December 2004
How Important is the New Goods Margin in International Trade?
Timothy J. Kehoe
University of Minnesota
Federal Reserve Bank of Minneapolis
Kim J. Ruhl
University of Texas at Austin
ABSTRACT______________________________________________________________
We examine the bilateral trade patterns of countries involved in significant trade
liberalizations using detailed data on the value of trade flows by commodity. We find a
striking relationship between a good's pre-liberalization share in trade and its growth
subsequent to liberalization. The goods that were traded the least before the liberalization
account for a disproportionate share in trade following the reduction of trade barriers. The
set of goods that accounted for only 10 percent of trade before the liberalization may
account for as much as 40 percent of trade following the liberalization. This new finding
cannot be accounted for by the standard models of trade, which rely on increases in
previously traded goods to produce trade growth. We modify the standard Dornbusch-
Fischer-Samuelson model of Ricardian trade to provide a model capable of delivering these
new facts. Our specification improves on previous Ricardian models by providing a
technology process that can be calibrated using data on intra-industry trade.
________________________________________________________________________
*© 2004, Timothy J. Kehoe and Kim J. Ruhl. We would like to thank Tom Holmes, Sam Kortum, Ed
Prescott, James Schmitz, Kei-Mu Yi and participants at the IO-Trade and Macro workshops for many valuable
comments. The views expressed herein are those of the authors and not necessarily those of the Federal
Reserve Bank of Minneapolis or the Federal Reserve System.
2
1 Introduction
The real value of trade between two countries can grow in only two ways. The
countries can export more of the goods they had already been trading, which we define as
growth on the intensive margin, or the countries can begin exporting goods they had not been
previously trading, which we define as growth on the extensive margin. In the bulk of applied
international trade models, growth in trade following a reduction in tariffs is driven by
increases in the trade on the intensive margin. The factor-proportions models, as well as
those employing imperfect competition, rely heavily on fixed trade patterns.
1
Few models of
international trade have incorporated an extensive margin, in which goods that were not
previously traded could become traded following a decrease in trade barriers.
In this paper we study the detailed trade statistics of 6 different countries during
significant trade liberalizations to determine the presence and importance of the extensive
margin for trade growth. Our study spans all of North America and most of Europe, and
includes such large-scale trade liberalizations as the North American Free Trade Agreement,
the Canada-U.S. Free Trade Agreement (CAUSFTA) and the accessions of Greece, Spain
and Portugal to the European Economic Community. We construct a new measure of the
extensive margin that takes into account the relative importance of a good in a county’s
trade, rather than imposing fixed dollar value cutoffs for determining whether a good is
traded or not.
We find significant evidence of growth in the extensive margin following a decrease
in trade barriers. The set of least traded goods which accounted for only 10 percent of trade
before the trade liberalization may grow to account for as much as 50 percent of trade
following the liberalization. We find extensive margin growth for almost all of the country
pairs we examine. Furthermore, we construct a time series measure and find that the growth
in the extensive margin coincides with the timing of the trade liberalization, supporting our
hypothesis that the extensive margin growth is driven by the trade liberalization and is not
the consequence of other factors, such as the product cycle. In contrast, when we examine
the extensive margin for the United States and trade partners with which there has been no
major trade policy changes, we find little evidence of growth on the extensive margin. This
1
The same holds true for International Real Business Cycle models, models that exploit the Armington
aggregator and others that feature trade in a composite good.
3
suggests that extensive margin growth is brought about by changes in trade policy and
changes little in response to events like business cycles.
The method we develop for measuring the extensive margin represents a substantial
break from the few previous studies of the extensive margin. These other studies use fixed
dollar value cutoffs to determine whether a good is traded or not in a particular period.
Hummels and Klenow (2004) classify a good as not traded if the value of trade is 0, while
Evenett and Venables (2002) classify a good as not traded if its yearly value of trade is 50,000
1985 U.S. Dollars or less, regardless of the country in question. A 50,000 Dollar cutoff
implies that a good trading for 0.03709 percent or less of Nepal’s total trade is not
considered traded, while a good trading for anything more than 0.00018 percent of total
trade in China is counted as traded.
2
In our measure of tradedness, we allow the actual
dollar value cutoff to differ across countries, relying instead on the relative importance of
these goods in a country’s trade.
Evenett and Venables (2002) considers the extensive margin while studying the
geographic distribution of exports in developing countries. They find that a significant
fraction of a developing country’s trade growth can be attributed to exports of “long
standing exportables” to new destinations. Thus, their concept of the extensive margin is a
cross-country aggregation of our bilateral concept. While they study a country’s exports to
many partners at two different points in time, we concentrate on single country pairs and
construct a measure that allows us to study the extensive margin across all the years in our
sample.
Hummels and Klenow (2004) uses detailed trade data to decompose a nation’s trade
into an extensive component and an intensive component for a large cross-section of
countries. They find that the extensive margin is important in explaining why big countries
trade more than small countries, in that, big countries trade more kinds of goods than
smaller countries. Hillberry and McDaniel (2002) uses the Hummels and Klenow
decomposition to study the growth in trade between the United States and its NAFTA
partners. They find growth in both the intensive and extensive margins. We extend the
decomposition in Hummels and Klenow (2004), in a way similar to that in Hillberry and
2
China’s 1985 value of total trade was $27 billion, while Nepal’s was only $134 million.
4
McDaniel (2002), to create a time series analog that is comparable to our measure of the
extensive margin.
Following the methodology laid out here, Mukerji (2004) studies the liberalization of
trade in India in the 1990s. Growth in the extensive margin is found in both Indian exports
and imports. Sandrey and van Seventer (2004) also use the methodology developed here to
study the liberalization of trade brought about by the Closer Economic Relationship
between Australia and New Zealand starting in 1988. They find evidence that the extensive
margin was growing for New Zealand exports to Australia during this period while the
export share of these goods from New Zealand to the rest of the world was relatively stable.
Their results in this respect are similar to ours in that extensive margin growth seems to
coincide with trade liberalization.
The few models developed to incorporate the extensive margin have generally fallen
into two classes. The first class of models features heterogeneous firms who produce
differentiated products, who must pay a fixed cost in order to export and whose exports are
subject to a trade barrier, usually a tariff, as in Melitz (2003), Alessandria and Choi (2003)
and Ruhl (2003). The firms are heterogeneous in their productivity; low productivity firms
could not make enough profits from exporting to cover the fixed cost, so they do not
export; high productivity firms can generate enough export profits to cover the fixed cost, so
they do export. These firms that do not export are producing nontraded goods. As trade
barriers fall, the potential profits of nonexporting firms increase and, for some firms, these
profits increase enough to cover the fixed export cost, inducing these firms to begin
exporting. The firms that switch from not exporting to exporting are creating extensive
margin growth in response to the change in trade barriers.
The second class of models are built around Ricardian models with many goods, as
in Dornbusch, Fischer and Samuelson (1977). In these models each country can produce
any good in the product space, but the counties have different productivities in the
production of each good. Since the same good can be produced in either country, in the
absence of trade barriers, a good is either produced in one country or the other and
exported. When there are trade barriers, there will be goods that neither country can
produce at a cost low enough to export, so both countries produce