-attribute firms times exports per such firm. Since trade intensifies competition
in Foreign relatively more in the H sector, each Home H-attribute exporter sells a smaller amount than she would in an
economy without product heterogeneity. In contrast, each Home L-exporter sells a larger amount that she would in an
economy without product heterogeneity.
Since ideal price indices can be mapped one-to-one into welfare changes, Lemma 1 thus implies that when countries
differ in their distributions of tastes, it is the relatively smaller group of consumers that gains relatively more from trade at
the moment of liberalization. This result is intuitive: if markets are opened to trade, a French consumer with a preference
for large cars suddenly gains access to many German large car varieties. In contrast, a French consumer with a preference
for small and fuel-efficient cars gains relatively little, since Germany offers few of these varieties compared to the French
9
This prediction of the model contrasts with the empirical finding that there is export selection in almost all differentiated goods industries (see e.g.
Bernard and Jensen, 1999).
industry. What is the direction of trade in the short run after liberalization? The following proposition summarizes the
prevailing patterns of trade.
Lemma 2 (Short run attribute content of trade). Assume that parameters are such that nAH and nAH∗ [0, 1]. At the moment after
trade liberalization, if L = L∗ , Home is a net exporter of H-attribute goods iff πH > πH∗ . If L = L∗ Home’s manufacturing exports
contain a larger fraction of H-attribute goods than do Foreign’s exports.
Proof. See Appendix
Lemma 2 documents net trade patterns after liberalization, which is not to be confused with the home market effect
(which is about trade patterns after the industry has adjusted to trade). Turning to the aggregate volume of trade XS (also
measured in terms of the numeraire), which is composed of H- and L-attribute goods and equal to

τ −σ LL∗ πH πH∗ τ −σ LL∗ (1 − πH ) 1 − πH∗
X S = XHS + XLS = + . (12)
LπH + τ −σ L∗ πH∗ L(1 − πH ) + τ −σ L∗ 1 − πH∗
If there are no differences in the distribution of valuations in Home and Foreign (πH∗ = πH ), the volume of Home’s ex-
−σ ∗
ports is equal to Lτ+τ −LL
σ L∗ for any value of π H . It is straightforward to verify that for any level of πH , in the above equation,
∗
the volume of trade is indeed maximized when πH = πH∗ , i.e., the volume of trade is lower if there are cross-country taste
differences than if there are not. The following proposition summarizes the extent to which taste differences impede the
volume of trade.
Proposition 3 (Short run trade volume). Assume that parameters are such that nAH , nAH∗ [0, 1]. At the moment after trade liber-
alization, the following holds. If πH∗ = πH , the volume of trade is the same as in the absence of consumer heterogeneity and home
is a net exporter of M goods iff L > L∗ . If πH = πH∗ , the volume of trade is lower than in the absence of consumer heterogeneity.
Regarding the volume of trade,

I. For given τ −σ , the volume of trade is decreasing in πH − πH∗ .
II. The impact of taste heterogeneity on the volume of trade is decreasing in τ σ .
Proof. See Appendix
In the short run, the composition of the domestic industry thus does not conform to the demand of the average foreign
consumer. In autarky, each country develops an industry that is suited best to the tastes of the local consumer. In the short
run after opening to trade, the country’s export bundle is inappropriate for the taste distribution of foreign consumers.10
Summarizing, two major trade patterns arise immediately after opening markets to trade. First, if countries are of
unequal size, the larger country becomes the net exporter of manufactured goods, while the smaller country becomes the
net exporter of agricultural goods. Second, owing to the differences in countries’ average tastes, trade volume is lower than
what one would observe in Krugman’s (1980) model. I next examine whether and to what extent these predictions hold
when the industry structure is allowed to adjust to the changed demand patterns after trade liberalization.
4. The long-run effects of trade liberalization
This section analyzes the resulting long-run equilibrium in which the industry structure has adapted to the demand
structure of the integrated market. It then analyzes the extent to which cross-country taste differences affect the volume
of trade in the long run. With separately additive and homothetic preferences featuring the same elasticity of substitution
across all market segments and in the absence of comparative advantage, cross-country taste differences may actually not
affect the volume of trade at all. They do, however, affect the volume of trade during the transition to the open economy
steady state – examined in Section 4.2 – and further the long-run volume of trade in the case when countries specialize
completely, a case examined in Section 4.3.
4.1. The case of incomplete specialization
After trade liberalization, the industrial structure of the two countries diverges in order to restore the free entry
condition. This is a direct consequence of the fact that after opening to trade, in Home, competition is relatively tougher in
the market for L-attribute goods than in the market for H-attribute goods (see Lemma 1). Although export possibilities are
better in the market for H-attribute goods than in the market for L-attribute goods, overall, producers of H-attribute goods
make strictly higher profits because the home market is relatively more important than the foreign market. Consequently,
all newly entering firms choose to enter the industry with the H-attribute good until the new equilibrium is reached.
10
In addition to cross-country differences in tastes, there could also be a simple “distrust” in foreign goods that can explain why consumption is home-
biased. Evans (2001) compares the local sales of foreign affiliates of U.S. multinational enterprises to data on U.S. bilateral exports and domestic sales by
host-country firms, finding that the effect of such distrust is negligible. Similarly, Cosar et al. (2015) quantify preference for home-produced goods in the
automobile industry.
But how far does specialization go? An interesting result is that steady state sales per firm are not affected by the
underlying taste differences across nations because the number of firms in each market segment adjusts such that all firms
have sales that are proportional to their entry costs. Denoting the value that a variables takes in the long-run equilibrium
with open markets by a T superscript, the following holds.
Lemma 3 (Domestic revenue in the open economy). If parameters are such that in the long-run equilibrium of the open econ-
T ∗ [0, 1], it is true that the domestic revenue (a ) of a firm is independent of the type of the good it produces, i.e.
omy nTH , nH j
(aH ) = (aL ) and ∗ (aH ) = ∗ (aL )

Proof. See Appendix
Lemma 2 states a surprising result: regardless of the relative size of countries and the distribution of the preferences
in the other nation – as long as there is incomplete specialization – domestic revenue is equalized across the H and L
segments in each country.
The intuition underlying Lemma 2 is trivial. If a Home H-attribute firm has higher sales at Home than a Home L-attribute
firm, there must be some offsetting advantage for L-attribute firms in the Foreign market for the free entry condition in
Home to hold. Moreover, because the foreign market matters relatively less than the domestic market due to the existence
of transportation costs, the free attribute entry condition in Home requires that the offsetting advantage for L-attribute
firms in Foreign must be larger than the advantage for H-attribute firms at Home. In contrast, the same argument from
a Foreign perspective requires that the Home advantage for H-attribute firms is relatively stronger than the L-attribute
advantage in Foreign. Together, the two free attribute entry conditions can only be satisfied if there is no advantage for
either type of firm in either market.11
The sales per firm, in turn, pin down the number of firms. If πH > πH∗ , it is true that
nTH ≥ nAH ≥ nAH∗ ≥ nH

T∗
holds for any τ −σ > 0.12 Furthermore, under incomplete specialization, which is more likely if countries are more equal in
tastes (higher |πH − πH∗ |) or if countries are of more unequal size,13 the total number of entering firms is equal to
L − τ −σ L ∗ L ∗ − τ −σ L
NT = and N T ∗ = .
(1 − τ −σ)(T + F )σ (δ + ρ ) (1 − τ )(T + F )σ (δ + ρ )
− σ
As is summarized by the following proposition, specialization also has implications for net attribute content of trade, as
well as for the way in which cross-county taste differences affect the aggregate volume of trade.
Proposition 4 (Trade and its net attribute content). Assume that πH > πH∗ and that parameters are such that nTH , nH
T ∗ [0, 1].
−σ ∗
Then, the volume of Home’s manufacturing exports is equal to Lτ+τ −LL
σ L∗ for any πH , πH that satisfy the stated assumptions. If L =
∗ ∗
L, Home has 0 net exports of manufactured goods. If L > L∗ , Home is a net exporter of manufactured goods. For any combination
of L and L∗ that is consistent with the stated assumptions, Home’s manufacturing exports are more H-attribute intensive than
Foreign’s manufacturing exports. Home’s manufacturing exports are more H-attribute intensive in the open economy equilibrium
than just after trade liberalization.
Proof. See Appendix.
In models following Krugman (1980), a country with a larger home market for manufacturing goods becomes the net
exporter of industrial output after markets are opened to trade. As is demonstrated in the Appendix, this effect is also
present in the model at hand. Further, net exports of the manufacturing M good can be nonzero even in the case of equal
country sizes. In this case, the direction of net exports is the following: if πH + πH∗ > 1, i.e., if the global market for the type
of good that Home’s exports are concentrated in is large, Home is a net importer of manufacturing goods. If πH + πH∗ > 1
there are more H-valuation consumers than L-valuation consumers in the world (since L = L∗) and accordingly, there are
also more H-attribute firms in the world than L-attribute good ones. Global competition is thus tougher in the H segment
of the industry, which happens to be the segment home’s exports are concentrated in. Similarly, competition is less tough
in the market segment Foreign’s exports are concentrated in. Thus, home’s overall manufacturing exports are smaller than
its manufacturing imports from foreign.
11
To formalize this insight, denote the difference in domestic sales at home between a H-attribute good and a L-attribute good firm by Z (Z = (aH ) −
(aL )) and the same difference in foreign by Z∗ (Z ∗ = ∗ (aH ) − ∗ (aL )). The free attribute-entry condition at home implies that Z + τ −β Z ∗ = 0, while the
same condition in Foreign is τ −β Z + Z ∗ = 0. Only Z = Z ∗ = 0 can satisfy the free attribute-entry conditions both at Home and in Foreign.
T∗
τ −σ (τ −σ + NNT )
12
The open-economy fraction of H-type firms in home and foreign is equal to nTH = min(1, nAH + (πH − πH∗ ) 1−τ −2σ
) and nTH∗ = max(0, nHA∗ − (πH −
NT
τ −σ (τ −σ + ) eσ vL aL σv a
πH∗ ) 1−τ −2σ
NT ∗
), where ≡ eσ aL vL −eσ aH vL
+ eσ aH evH −e
H L
σ aL vH .
σa v σa v
13
The parameter restrictions for are that πH > πH∗ > eσ vH aL e L L −e
incomplete specialization φ
H L
(1 − τ −σ LL∗ ) + πH τ −σ LL∗ and that πH <
eσ aL vL −eσ aH vL
eσ vH aH (1 − τ −σ LL ) + πH∗ τ −σ LL . Note that the condition that countries do not specialize in the O sector is less restrictive than the condi-
∗ ∗
φ
tion that nTH and nTH∗ ]0, 1[ if πH = πH∗ . If πH = πH∗ the parameter restriction τ −σ < LL∗ < τ σ ensures the existence of manufacturing firms in both
nations.
In the model at hand, a further within-industry home market arises that is closely related to Hanson and Xi-
ang (2004) and Fajgelbaum et al. (2011) relative notion: a home market effect can arise in the type of manufactured products.
The main result of Proposition 4 regards how taste differences affect the volume of trade: the home bias of consumption
vanishes as long as specialization is incomplete. This is intriguing: whereas in the short run, differences in the composition
of industry are associated with low trade volume, in the long run, further specialization is associated with increasing trade
volume.14
To understand why the consumption home bias vanishes, it is expedient to recall why it arises in the first place. The
dollar volume of trade is equal to the number of firms times the sales per firm. At the moment after opening to trade, there
are nAH H-attribute good exporters that sell relatively less than τ −σ times their domestic sales on the export markets because
competition is relatively stronger in the foreign H sector than in the domestic H sector. This is partly offset because the 1 −
nAH L-attribute good exporters sell more than τ −σ times their domestic sales on the export markets. However, overall, the ef-
fect on trade volume is negative since the H-sector is the more important one for home firms. In contrast, steady state trade
flows are not affected by the underlying taste differences across nations. Since the import competition is biased towards
one sector, the domestic industry concentrates into the other sector. With equally sized countries, this adjustment continues
until πH /P T (vH )−σ = πH∗ /P T ∗ (vH )−σ , hence implying that exports & domestic revenue per firm are the same for H- and L-
attribute firms. Thus, in steady state, heterogeneity and the composition of exports does not matter for average trade flows.
This sheds a new light on Linder’s (1961) hypothesis. His hypothesis hinges on the intuitive notion that low domestic
taste for an attribute is associated with a low volume of imports of goods embodying this attribute. While this insight holds
for a given industry structure, in general equilibrium the country in question loses firms that produce the type of good for
which domestic demand is low, and thus the country becomes a net importer of the good. In general equilibrium, a low
taste for an attribute is thus associated with a large amount of imports embodying the attribute.
The response of industrial composition to trade also has implications for the welfare effects of liberalization, which are
summarized by the following corollary.
Corollary 2 (The gains from trade under incomplete specialization). Assume that πH > πH∗ and that parameters are such that
in the long-run equilibrium of the open economy nTH , nH T ∗ [0, 1]. Then, it is true that at Home, the expected welfare of an H-
valuation consumer is smaller than in the short run after liberalization, but larger than in autarky, while the expected welfare of
an L-valuation consumer is larger than in the short run after liberalization and larger than in autarky. The unweighted average
welfare gain from trade is larger in the long run than in the short run.
Proof. See Appendix
At the moment of opening to trade, the relatively larger domestic group gains less from trade than the smaller group. For
example, given that Germany produces many large car varieties, French consumers with a preference for large cars do gain
more at the moment of liberalization than do the French consumers with a preference for small cars. Dynamically, however,
trade induces French producers of large cars to switch into the small car segment, which favors the lovers of small cars.
4.2. The transition to the open economy steady state
This subsection establishes the transition to the new steady state of the open economy. The first result emerging from
the analysis of this transition is that the static effects of liberalization examined in Section 3 may persist for a nontrivial
amount of time. The reason for this is that immediately after liberalization, the global economy is characterized by the
presence of too many firms so that entry of any kind of firm is unprofitable (see Chaney, 2005 and the survey of firm
dynamics after liberalization in Burstein and Melitz, 2013).
Since firms incur no fixed cost to keep their business alive, there is no active exit and the number of firms only
decreases due to exogenous firm-death at rate δ . Consequently, the autarky industry composition is preserved for some
time and, with it, the impact of cross-country taste differences on the volume of trade in the immediate aftermath of
liberalization described in Section 3 continues to hold. The model differs from what one would expect in a dynamic analysis
of a standard (Melitz, 2003) model, in which no such dynamics materialize due to the absence of a one-time fixed cost.
With flow costs only, the country moves instantaneously to the new steady state when experiencing a trade shock. In the
model presented here, firms make strictly positive instantaneous profits so that an adjustment in the composition of firms
can only materialize via exogenous firm death, thus explaining the inertia.
The second result is that along the transition path, the volume of trade starts to rise gradually once entry occurs. This
happens as entry is entirely directed towards the segment the country specializes in. The increasing specialization, in
turn, leads to the domestic industry leaving precisely the market segments the foreign industry is specialized in and, thus,
increases the volume of trade. The following proposition summarizes firm dynamics after liberalization.
Proposition 5 (Transitional dynamics). Assume that countries are equal-sized and that preferences are symmetric (πH = 1 − πH∗
and eσ vL aL = eσ vH aH = H and eσ vH aL = eσ vL aH = 1). Further, denote the moment of trade liberalization by t0 and assume that
14
Hanson and Xiang (2004) allow for the degree of returns to scale to vary across industries. Similarly, in Fajgelbaum et al. (2011) markups are different
for high quality goods and low quality goods. In these two frameworks, cross-country taste differences can influence the long run volume of trade also in
the case of incomplete specialization, since some countries may end up with industries that are characterized by a high degree of returns to scale.
Number
of firms
NAH
NAL
NTH
NTL
t0 t1 t2 Time
Fig. 1. It shows the laws of motion of NL, t and NH, t , the number of L- and H-type firms after an unanticipated liberalization at time t0 for the case of
πH > πH∗ . There is no entry until t1 and both NL, t and NH, t decrease exponentially at rate δ from the autarky steady state levels NLA and NHA . From t1 to t2 ,
there is entry of H-type firms according to Proposition 5. There is no entry of L-attribute producers and NL, t continues to decrease exponentially at rate δ .
The new steady state is reached at t2 . Thereafter, entry is such to replace firm death, thus keeping NL, t and NH, t stable over time at the open economy
steady state levels NLT and NHT .
export selection costs T are sufficiently large such that

T +F πH ( H + τ − σ ) ( τ − σ πH + ( 1 − πH ) ) + ( 1 − πH ) ( 1 + τ − σ H ) ( πH + τ − σ ( 1 − πH ) )
> .
F ( πH + τ − σ ( 1 − πH ) ) ( τ − σ πH + ( 1 − πH ) ) ( H + 1 )
Then, at home, there is no entry of any firm from time t0 to t1 and the law of motion for the number of home firms is given
by NH,t = nAH N A e−ρ (t−t0 ) and NL,t = nAL N A e−ρ (t−t0 ) . From time t1 to t2 there is only entry of H-type producers and the number
of L-type producers decays exponentially from the autarky level at rate δ . At t2 > t1 the new steady state is reached and there is
entry of both types of producers exactly compensating exogenous firm death. The evolution of firms is given by
NL,t = nAL N A e−ρ (t−t0 )
2
12 + 22 3NL,t − 2 1
− + 2 X NL,t − 2 1 +
2 2
−41 2 3 1 2 3 NL,t − πH 12 + (1 − πH )22 NL,t
1 2
NH,t =
21 2 3
where
T +F
1 = eσ vH aH + τ −σ eσ vL aH , 2 = eσ vL aH + τ −σ eσ vL aL , 3 = L−1 , and
σ (ρ + δ )
2
4 = 12 + 22 3 NL,t − 2 1 − 41 2 3 1 2 3 NL,t 2 − π 2 + 1 − π
( H )2 NL,t
2
H 1
σv a
e H H
σv a σ v a −πH
and the time tome to transition is equal to ln( σv a
e H H
e H H −e H L
τ −σ
) + ln( T +F F ). t1 and t2 are derived in the Ap-
σ v a −πH −(πH −πH ) 1−τ −σ
∗
σv a
e H H −e H L
pendix.
Once entry occurs, it is entirely directed towards the segment the country specializes in. With such increasing special-
ization, the aggregate volume of trade is gradually rising due to the extensive margin as firms with low export revenue exit
the industry and are being replaced with high export revenue firms that produce goods better suited to the foreign demand
structure. Trade also grows along the intensive margin as countries specialize into market segments with comparatively
large domestic demand, implying that domestic firms leave the market segments the foreign industry is specialized in.
This increasing specialization hence increases the average sales of continuing exporters. This transition highlights that
cross-country taste differences can be behind the sluggish response of trade volume to liberalization (see also Yi, 2003;
2010; Ruhl, 2008 and Hummels, 2007).
Fig. 1 summarizes the laws of motion for the number of firms NL, t and NH, t after an unanticipated liberalization at time
t0 . The transitory path of the economy from the autarky steady state to the open economy one is characterized by three
phases. In the first phase between t0 and t1 , there is no entry of any kind of firm. The reason for this is that compared to
autarky, there are too many L-type firms, which also compete for H-type consumers and so markets are too tight even for
H-type firms to enter the market. As there is no entry, the number of both types of firms decreases at the rate of exogenous
firm-death δ . In the second phase between t1 and t2 , there is entry of only H-type firms, while the number of L-type firms
Fig. 2. It shows the evolution of the volume of trade after an unanticipated liberalization at time t0 . The dashed line shows the evolution under πH = πH∗ .
In this case, the volume of trade jumps directly to the steady state level XT at the time of liberalization. If πH = πH∗ , the short-run volume of trade XS is
strictly lower than the steady state level and trade increases once the industry structure starts adjusting at t1 . The volume of trade gradually increases
until it reaches the steady state level at t2 .
continues to decrease. Third, once the number of L-firms has decreased enough so that the new steady state is reached at
t2 , there is entry of both types of firms sufficient to replace firm death, and the size and composition of industry remains
constant thereafter.
Fig. 2 shows the evolution of the volume of trade and how it is affected by cross-country taste differences along the
transition. Consider first the case of countries with identical distributions of tastes (πH = πH∗ ). In this case, the volume of
τ −σ
trade jumps right to its steady state level 1+ τ −σ L at the moment of trade liberalization t0 (see (Eq. (12)). In the presence
of cross-country taste differences, the short-run volume of trade is strictly lower than the steady state level (again, see Eq.
(12)). With the industry composition being unchanged until t1 , the volume of trade remains at the short-run level. After
t1 , the home market effect kicks in, and the volume of trade gradually increases, reaching the steady state level 1+ τ −σ L in
τ −σ
finite time at t2 .
4.3. The case of complete specialization
The extent to which cross-country taste differences affect the volume of trade in the long run is determined by the
degree to which the domestic industry can react to trade liberalization. In the case that countries specialize completely, the
consumption home bias does not disappear entirely, as is shown in the following corollary.
Corollary 3 (Complete specialization in symmetric countries). Assume that countries are of equal size and that preferences are
σ v a τ −σ e σ v L a H
symmetric (πH = 1 − πH∗ and eσ vL aL = eσ vH aH and eσ vH aL = eσ vL aH ). If πH > πH∗ and πH ≥ e eσL vLL −
aL
+eσ vH aL
(1 + τ −σ )−1 , countries
T ∗ = 0) and the volume of trade is equal to
are completely specialized (nTH = 1 and nH

evH aH evH aL
X T = τ −σ L ( 1 − π ) + π −σ v a ,
τ e
− σ v H a H +e v H a L τ e L H + evL aL
where it is true that
LL∗
XS ≤ XT ≤
L + τ −σ L ∗
∂XT
and ∂π < 0.
Large cross-country taste differences thus impede the long-run volume. In the above-introduced example, Germany then
produces only fast cars, for which demand in France is so low that even in the absence of any French fast-car producers
there is still little demand for these type of cars; German exports are thus not “appropriate” for the tastes of French
consumers and taste differences are associated with a consumption home-bias.15
The pervasiveness of both complete specialization and zero trade flows (see, for example, Schott, 2004 and Helpman
et al., 2008) suggests that complete specialization is a relevant case. Indeed, Bernasconi (2013) finds direct evidence that
the degree of country-similarity (in terms income similarity) is a strong predictor of zero trade flows.16
15
A longer working paper version of this study (see Auer, 2010) examines the case of complete specialization by only one country.
16
Still, it is true that the adjustment of industry composition to trade has affected international taste differences in the sense that the volume of trade is
larger in the long run than just after opening markets to trade also for the case of complete specialization.
Together, Corollaries 2 and 3 show the difference between frameworks allowing for endogenous firm entry and those
assuming exogenously given differences in supply, as for example in Hallak’s (2010) sector-level prediction of the Linder
hypothesis. In the latter work, within each sector, cross-country taste differences take the form of country-specific taste
shifters for quality and each country supplies a differentiated good, the quality of which is assumed to be increasing
in the exporter’s income level. The setup of the current paper differs from Hallak (2010) in two important dimensions.
First, cross-country taste differences are conceptually modeled as in Krugman (1980): consumers are heterogeneous within
all nations and cross-country taste differences are modeled as differences in the composition of each population with
regard to preferences for various types of goods. Second, entry into the industry is endogenous and can thus react to
trade liberalization. Together, these differences imply that in the presented model, Hallak’s (2010) prediction does arise if
specialization is complete, but also, that there exists a parameter region in which it does not hold and the long-run volume
of trade is unaffected by taste differences.
5. Taste differences, liberalization and trade in the European automotive industry
This section examines the empirical validity of the above-proposed notion that the adaptation of a country’s industry
composition to the taste structure of a globalized economy increases the volume of trade. For this, I uncover taste differ-
ences along the product attribute dimension relying on assumptions regarding the price elasticity of demand and using
the structure of the model. I then show that in the European automotive industry, cross-country taste differences and the
fact that local industries produce goods that are optimized for domestic consumers initially reduced within-European trade
flows. Only once industrial composition adapted to trade liberalization, within-European trade grew.
Trade in the European automotive industry. The data set examined in this section includes information on prices, sales,
and characteristics of all car models sold in Belgium, France, Germany, Italy, and the UK during 1970–1999. The data has
been collected by Goldberg and Verboven (2001) and Goldberg and Verboven (2005), who also describe it in detail (see
also Verboven, 1996).
The empirical analysis focuses on the study of this industry and this period for two reasons. First, the automotive in-
dustry is well-suited to examine the effect of product attributes on trade since a car is a well-defined unit and information
on a model’s attributes is readily available. The second advantage of analyzing this data is that the European car market
has been subject to considerable trade liberalization during the 1980s and early 1990s. The European Commission had seen
the automotive industry as a “test case” for European integration and had thus initiated a wide spectrum of policies aimed
at integrating the national markets into a single one. The latter policies where focused on removing trade barriers of any
form, encouraging cross-border shopping, and increasing transparency. Although these efforts were started already in the
1980s, they intensified before the launch of the European single market in 1992.17
Fig. 3 documents that the adopted measures greatly increased the volume of intra-European trade. The three lines in
Fig. 3 show the evolution of the number of domestically consumed cars in the five markets (i.e., German cars sold in
Germany; dashed line), imports from the respective 11 other initial European Community members (i.e., Spanish cars sold
in Germany; solid line), and imports from all other countries (dash–dotted line). Whereas domestic sales stagnated and then
declined in the 90s, within-European imports grew more or less steadily and then rose substantially in the 90s. Imports
from non-EC/EU countries grew substantially until 1992, but thereafter stagnated as within-European trade took over.18
Fig. 4 documents that while trade grew substantially, the European market is still surprisingly little integrated. Making
this observation requires establishing a frictionless benchmark of the volume of trade that would prevail if integration were
complete. Following Deardorff (1998) and Yi (2010), the simplest possible benchmark is that in the absence of effective
trade barriers of any kind – may they arise from taxes and physical transportation costs, regulatory requirements, taste
differences, or cross-country price discrimination – a car model should on average sell in proportion to total sales in each
market. Fig. 4 documents the evolution of the median “relative sales abroad” of all car models that are both produced and
sold in the five European markets. For each car model, “relative sales abroad” is defined as the ratio of the model’s sales
abroad compared to sales in the home market, where this ratio is adjusted for differences in total market size. For example,
in 1972, the Fiat 500 model sold 180, 0 0 0 units in Italy and 12, 0 0 0 in Germany. Taking into account that in 1972, in total
1.4 and 1.9 million cars were sold in Italy and Germany, respectively, the relative sales of the Fiat 500 in Germany were
0.05 (≈ ((12, 0 0 0/180, 0 0 0 )/(1.9/1.4 ))) of what they were in Italy. If the European market were truly a “common” one, this
ratio should on average equal 1. However, Fig. 4 documents that the relative sales abroad are surprisingly low on average
(19%) and topped out at 29.5% towards the end of the sample.
It is hard to explain the low degree of integration with observed trade barriers. For example, Broda and Weinstein
(2006) estimate the demand elasticity in SITC Industry 782 (Motor Cars and other Motor Vehicles Principally designed for
the transport of persons and excluding public transport) to equal −3.02. Using this estimate, to match even the maximum
rate of 29.5% of the relative sales abroad observed in the data implies that total effective trade barriers have the same
effect as an 49.5% add-valorem tax (as 1.495−3.02 ≈ 0.295). In 1999, the degree of price discrimination was minuscule (see
Goldberg and Verboven, 2005), tariff rates were zero, and regulatory requirements should in principle have been eliminated.
17
See BEUC (1992) for an outline of the policy changes aimed at increasing the speed of integration and Goldberg and Verboven (2005) and Brenkers
and Verboven (2006) for an analysis of the effects of these measures on price convergence.
18
The decline in sales in 1992 owes to the Europe-wide recession that followed the German Re-unification boom.
6
Annual sales (in mil. units)
2 0 4
1970 1980 1990 2000
Domestic sales Imports from other nations

Imports from EC/EU−12
Fig. 3. It shows the evolution of the number of domestically consumed cars in the five markets (dashed line), imports from the respective 11 other initial
European Community members (solid line), and imports from all other countries (dash–dotted line). All numbers are expressed in million cars and the
data is annual.
.3 .25
Relative sales abroad
.2 .15
.1
1970 1980 1990 2000
Fig. 4. It shows the average over all markets and models of “relative sales abroad,” defined as the ratio of the model’s sales abroad compared to sales
in the home market, where this ratio is adjusted for differences in total market size. For each car model j and market c, relative sales abroad is equal to
Q j,c /Qc
Q j,o /Qo
, where Qj, c is the quantity of j sold on market c, Qc is total sales on market c, and Qj, o is the quantity of j sold on the home market O of model j.
QO is total sales on market O.
Physical transportation costs are small: a door to door delivery of a single car anywhere within these 5 markets costs less
than 500 euros and bulk transport of cars is substantially cheaper.
There are also pronounced differences in the type of cars that countries produce, consume, and trade. Fig. 5 describes
the evolution of the average engine strength of cars that are produced and sold in France and Germany. Engine strength
is measured in KW and all presented averages are volume-weighted. The four lines correspond to cars that are produced
and sold in Germany (solid line), produced in Germany and exported to France (dotted line), produced and sold in France
70 60
Engine power (KW)
50 40
30
1970 1980 1990 2000
DE to FR DE Domestic
FR Domestic FR to GE
Fig. 5. It shows the evolution of the average engine strength of cars that are produced and sold in France and Germany. Engine strength is measured
in KW. The four lines correspond to the average engine strength of cars that are produced and sold in Germany (solid line), produced in Germany and
exported to France (dotted line), produced and sold in France (dashed line), and produced in France and exported to Germany (dash–dotted line). All
presented averages are volume-weighted.
1
Technical fuel efficiency
0 −.5 .5
1970 1980 1990 2000
DE to FR DE Domestic
FR Domestic FR to GE
Fig. 6. It shows the evolution of the average technical fuel efficiency of cars that are produced and sold in France and Germany. Technical fuel efficiency
is equal to − j in the regression F uelConsumption j = α + β KW j + j . The four lines correspond to the average technical fuel efficiency of cars that are
produced and sold in Germany (solid line), produced in Germany and exported to France (dotted line), produced and sold in France (dashed line), and
produced in France and exported to Germany (dash–dotted line). All presented averages are volume-weighted.
(dashed line), and produced in France and exported to Germany (dash–dotted line). The first pattern emerging from this
figure is that German cars have stronger engines than French cars irrespective of where they are sold. The second pattern
20
Standard deviation in KW
10 5 15
1980 1985 1990 1995 2000
Imports vs dom. prod. Dom. prod. across markets
Fig. 7. It shows the standard deviation of differences in the national averages of cars’ engine strength (in KW, solid line). For this line, for each country
A, the average engine strength of all domestically produced cars Avg Engine StrengthDom,Country A is constructed for each year. Second, the standard deviation
of Avg Engine StrengthDom,Country A − Avg Engine StrengthDom,Country B over all bilateral combinations A, B is constructed. The solid line shows the evolution of
this standard deviation over time. The dashed line presents the standard deviation of the engine strength of imported cars versus the one of domestic
production. For this line, the standard deviation of Avg Engine StrengthDom,Country A − Avg Engine StrengthImp,Country A is constructed for each year over all
markets A. The dashed line shows the evolution of this standard deviation over time.
is that German car producers tend to sell stronger cars domestically than when exporting. Fig. 6 shows that on the flip side
of this result, cars produced in France are more fuel-efficient than the ones produced in Germany.19
Cross-country differences in the attribute composition of national output have increased after the creation of the
common market. The solid line in Fig. 7 plots the evolution of the standard deviation of differences in the national averages
of cars’ engine strength. For this, I first construct the average engine strength of all domestically produced cars and second,
I calculate the standard deviation of Avg Engine StrengthDom,Country A − Avg Engine StrengthDom,Country B over all bilateral combi-
nations and for each year. The dashed line in Fig. 8 further shows that over time, also the import baskets have become more
dissimilar to the domestic production basket. For this, I first construct the average engine strength of domestically produced
cars and of the national import baskets. Second, I calculate the standard deviation of AvgEngine StrengthDom,Country A − Avg
Engine StrengthImp, Country A over all five markets and for each year.
Mapping the model to the data. Differences in production, consumption, and trade patterns can arise because of taste
differences, but also because of differences in prices, the set of offered cars, and other factors such as firm-specific
distribution networks. To uncover tastes differences, I start by mapping the sales of each car model on each market to the
above-developed preferences. If within each country, there are two kinds of consumers with taste parameters vL ≡ σ vL and
vH = σ vH , respectively, Eq. (4) implies
⎛ ⎞
⎜ exp[vL a j,c ](1 − πH,c )Lc exp[vH a j,c ]πH,c Lc ⎟
q j,c = p−j,cσ ⎝ (1−σ ) + ( 1 −σ ) ⎠ε j,c . (13)
pn,c exp[vL a j,c ] pn,c exp[vH a j,c ]
n Jc n Jc
Eq. (4) relates the sales of good j in market c to the price pj, c , to how well the attribute aj, c fits with the preferences of
H- and L-valuation consumers, and to how many consumers there are of each type ((1 − πH,c )Lc and Lc π H,c , respectively).
Sales are also affected by the toughness of competition for each type of consumer, which depends on prices and the
attribute composition of competing car models that are offered on the market in question.
Importantly, (13) augments the theory-implied demand in Eq. (4) by an error term ε j, c , which may reflect the impact
of distribution networks, non-tariff trade barriers, and the intrinsic preference for home goods. First, given the importance
19
In Fig. 6, “fuel efficiency” is not equal to fuel consumption (which would simply reflect the fact that cars with smaller engines consume less fuel), but
it graphs “technical fuel efficiency,” defined as a vehicle’s fuel consumption conditional on its engine power. Technical fuel efficiency is constructed by first
regressing fuel consumption on engine power: F uelConsumption j = α + β KW j + j . Technical fuel efficiency is equal to the negative of this model’s residual
j.
0
Effect on trade in percent
−10 −15 −5
1980 1985 1990 1995 2000
Fig. 8. It shows by how much trade flows have been impeded by cross-country taste differences compared to the benchmark of exporter and importer
having identical preferences (see main text for details on the computation). The upper panel A computes the benchmark assuming that the elasticity of
substitution between individual car varieties is equal to 3.02, whereas the lower panel B assumes that this elasticity is equal to 10.45.
of distribution costs and dealer networks in this industry, not only the characteristics of the car itself, but also those of the
company and its local dealer network matter for sales. Second, there may also be an intrinsic preference for domestically
produced goods and it may also be the case that regulation is tailored towards domestically produced cars. These consid-
erations imply that ε j, c depends on the distribution network of the associated brand in market c and on whether the car is
produced in its home market or imported from abroad.
Denoting the brand associated with model j by bj , the impact of the brand-and-market specific distribution network on
sales by d(bj , c), and a dummy that takes the value 0 if the car originates from market c and 1 otherwise by j,c , it holds
that
ln(ε j,c ) = d (b j , c ) + hb( j,c ) + ln(ε
j,c ), (14)
where ln(ε j,c ) denotes the residual error. It is possible to account for the effect of home bias and distribution networks in
(13) because dealer networks affect all car models of a firm in any given market equally. The same holds for the effect of
the home bias hb(j, c ) if the car models from each brand originate from a common location of production.
Consider an estimation restricted to a sample with unique brand-origin combinations (i.e. the sample includes only
those BMWs that are produced in Germany while only the Renaults that are produced in France are included). Since
d (b j , c ) + hb( j,c ) only varies at the level of the brand and the market, the term can be subsumed in brand-and-market
specific fixed effects. Taking this into account yields the relation
ln(q j,c ) = −σ ln p j,c + ln(exp[vL a j,c ]Lc + exp[vH a j,c ]H
c ) + b j ,c + ln (ε j,c )
(15)
where b j ,c corresponds to the brand and market-specific fixed effect. The country- and valuation-specific constants are
(1−σ ) (1−σ )
defined as Lc ≡ (1 − πH,c )Lc ( pn,c c ≡ πH,c Lc (
exp[vL a j,c ] )−1 and H pn,c exp[vH a j,c ] )−1 .
n Jc n Jc
Under the assumption that ε j,c is orthogonal to the attribute of the model, estimation of (15) allows to uncover the
extent to which π H, c varies over the markets c. However, a key concern is that ε j,c is correlated with pj, c . Since structural
estimation of σ that addresses this endogeneity is beyond the scope of the paper, I instead first assume that σ is equal
to 3.02, as estimated by Broda and Weinstein (2006). I then estimate (15) using nonlinear least squares to uncover the
parameters (vL , vH ) and the fixed effects of interest. I then back out the country-specific values of π H, c using the definitions
of Lc and H c . When estimating the model for the luxury dummy (a j,c = [0, 1], where 1 indicates that the model is a
luxury car) and assuming σ = 3.02, I find that vL = 1.85 and vH = 2.13. The backed-out π H, c are as follows: Germany has
the highest preference for luxury cars with πH,c = 0.90, followed by the UK (πH,c = 0.60), France (πH,c = 0.55), Belgium
(πH,c = 0.01), and Italy (πH,c = 0).
An alternative estimate for σ is available from Blonigen and Soderbery (2010), who provide estimates on the elasticity of
substitution between car varieties in the USA. Their approach differs from that of Broda and Weinstein (2006) as their data
also includes information on domestic sales. The average elasticity in the sample of compact, mid-size and full-size cars is
10.45 (reported in their Table 7). When estimating the model for the luxury dummy and assuming σ = 10.45, I find that
vL = 4.96 and vH = 8.08. The backed-out π H, c are different quantitatively, but confirm the ordering of tastes found when
assuming σ = 3.02: Germany has the highest preference for luxury cars with πH,c = 0.83, followed by the UK (πH,c = 0.75),
France (πH,c = 0.73), Belgium (πH,c = 0.61), and Italy. (πH,c = 0.41). Table A.1 in the Appendix lists the estimated coefficients
and the associated standard errors for the two estimations assuming either σ = 3.02 or σ = 10.45.20
The impact of taste heterogeneity on trade. In next use the model and the uncovered parameters to examine by how much
taste differences have impeded trade flows. Quantifying the impact of taste differences on trade flows implies comparing
the actual volume of trade to the one prevailing under a benchmark in which importer and exporter share the same
preferences. The specific benchmark I adopt is the following: at every given point in time, I take the set of cars that are
offered on each market, as well as their attributes and prices, as a given. I also take the residual demand for each model
as given (see (14)) and examine how high the sales of each exported model were if the importing country c shared the
preference of the exporter e : πH,c = πH,e . I calculate the impact of this counterfactual for all bilateral trade flows using the
tastes uncovered in (15), including not only the luxury dummy as attribute, but also the car’s class, engine power, and fuel
efficiency. That is, I uncover four taste parameters π H, c for each country and for each of the four attributes luxury dummy,
car class (aj, c can take five different values), engine power (aj, c is a continuous variable) and fuel efficiency (again aj, c
continuous variable). I then calculate the total impact on trade in percentage terms of actual trade flows comparing actual
trade flows to the one that are predicted by Eq. (15) when setting πH,c = πH,e .
The two panels of Fig. 8 document that the slow response of industrial composition to the demand structure of the open
market implied that it took time until within-European trade reached its new steady state. I calculate the above-described
counterfactual for the full sample and for various sub periods. The figure presents the results of this counterfactual exercise
over a six-year rolling window from 1976 to 1999. The upper Panel A documents the resulting estimate of how taste
differences have impeded trade flows over the last 30 years under the assumption σ = 3.02: trade could have been up to
10% higher than observed in the mid-90s had the industry structure already adjusted. The lower Panel B documents the
same counterfactual for the scenario uncovering taste differences under the assumption that σ = 10.45. In this scenario, the
magnitude of the effects is lower, but the qualitative pattern is similar to the one in Panel A.
These results show that the key prediction of the above-developed theoretical analysis is strongly supported in the
data. The initial mismatch between the demand structure of the open market and the industry structure prevailing just
after liberalization has led to a sizeable reduction of within-European trade (see Fig. 8). Only once the industry structure
adjusted to liberalization and specialization increased (see Fig. 7), cross-country differences in production ceased to reduce
trade flows. After the creation of the European common market, the slow adaptation of industrial composition to trade
liberalization was hence associated with sluggish growth of within-European trade in the automotive industry.
6. Conclusion
This paper examines the dynamic impact of trade liberalization in the presence of cross-country taste differences and
the associated home market effects. It first develops a structural model of demand featuring consumers with homothetic
preferences and heterogeneous tastes over attributes. The paper thus provides an extension of Krugman (1980), in particular
relaxing his assumption that expenditures for any type of good are fixed exogenously and do not react to the availability of
fitting varieties and their relative prices.
Nesting these preferences in an international economy setting featuring iceberg trade costs and costly export market
access, the model predicts that consumption is home-biased in the immediate aftermath of liberalization since each
country’s industry structure is optimized for the preferences of domestic consumers and domestic output thus does not
match well with preferences abroad.
Along the transition to the open economy steady state, national industries specialize away from market segments in
which other nations have comparatively large domestic demand. This increasing specialization that underlies the home mar-
ket effect increases the average demand for foreign goods, the volume of trade, and the average gains from liberalization. The
theoretical analysis thus predicts that cross-country demand differences have important implications for the dynamic impact
of trade liberalization: the volume of trade only grows sluggish after liberalization, since each country’s industrial composi-
tion has to adapt to the demand structure of the open economy, which requires firm exit and entry and, therefore, time.
I conclude by providing empirical evidence that such industrial reshuffling was indeed an important determinant of the
dynamics of trade in the automotive industry following the creation of the European common market.
These findings highlight that endogenizing how a nation’s industrial composition responds to trade liberalization is of
first-order importance for understanding trade patterns and the welfare gains from open markets. The paper also offers a
new explanation of why trade grows sluggish after liberalization (as for example documented by Yi (2003) or Hummels
(2007). After a liberalization, each country’s industrial composition has to adapt to the demand structure of a globalized
economy, which requires firm exit and entry and, therefore, time. In contrast to the existing literature, this is not driven
20
A worry that is not addressed in this estimation is the oligopoly nature of this industry, giving rise to strategic interactions and hence potentially
perceived price elasticities that are heterogeneous for each model and market. Such differences are assumed away in (15). Cosar et al. (2015) uncover taste
differences in an estimation building on Berry et al. (1995, 1999) that accounts for such strategic interactions.
by a trade-induced shift towards ex-ante more profitable entities (i.e. within sectors), but rather, by the adaptation of a
country’s sectoral composition to the taste structure of a globalized economy.
Appendix
A1. Coefficient estimates
Table A.1.
A2. Proofs
Proposition 1 (Demand) (reminded) Denote the demand function of a firm with attribute aj and charging price pj by D(aj ,
pj ). Demand is determined by

exp[σ va j ]
D(a j , p j ) = (1 − α )θ L(1 − σ ) p−j (1+σ ) f v (v ) −σ d v, (16)
v V
P (v )
where (...) is the gamma function and P (v ) denotes the ideal price index for all consumers with vi = v and is equal to
−σ −1/σ
pn
P (v ) = . (17)
exp [van ]
n J
Proof. A consumer with valuation v buys only from the firm offering the good with the lowest effective unit price, i.e.,
a j v+xi, j
each consumer chooses
j = arg max e p . Since the distribution of xi, j is continuous the probability of ties is 0. From the
j
i J
firm side, (expected) demand from consumer v with an unknown realization of xi, j is then equal to the probability that the
firm’s draw xi, j , adjusted for the firms’ price and the match of aj and v is the maximum of all adjusted draws. Since each
consumer spends (1 − α )θi on the manufacturing composite, spends it all on one variety only, sales are given by

( 1 − α ) θi ea j v+xi, j ean v+xi,n
D j ( a j , p j , vi = v ) = gx xi, j Pr = max dxi, j
pj xi, j X pj n J pn

= Pr(xi,n < ln pn ) − ln( p j ) + (a j − an )v + xi, j dxi, j
If all x are distributed Gumbel with scale parameter 0 and shape parameter 1/σ , the following holds
1
gx ( x ) = exp[−xσ ] exp[− exp[−xσ ]]
σ
and thus
Pr(xi,n < ln( pn ) − ln( p j ) + (a j − an )v + xi, j ) = exp[−pσj p− σ
n exp[−σ (a j v + xi, j )] exp[vσ an ]]
so that
!

Pr xi,n < ln( pn ) − ln( p j ) + (a j − an )v + xi, j = exp −p1j /σ exp −σ (a j v + xi, j ) ( p−
n
σ exp[vσ an ] )
n= j J= j
Table A1
Coefficient estimates in Eq. (17) for σ = 3.02 and σ = 10.45.
σ = 3.02 σ = 10.45
Coefficient Estimate Standard error Estimate Standard error
H
B elgium −502.7 Na −10.04 0.74
H
F rance −3.48 0.68 −10.30 1.42
H
G ermany −3.17 0.62 −10.80 −10.79
H
I taly −229,0 0 0 Na −11.47 Na
H
UK −3.09 4.66 −9.22 1.42
H
B elgium −2.78 0.45 −10.49 2.88
L
F rance −3.70 0.59 −11.31 3.21
L
G ermany −5.37 Na −12.35 3.27
L
I taly −3.05 0.55 −11.11 3.21
L
UK −3.51 0.59 −10.31 3.18
vH 2.13 0.047 8.08 1.27
vL 1.83 0.027 4.96 Na

Since (1 + pσj exp[−vσ a j ] ( p−n σ exp[vσ an ] )) = ( p−n σ exp[vσ an ] ). Now substituting: zi, j =
J= j J= j

pσj exp[−vσ a j ] ( p−n σ exp[vσ an ] )xi, j in D j (a j , p j , vi = v ), leads to
j J

( 1 − α ) θi 1 σ
D j (a j , p j , v ) = ( p exp[−vσ a j ] ( p−n σ exp[vσ an ] ))−1 exp[−zi, j σ ] exp[− exp[−σ zi, j ]]dzxi, j .
pj σ j j J
zi, j X
Since the latter part can be expressed as the CDF of a Gumbel shock, we get demand per mass 1 of consumers with valuation
v
( 1 − α ) θi w p−j σ exp[σ va j ]
D j (a j , p j , v ) = . (18)
pj pj ( p−n σ exp[σ van ] )
n J
Given the wage w is normalized to 1 implies a firm’s total demand Dj (aj , pj ) is equal to the integral over all possible
valuations v.
p−j σ exp[σ va j ]
( 1 − α ) θi
D j (a j , p j ) = L jv ( v ) dv
pj v V ( p−n σ exp[σ van ] )
n J
Corollary 1 (Expected consumer welfare) (reminded). Denote the expected welfare of consumer i with vi = v and income θ i
by E(Ui |v, θ i ). If pO = 1,
σ
E (Ui ) = (1 − α )1−α α α 1 − ( P ( v ) ) −α θi
α
where the ideal price index P (v ) is as defined in (5) and (..) is the gamma function.
exp[vi a j +xi, j ]
Proof. The consumer only buys from the draw and match-adjusted cheapest firm. Define j ∗ (i ) ≡ arg max( pj ). Con-
j J
ditional on this j∗ (i), consumer i maximizes
Ui = max O1i −α (qi, j∗ (i ) exi, j∗ (i) +a j∗ (i) vi )α − λi [Oi pO +qi, j∗ (i ) p j∗ (i ) − θi ]

Oi ,qi, j∗ (i ) i I
Implying that the value of the Langragian multiplier, or the marginal utility with respect to increasing income θ i , equals λi =
(1 − α )1−α α α p−O (1−α ) ( xi, j∗ (i)j +(ai)j∗ (i) vi )−α . With pO normalized to 1, the utility for a given maximum realization of xi, j∗ (i) +
p∗
e
a j∗ (i ) vi is thus
α
1 −α exi, j∗ (i) +a j∗ (i) vi
Ui = (1 − α ) αα θ i .
p j ∗ (i )
) denote the cdf of U
How is the expectation of the maximized utility distributed? Let F (U = U /(1 − α )1−α α α θ , which is
i i i i
distributed

xi, j +a j vi
α
i ) = Pr max e
F (U i
<U
j J pj
" #
i α + ln p j
1
= Pr xi, j < ln U
ea j vi
j J
" " " 1
p # ##
= i α + ln
exp − exp − ln U
j
σ
a j vi
e
j J
⎡ − σα ⎤
i
U
= exp −⎣ σα ⎦
j J p−j σ eσ a j vi
is distributed Frechet with scale parameter ( p−σ eσ a j vi ) σα and shape parameter σ .

Ui j J j α
−α
i ) = P (vi )
E (Ui ) = (1 − α )1−α α α θi E (U ,
where = (1 − α )1−α α α (1 − σ α ).
Lemma 1 (Liberalization and short-run relative competition) (reminded) Assume that πH > πH∗ and nAH , nAH∗ [0, 1]. When
opening markets to trade, competition in Home intensifies more in the L-attribute segment of the industry than in the
H-attribute segment, while competition in Foreign intensifies more in the H-attribute segment of the industry than the
L-attribute segment. I.e., it is true that
P S ( vH ) P A ( vH ) P S∗ (vH ) P A∗ ( vH )
> A and S∗ < A∗ .
P ( vL )
S P ( vL ) P ( vL ) P ( vL )
Proof. Since all firms export and entry is given by the autarky equilibrium values, it holds that P S (vH )−σ = (N A nAH +
τ −σ NA∗ nAH∗ )eσ vH aH + (NA (1 − nAH ) + τ −σ NA∗ (1 − nAH∗ ))eσ vH aL , implying
−1σ
−1 πH L + τ −σ L∗ πH∗
P ( vH ) =
S
(σ (δ + ρ )F ) φ and, similarly,
eσ vL aL − eσ vL aH
−1σ
−1
L(1 − πH ) + τ −σ L∗ 1 − πH∗
P ( vL ) =
S
(σ (δ + ρ )F ) φ .
eσ vH aH − eσ vH aL
it thus holds that

−1σ
P S ( vH ) πH L + τ −σ L∗ πH∗ eσ vH aH − eσ vH aL
= σ vL aL
P S ( vL ) L ( 1 − πH ) + τ L 1 − πH e
− σ ∗ ∗ − eσ vL aH
which compares to the relative level of price indices in autarky (see (11)) of
−1σ
P A ( vH ) πH e σ v H a H − e σ v H a L
= ,
P A ( vL ) 1 − πH e σ v L a L − e σ v L a H
P S ( vH ) P A ( vH ) πH L+τ −σ L∗ πH∗ πH
documenting that > is identical to showing that < , which holds iff πH > πH∗ . A sym-
P S ( vL ) P A ( vL ) L (1−πH )+τ −σ L∗ (1−πH∗ ) 1−πH
P S∗ (vH ) P A∗ ( vH )
metrical argument for foreign proves that < iff πH > πH∗ .
P S∗ (vL ) P A∗ ( vL )
Lemma 2 (Short-run attribute content of trade) (reminded) Assume that parameters are such that nAH , nAH∗ [0, 1]. At the
moment after trade liberalization, if L = L∗ , Home is a net exporter of H-attribute goods iff πH > πH∗ . If L = L∗ Home’s
manufacturing exports contain a larger fraction of H-attribute goods than do Foreign’s exports.
Proof. Home’s net exports of H-attribute goods are equal to the number of H-attribute Home firms times exports per such
firm minus the same multiplicative in Foreign

−σ σ v L a H
e τ −σ e σ v H a H τ −σ e σ v L a H
XHS − XHS∗ = N A nAH πH∗ L∗ ∗S + ( 1 − π )L ) − N A∗ ∗A
n π L + ( 1 − π )L
P (vH )−σ P ∗S (vL )−σ P S (vH )−σ P S (vL )−σ
H H H H
For the case of L = L∗ , taking into account the toughness of competition in Foreign as well as at Home yields
(1 − πH )πH (1 − τ −σ )
XHS − XHS∗ = (πH − (1 − πH ) )τ −σ L
( τ πH + ( 1 − πH ) ) ( τ − σ ( 1 − πH ) + πH )
−σ

πH ( 1 − πH )
+ τ − σ L ∗ − −σ πH
τ −σ 1−ππH H +1 τ 1−πH + 1
eσ vH aH eσ vH aL eσ vL aL −eσ vL aH σ v a eσ vL aH σv a σv a
σ vL aH > 0. Since both πH − (1 − πH ) and
−e H L
where = eσ vH aH eσ vL aL −eσ vL aH eσ vH aL eσ vH aH −eσ vH aL
+ eσ vH aH eeσ vLLaLL−e σ vH aL eσ vL aH · eeσ vHL aHL −e
πH (1−πH ) XHS XHS∗
1 −π − −σ π H are larger than 0, XH − XH > 0 for πH > πH∗ . Next,
∗ for the case of L = L∗ , to show that >
τ −σ π H +1 τ 1−π +1 XHS +XLS XHS∗ +XLS∗
H H
XHS XHS∗
it suffices to show that > .
XLS XLS∗
σv a σv a
XHS πH∗ P∗Se (vHH )H−σ + 1 − πH∗ P∗Se (vLL )H−σ
nAH
XLS
= σv a σv a
1 − nAH πH∗ P∗Se (vH )L−σ + 1 − πH∗ P∗Se (vL )L−σ
H L
eσ vL aL
eσ vH aL
π
eσ vL aL −eσ vL aH H
∗
− 1 − π ∗
H eσ vH aH −eσ vH aL
=
πH∗ eσ vH aL ( 1−πH∗ ) eσ vL aL
τ −σ πH L+L∗ πH∗ eσ vH aH −eσ vH aL + τ −σ L(1−πH )+L∗ (1−πH∗ ) eσ vL aL −eσ vL aH

eσ vH aH eσ vL aH
XHS∗ H πH P S (vH )−σ + (1 − πH ) P S (vL )−σ
n∗A
XLS∗
= σv a σv a

1 − n∗AH
πH PSe(vHH)−L σ + (1 − πH ) PSe(vLL)−L σ
eσ vL aL
σ vH aL
eσ vL aL −eσ vL aH
πH∗ − 1 − πH∗ eσ vHeaH −e σ vH aL
= eσ vH aH
σv a σ vH aL
eσ vH aH −eσ vH aL
− πH∗ eσ vL aeL −eL σL vL aH + eσ vHeaH −e σ vH aL
eσ vL aL σv a eσ vL aL σv a XHS XHS∗
Since both eσ vL aL −eσ vL aH
πH − (1 − πH ) eσ vHeaH −e
H L
σ vH aL and eσ vL aL −eσ vL aH
πH∗ − (1 − πH∗ ) eσ vHeaH −e
H L
σ vH aL 0, 1, it is true that
XLS
>
XLS∗
for πH > πH∗
Proposition 3 (Short-run trade volume) (reminded) Assume that parameters are such that nAH , nAH∗ [0, 1]. At the moment
after trade liberalization, the following holds. If πH∗ = πH , the volume of trade is the same as in the absence of consumer
heterogeneity and home is a net exporter of the M good iff L > L∗ . If πH = πH∗ , the volume of trade is lower than in the
absence of consumer heterogeneity. Regarding the volume of trade,
I. For given τ −σ , the volume of trade is decreasing in |πH − πH∗ |.
II. The importance of taste heterogeneity is decreasing in τ σ .
Proof. Denote the total value of exports at the moment after opening markets to trade by XS and XS∗ and the attribute-
specific trade flows by an additional H or L subscript. For each type of good, the value of trade is proportional to the
number of firms of each type and the sales per such firm, i.e., Home’s export volume equals

−σ σ v L a H
e
XHS = N A∗ nAH πH∗ L∗ ∗S + ( 1 − π H) L and
P (vH )−σ P ∗S (vL )−σ

τ −σ e σ v H a L ∗ ∗τ
−σ σ v L a L
e
XLS = N A∗ 1 − nAh πH∗ L∗ + ( 1 − π H) L .
P ∗S (vH )−σ P ∗S (vL )−σ
The two ideal price indices for foreign H- and L-valuation consumers are given by Lemma 1 thus yielding

τ −σ LL∗ πH πH∗ τ −σ LL∗ (1 − πH ) 1 − πH∗
X = S
+ (19)
LπH + τ −σ L∗ πH∗ L(1 − πH ) + τ −σ L∗ 1 − πH∗

τ −σ LL∗ πH∗ πH τ −σ LL∗ 1 − πH∗ (1 − πH )
S∗
X = −σ + (20)
τ LπH + L∗ πH∗ τ −σ L(1 − πH ) + L∗ 1 − πH∗
π∗ (1−πH∗ )
Next, note that πH τ −σ Lπ H+L∗ π ∗ + (1 − πH ) ∗ = τ −σ1L+L∗ and that
H H τ −σ L(1−πH )+L∗ (1−πH∗ ) πH =πH
(
∂ X S < 0 if πH∗ > πH
∗
= 0 if πH = πH
∂πH >0 if π < π ∗
H H
π πH∗ (1−π )(1−π ∗ )
so that τ −σ LL∗ ( τ −σ LπH ∗π ∗ + τ −σ L(1−πH )+L∗ (H1−π ∗ ) ) < if πH∗ = πH and it holds that
H +L H H H
) τ −σ LL∗
S
= N A +τ −σ N A∗
if πH∗ = πH
X τ −σ LL∗
,
< N A +τ −σ N A∗
if πH = πH∗
−σ ∗
which verifies the first two claims of Proposition 3. Next, setting πH = πH∗ in (20) and (19) yields X − X ∗ = Lτ+τ −LL
σ L∗ −
τ −σ LL∗ , which has the described sings depending on L, L∗ . For the second part of the claim, note that at L = L∗ , X −
τ −σ L + L ∗
(π 2 +π ∗2 )(1−π )(1−π ∗ )−((1−π )2 +(1−π ∗ )2 )π π ∗
X ∗ = (τ −σ π +π ∗H)(π H+τ −σ π H∗ )(τ −σ (H1−π )+(1H−π ∗ ))((1H−π )+Hτ −Hσ (1−π ∗ )) πH − πH∗ τ −σ (1 + τ −σ ). The latter expression is 0 when-
H H H H H H H H
ever πH + πH∗ = 1, positive if πH + πH∗ < 1 and πH ≥ πH∗ , and negative if πH + πH∗ < 1 and πH ≥ πH∗ . The latter sign is reversed
if πH < πH . A similar calculation for foreign yields (20). Thus, if L = L∗ and πH = πH∗ net trade flows satisfy
∗
(
>0 if πH + πH∗ < 1
S∗
X − X L=L∗ ;πH ≥πH = = 0 if πH + πH∗ = 1 .
S ∗
< 0 if πH + πH∗ < 1

It is noteworthy that due to the presence of the O sector, wages are equal across the two countries and thus a net trade
flow in labor units is equal to a net trade flow in Dollars.
Lemma 3 (Domestic revenue in the open economy) (reminded) If parameters are such that in the long-run equilibrium
T ∗ [0, 1], it is true that the domestic revenue (a ) of a firm is independent of the type of the
of the open economy nTH , nH j
good it produces, i.e.
(aH ) = (aL ) and ∗ (aH ) = ∗ (aL )
σ vH a j σ vL a j
Proof. Denote by (a j ) ≡ LπH Pe(v )−σ + L(1 − πH ) Pe(v )−σ the domestic revenue of a Home firm with good aj in the Home
H L
σ vH a j σ vL a j
market and by ∗ (a j ) ≡ L∗ πH∗ Pe(v )−σ + L∗ (1 − πH∗ ) P∗e(v )−σ the domestic revenue of a Foreign firm with good aj in the For-
H L
eign market. Since all firms export, face a constant elasticity of demand, and are subject to iceberg transportation costs, the
export revenue of a Home firm is equal to τ −σ ∗ (a j ) so that total revenue of a home firm is equal to (a j ) + τ −σ ∗ (a j ).
Similarly, the total revenue of a Foreign firm equals τ −σ (a j ) + ∗ (a j ). An equilibrium without complete specialization
requires that the discounted sales of an H-attribute and an L-attribute firm are equal in Home and in Foreign:
∞ ∞
1 1
e−(δ +ρ )t ((aH ) + τ −σ ∗ (aH ))dt = e−(δ +ρ )t ((aL ) + τ −σ ∗ (aL ))dt
t=0 σ t=0 σ
∞ ∞
1 1
e−(δ +ρ )t (τ −σ (aH ) + ∗ (aH ))dt = e−(δ +ρ )t (τ −σ (aL ) + ∗ (aL ))dt
t=0 σ t=0 σ
Solving the integral and subtracting τ −σ times the upper equation from the lower equation yields (aH ) = (aL ) and
subtracting τ −σ times the lower equation from the upper one yields ∗ (aH ) = ∗ (aL ).
Proposition 4 (Trade and its net attribute content) (reminded) Assume that πH > πH∗ and that parameters are such that
∗
T ∗ [ 0 , 1 ].
nTH , nH Then, the volume of Home’s manufacturing exports is equal to L+τLL−σ L∗ for any πH , πH∗ that satisfy the stated
assumptions. If L∗ = L, Home has 0 net exports of manufactured goods. If L > L∗ , Home is a net exporter of manufactured
goods. For any combination of L and L∗ that is consistent with the stated assumptions, Home’s manufacturing exports
are more H-attribute intensive that Foreign’s manufacturing exports. Home’s manufacturing exports are more H-attribute
intensive in the open economy equilibrium than just after trade liberalization.
Proof. Home’s Exports and Imports of goods that embody the H-attribute are proportional to the number of domestic,
respectively, foreign H firms. Home’s total exports of H-attribute goods equal

τ −σ e σ v H a H ∗ τ
−σ σ v L a H
e
XHT = N T nTH L∗ πH∗ − σ + L ( 1 − πH ) T ∗
∗
P ( vH )
T ∗ P (vL )−σ
L ∗ τ −σ
= N T nTH
τ −σ N T + N T ∗
Similarly,
L ∗ τ −σ
XLT = 1 − nTH N T
τ −σ N T + NT ∗
Lτ − σ
XHT ∗ = nTH∗ N T ∗
N + τ −σ N T ∗
T
T∗ Lτ − σ
XLT ∗ T∗
= 1 − nH N
N T + τ −σ N T ∗
Because each H-attribute and L-attribute producer of a country exports the same amount, the composition of the
industry does not affect the overall volume of exports.
L ∗ τ −σ τ −σ
X T = XHT + XLT = N T = L − τ −σ L ∗
τ −σ N T + N T ∗ 1 − τ −2σ
L τ −σ −σ
τ −σ
XT∗ = XHT ∗ + XLT ∗ = N T ∗ T = L ∗
− τ L
N + τ −σ N T ∗ 1 − τ −2σ
and Home’s net exports of manufactured goods equal
τ −σ
X T − X T ∗ = ( L − L∗ )
1 + τ −σ
Home’s net exports of H-attribute equal
eσ aL vL eσ vH aL
XHT − XHT ∗ = (πH − πH∗ ) σ a v − ( ( 1 − π H ) − ( 1 − π ∗
) )
e L L − eσ aH vL H
eσ vH aH − eσ vH aL
and the labor intensity of
XHT XHT ∗
= nTH > nTH∗ =
XHT + XLT XHT ∗ + XLT ∗
Last, it is also true that
XHT XHS
= nTH >
XHT + XLT XHS + XLS
Since both nTH > nAH and after trade liberalization, an H-attribute firm exports less than an L-attribute firms so that
σv a σv a
πH∗ e H H + (1−πH∗ ) ∗S
e L H
P ∗S (vH )−σ P (vL )−σ
−σ σ v a −σ σ v L a L < 1.
πH∗ L∗ τ ∗S e H−σL + (1−πH∗ )L∗ τ ∗S e −
P ( vH ) P ( vL ) σ
Corollary 2 (The gains from trade under incomplete specialization) (reminded) Assume that πH > πH∗ and that parameters
are such that in the long-run equilibrium of the open economy nTH , nH T ∗ [0, 1]. Then, it is true that at Home, the expected
welfare of an H-valuation consumer is smaller than in the short run after liberalization, but larger than in autarky, while
the expected welfare of an L-valuation consumer is larger than in the short run after liberalization and larger than in
autarky. The unweighted average welfare gain from trade is larger in the long run than in the short run.
Proof. Note that by Corollary 1, the gains from trade are exclusively dependent on the ideal price indices. Lemma 3 pins
down the sales of firms, and recalling the definition of domestic revenue (aj ) and ∗ (aj ) in Eq. (7) yields
P T (vH )−σ e σ a H v H − e σ a L v H πH P T ∗ (vH )−σ eσ aH vH − eσ aL vH πH∗

−σ = σ a L v L and −σ = σ a L v L ,
P ( vL )
T e −e σ a H v L 1 − πH P ( vL )
T ∗ e − eσ aH vL 1 − πH∗
which, when compared to Lemma 1 and the autarky ideal price indices (see Proposition 2), satisfies the stated relations.
Proposition 5 Transitional dynamics (reminded). Assume that countries are equal-sized and that preferences are symmetric
(πH = 1 − πH∗ and eσ vL aL = eσ vH aH = H and eσ vH aL = eσ vL aH = 1). Further, denote the moment of trade liberalization by t0 and
assume that export selection costs T are sufficiently large such that
T +F πH ( H + τ − σ ) ( τ − σ πH + ( 1 − πH ) ) + ( 1 − πH ) ( 1 + τ − σ H ) ( πH + τ − σ ( 1 − πH ) )
> .
F ( πH + τ − σ ( 1 − πH ) ) ( τ − σ πH + ( 1 − πH ) ) ( H + 1 )
Then, at home, there is no entry of any firm from time t0 to t1 and the law of motion for the number of home firms is given
by NH,t = nAH N A e−ρ (t−t0 ) and NL,t = nAL N A e−ρ (t−t0 ) . From time t1 to t2 there is only entry of H-type producers and the number
of L -type producers decays exponentially from the autarky level at rate δ . At t2 > t1 the new steady state is reached and
there is entry of both types of producers exactly compensating exogenous firm death. The evolution of firms is given by

2
12 + 22 3NL,t − 2 1
− 12 + 2 X NL,t − 2 1 +
2
−41 2 3 1 2 3 NL,t
2
− πH 12 + (1 − πH )22 NL,t
NH,t =
21 2 3
where
T +F
1 = eσ vH aH + τ −σ eσ vL aH , 2 = eσ vL aH + τ −σ eσ vL aL , 3 = L−1 , and
σ (ρ + δ )
2
4 = 12 + 22 3 NL,t − 2 1 − 41 2 3 1 2 3 NL,t 2 − π 2 + 1 − π
( H )2 NL,t
2
H 1
σv a
e H H
σv a σ v a −πH
and the time tome to transition is equal to ln( σ v
e H H
a
e H H −e H L
τ −σ
) + ln( T +
F ). t1 and t2 are derived in the
F
σv a σv a − π H − (πH −πH ) 1−τ −σ
∗
e H H −e H L
Appendix.
Proof. Note that:

∗ σv a
πH Leσ vH aH (1 − πH )Leσ vL aH π ∗ ∗ σ vH aH
L e 1 − π ∗
L e L H
+ τ −σ = σ (ρ + δ ) (T + F )
H H
−σ + −σ σ +
P ( vH ) P ( vL ) P ∗ ( vH )−
P ∗ (vL )−σ
whereas in foreign

−σ πH Leσ vH aL (1 − πH )Leσ vL aL πH∗ L∗ eσ vH aL 1 − πH∗ L∗ eσ vL aL T +F
τ + + + =
P (vH )−σ P (vL )−σ P ∗ (vH )−σ P ∗ (vL )−σ σ (ρ + δ )
Note by perfect symmetry P (vH )−σ = P ∗ (vL )−σ and P (vL )−σ = P ∗ (vH )−σ and πH∗ = (1 − πH ) together
πH ( 1 − πH ) σ v L a H T +F
e σ v H a H + τ −σ e σ v L a H + e + τ −σ eσ vH aH = L−1
P (vH )−σ P (vL )−σ σ (ρ + δ )
πH σ v H a L − σ ( 1 − πH ) − σ σ v a T +F
e τ + eσ vL aL + −σ τ e L L + eσ vH aL = L−1
P (vH )−σ P ( vL ) σ ( ρ + δ)
Noting that (eσ vH aH + τ −σ eσ vL aH ) = (eσ vH aL τ −σ + eσ vL aL ) and (τ −σ eσ vL aL + eσ vH aL ) = (eσ vL aH + τ −σ eσ vH aH ), and by symmetry
it is true that

P (vH )−σ = eσ vH aH + τ −σ eσ vH aL NH,t + eσ vH aL + τ −σ eσ vH aH NL,t = 1 NH,t + 2 NL,t

P (vL )−σ = eσ vL aH + τ −σ eσ vL aL NH,t + eσ vL aL + τ −σ eσ vL aH NL,t = 2 NH,t + 1 NL,t
Next, defining 1 = eσ vH aH + τ −σ eσ vL aH > 2 = eσ vL aH + τ −σ eσ vL aL it holds that

πH eσ vH aH + τ −σ eσ vL aH P (vL )−σ = πH 1 (NH,t 2 + 1 NL,t )
and setting L−1 σ (Tρ++Fδ ) = 3 , leads to

1 2 3 NH,t
2
+ NH,t 12 + 22 3 NL,t − 2 1 + 1 2 3 NL,t
2
− πH 12 + (1 − πH )22 NL,t = 0
with the solution of the quadratic root. Last, to solve for t1 , note that from liberalization onwards, both types of firms are
decreasing at rate e−ρ (t−t0 ) . Thus

πH Leσ vH aH (1 − πH )Leσ vL aH −σ πH∗ L∗ eσ vH aH 1 − πH∗ L∗ eσ vL aH T +F
+ +τ + =
P (vH )−σ P (vL )−σ P ∗ (vH )−σ P ∗ (vL )−σ σ (ρ + δ )
so that
πH 1 (1 − πH )2 T +F
+ = L−1 ,
1 NH,t + 2 NL,t 2 NH,t + 1 NL,t σ (ρ + δ )
where since NH,t = nAH N A e−ρ (t−t0 ) and NL,t = nAL N A e−ρ (t−t0 ) and N A = σ (ρ L+δ )F t1 − t0 is defined by
T + F
πH 1 (1 − πH )2
ρ (t1 − t0 ) = ln − ln + + ln eσ vL aL + eσ vH aL
F πH + τ − σ ( 1 − πH ) ( 1 − πH ) + τ − σ πH

− 12 + 22 NL,t − 2 1 + 4
NH,t = .
21 2 3
This yields
T + F
πH 1 (1 − πH )2
t1 = t0 + ρ ln −1
− ρ ln −1
+ + ρ −1 ln eσ vL aL + eσ vH aL
F πH + τ − σ ( 1 − πH ) ( 1 − πH ) + τ − σ πH

T + F π σ vH aH − eσ vH aL π
( 1 − )e
t2 = t0 + ρ −1 ln − ρ −1 ln
H H
−σ .
F (1 − πH )eσ vH aH − eσ vH aL πH − (eσ vL aL + eσ vH aL ) πH − πH∗ 1−τ τ −σ

Corollary 3 (Complete specialization) (reminded) Assume that countries are equal-sized and that preferences are symmet-
σ v a τ −σ e σ v L a H
ric (πH = 1 − πH∗ and eσ vL aL = eσ vH aH and eσ vH aL = eσ vL aH ). If πH > πH∗ and πH ≥ e eσL vLL −
aL
+eσ vH aL
(1 + τ −σ )−1 , countries are
T ∗ = 0) and the volume of trade is equal to
completely specialized (nTH = 1 and nH

evH aH evH aL
X T = τ −σ L ( 1 − π ) + π −σ v a ,
τ e
− σ v H a H +e v H a L τ e L H + evL aL
where it is true that
LL∗
XS ≤ XT ≤
L + τ −σ L ∗
∂XT
and ∂π < 0.
Proof. Under the stated assumptions, nH = 1, n∗H = 0 and
P (vH )−σ = Neσ vH aH + τ −σ N ∗ eσ vH aL , P (vL )−σ = Neσ vL aH + τ −σ N ∗ eσ vL aL ,

P ∗ (vH )−σ = τ −σ Neσ vH aH + N∗ eσ vH aL , and P∗ (vL )−σ = τ −σ Neσ vL aH + N∗ eσ vL aL
hold. Free entry in Home and Foreign satisfies

eσ vH aH eσ vL aH −σ ∗ ∗ e
σ vH aH eσ vL aH
(F + T )δσ = L π + (1 − π ) +τ L π + (1 − π )
∗
P (vH )−σ P (vL )−σ P ∗ (vH )−σ P ∗ (vL )−σ
σ vH aL

∗ e eσ vL aL −σ eσ vH aL eσ vL aL
(F + T )δσ = L π ∗
∗
+ (1 − π )∗
+τ L π + (1 − π )
P (vH )−σ P ∗ (vL )−σ P (vH )−σ P (vL )−σ
under the stated assumptions of symmetry, P (vH )−σ = P ∗ (vL )−σ and P (vL )−σ = P ∗ (vH )−σ , which solves for N = (F +TL )δσ =
N ∗ . This solves for exports as stated in the corollary.
Supplementary material
Supplementary material associated with this article can be found, in the online version, at 10.1016/j.euroecorev.2017.05.
010.
References
Aizenman, J., Brooks, E., 2008. Globalization and taste convergence: the cases of wine and beer. Rev. Int. Econ. 16 (2), 217–233. 05
Anderson, S.P., de Palma, A., Thisse, J.-F., 1987. The CES is a discrete choice model? Econ. Lett. 24 (2), 139–140.
Anderson, S.P., de Palma, A., Thisse, J.-F., 1992. Discrete Choice Theory of Product Differentiation. MIT Press, Cambridge.
Anderson, S.P., van Wincoop, E., 2004. Trade costs. J. Econ. Lit. 42 (3), 691–751. American Economic Association.
Atkin, D.G., 2013. Trade, tastes and nutrition in India. Am. Econ. Rev. 103 (5), 1629–1663. August 2013.
Auer, R. A., 2010. Consumer Heterogeneity and the Impact of Trade Liberalization: How Representative is the Representative Agent Framework?Working
Papers 2010-13, Swiss National Bank.
Baier, S.L., Bergstrand, J.H., 2001. The growth of world trade: tariffs, transport costs, and income similarity. J. Int. Econ. 53 (1), 1–27.
Behrens, K., Lamorgese, A.R., Ottaviano, G.I.P., Tabuchi, T., 2009. Beyond the home market effect: market size and specialization in a multi-country world. J.
Int. Econ. 79 (2), 259–265.
Bergstrand, J.H., 1990. The Heckscher–Ohlin–Samuelson model, the linder hypothesis and the determinants of bilateral intra-industry trade. Econ. J. 100
(403), 1216–1229.
Bernard, A.B., Jensen, J.B., 1999. Exceptional exporter performance: cause, effect, or both? J. Int. Econ. 47 (1), 1–25. Elsevier.
Bernasconi, C., 2013. Similarity of Income Distributions and the Extensive and Intensive Margin of Bilateral Trade Flows. Working Papers 115, ECON –
Department of Economics – University of Zurich.
Berry, S., Levinsohn, J., Pakes, A., 1995. Automobile prices in market equilibrium. Econom. Econom. Soc. 63 (4), 841–890.
Berry, S., Levinsohn, J., Pakes, A., 1999. Voluntary export restraints on automobiles: evaluating a trade policy. Am. Econ. Rev. 89 (3), 400–430.
BEUC, 1992. Parallel imports for cars in the EC. BEUC/222/92, 1992, Brussels.
Blonigen, B.A., Soderbery, A., 2010. Measuring the benefits of foreign product variety with an accurate variety set. J. Int. Econ. 82 (2), 168–180.
Brenkers, R., Verboven, F., 2006. Liberalizing a Distribution System: the European Car Market Randy Brenkers and Frank Verboven. J. Eur. Econ. Assoc. 4 (1),
216–251.
Broda, C., Weinstein, D.E., 2006. Globalization and the gains from variety. Q. J. Econ. 121 (2), 541–585. MIT Press.
Brülhart, M., Trionfetti, F., 2009. A test of trade theories when expenditure is home biased. Eur. Econ. Review 53 (7), 830–845.
Burstein, A., Melitz, M., 2013. Trade liberalizations and firm dynamics. In: Advances in Economics and Econometrics: Volume 2, Applied Economics: Tenth
World Congress, 50. Cambridge University Press, p. 283.
Chaney, T., 2005. Productivity Overshooting: The Dynamic Impact of Trade Opening with Heterogeneous Firms. Mimeo, University of Chicago.
Cosar, K., Grieco, P. L. E., Li, S., Tintelnot, F., 2015. What drives home market advantage? Mimeo, University of Chicago, February 2017.
Crozet, M., Trionfetti, F., 2008. Trade costs and the home market effect. J. Int. Econ. 76 (2), 309–321.
Cunat, A., Maffezzoli, M., 2007. Can comparative advantage explain the growth of US trade. Econ. J. R. Econ. Soc. 117 (4), 583–602.
Davis, D.R., 1998. The home market, trade, and industrial structure. Am. Econ. Rev. 88 (5), 1264–1276.
Davis, D.R., 2003. Market access, economic geography and comparative advantage: an empirical test. J. Int. Econ. 59 (1), 1–23.
Davis, D.R., Weinstein, D.E., 1999. Economic geography and regional production structure: An empirical investigation. Eur. Econ. Rev. 43 (2), 379–407.
Deardorff, A.V., 1998. Determinants of bilateral trade: does gravity work in a neoclassical world?. In: Frankel, J.A. (Ed.) The Regionalization of the World
Economy. University of Chicago Press, Chicago, pp. 7–22.
Dixit, A.V., Stiglitz, J.E., 1977. Monopolistic competition and optimum product diversity. Am. Econ. Rev. 67 (3), 297–308.
Evans, C.L., 2001. Home Bias in Trade: Location or Foreign-ness? Staff Reports 128, Federal Reserve Bank of New York.
Fajgelbaum, P., Grossman, G., Helpman, E., 2011. Income distribution, product quality and international trade. J. Polit. Econ. 119 (4), 721–765.
Feenstra, R.C., Markusen, J.R., Rose, A.K., 2001. Using the gravity equation to differentiate among alternative theories of trade. Can. J. Econ. 34 (2), 430–447.
Canadian Economics Association.
Fieler, A.C., 2011. Non-homotheticity and bilateral trade: evidence and a quantitative explanation. Econometrica 79 (4), 1069–1101.
Föllmi, R., Hepenstrick, C., Zweimueller, J., 2008. Income Effects in the Theory of Monopolistic Competition and International Trade. Mimeo, University of
Zurich.
Gabaix, X., Laibson, D., Li, D., Li, H., Resnick, S., de Vries, C.G., 2010. Competition and Prices: Insights from Extreme Value Theory. Mimeo, NYU Stern.
Gabaix, X., Laibson, D., Li, D., Li, H., de Vries, C.G., 2006. On extreme value theory and market demand. Mimeo, MIT.
Goldberg, P.K., Verboven, F., 2001. The evolution of price dispersion in the european car market. Rev. Econ. Stud. 68 (4), 811–848.
Goldberg, P.K., Verboven, F., 2005. Market integration and convergence to the law of one price: evidence from the european car market. J. Int. Econ. 65 (1),
49–73.
Hallak, J.C., 2010. A product-quality view of the linder hypothesis. Rev. Econ. Stat 92 (3), 453–466.
Hanson, G., Xiang, C., 2004. The home-market effect and bilateral trade patterns. Am. Econ. Rev. 94 (4), 1108–1129.
Helpman, E., Melitz, M., Rubinstein, Y., 2008. Estimating trade flows: trading partners and trading volumes. Q. J. Econ. 123 (2), 441–487.
Holmes, T.J., Stevens, J.J., 2005. Does home market size matter for the pattern of trade? J. Int. Econ. 65 (2), 489–505.
Hummels, D., 2007. Transportation costs and international trade in the second era of globalization. J. Econ. Perspect. 21 (3), 131–154.
Imbs, J., Mejean, I., 2015. Elasticity optimism. Am. Econ. J. Macroecon. 7 (3), 43–83.
Keith, H., Mayer, T., Ries, J., 2002. On the pervasiveness of home market effects. Economica 69 (275), 371–390.
Keith, H., Ries, J., 2001. Increasing returns versus national product differentiation as an explanation for the pattern of u.s.-canada trade. Am. Econ. Rev. 91
(4), 858–876.
Krugman, P., 1980. Scale economies, product differentiation, and the pattern of trade. Am. Econ. Rev. 70 (5), 950–959.
Linder, S.B., 1961. An Essay on Trade and Transformation. John Wiley & Sons, New York.
Markusen, J.R., 1986. Explaining the volume of trade: an eclectic approach. Am. Econ. Rev. 76 (5), 1002–1011.
Matsuyama, K., 1992. A simple model of sectoral adjustment. Rev. Econ. Stud. 59, 375–388.
Matsuyama, K., 20 0 0. A Ricardian model with a continuum of goods under non-homothetic preferences: demand complementarities, income distribution
and north-south trade. J. Polit. Econ. 108, 1093–1120.
McFadden, D.L., 1981. Econometric models of probabilistic choice. In: Manski, C.F., McFadden, D.L. (Eds.), Structural Analysis of Discrete Data with Econo-
metric Applications. MIT Press, Cambidge, pp. 198–242.
Melitz, M.J., 2003. The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica 71 (6), 1695–1725.
Ruhl, K.J., 2008. The International Elasticity Puzzle. Mimeo, New York University.
Schott, P.K., 2004. Across-product versus within-product specialization in international trade. Q. J. Econ. 119 (2), 647–678.
Uy, T., Yi, K.-M., Zhang, J., 2013. Structural change in an open economy. J. Monet. Econ. 60 (6), 667–682.
Verboven, F., 1996. International price discrimination in the european car market. RAND J. Econ. 27 (2), 240–268. Summer, The RAND Corporation.
Weder, R., 2003. Comparative home-market advantage: an empirical analysis of british and american exports. Rev. World Econ. 139 (2), 220–247.
Yi, K.-M., 2003. Can vertical specialization explain the growth of world trade? J. Polit. Econ. 111 (1), 52–102.
Yi, K.-M., 2010. Can multi-stage production explain the home bias in trade? Am. Econ. Rev. 100 (1), 364–393. March 2010.

Product Heterogeneity, Cross-Country Taste Differences, and

Uploaded by

Copyright:

Available Formats

You might also like

Product Heterogeneity, Cross-Country Taste Differences, and

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Product Heterogeneity, Cross-Country Taste Differences, and

Uploaded by

Copyright:

Available Formats

European Economic Review 100 (2017) 1–27

Contents lists available at ScienceDirect

European Economic Review

Product heterogeneity, cross-country taste differences, and

2. A model of demand for heterogeneous products

2.2. Demand and consumer welfare

Mi = (1 − α )/pM,i and Oi = α /pO ,

Proof. See Appendix

E (Ui ) = (P (v ) )−α θi

Proof. See Appendix

2.4. Equilibrium in the closed economy

3. The immediate aftermath of liberalization

Proof. See Appendix

Proof. See Appendix

4. The long-run effects of trade liberalization

4.1. The case of incomplete specialization

(aH ) = (aL ) and ∗ (aH ) = ∗ (aL )

nTH ≥ nAH ≥ nAH∗ ≥ nH

Proof. See Appendix.

Proof. See Appendix

4.2. The transition to the open economy steady state

export selection costs T are suﬃciently large such that

Proof. See Appendix.

4.3. The case of complete specialization

Proof. See Appendix.

5. Taste differences, liberalization and trade in the European automotive industry

1970 1980 1990 2000

Domestic sales Imports from other nations

1970 1980 1990 2000

1970 1980 1990 2000

1970 1980 1990 2000

1980 1985 1990 1995 2000

Imports vs dom. prod. Dom. prod. across markets

1980 1985 1990 1995 2000

A1. Coeﬃcient estimates

Ui = max O1i −α (qi, j∗ (i ) exi, j∗ (i) +a j∗ (i) vi )α − λi [Oi pO +qi, j∗ (i ) p j∗ (i ) − θi ]

 is distributed Frechet with scale parameter ( p−σ eσ a j vi ) σα and shape parameter σ .

it thus holds that

< 0 if πH + πH∗ < 1

P T (vH )−σ e σ a H v H − e σ a L v H πH P T ∗ (vH )−σ eσ aH vH − eσ aL vH πH∗

NL,t = nAL N A e−ρ (t−t0 )

Proof. Note that:

Next, deﬁning 1 = eσ vH aH + τ −σ eσ vL aH > 2 = eσ vL aH + τ −σ eσ vL aL it holds that

Proof. Under the stated assumptions, nH = 1, n∗H = 0 and

P (vH )−σ = Neσ vH aH + τ −σ N ∗ eσ vH aL , P (vL )−σ = Neσ vL aH + τ −σ N ∗ eσ vL aL ,

You might also like

E (Ui ) = (P (v ) )−α θi

is distributed Frechet with scale parameter ( p−σ eσ a j vi ) σα and shape parameter σ .

Next, deﬁning 1 = eσ vH aH + τ −σ eσ vL aH > 2 = eσ vL aH + τ −σ eσ vL aL it holds that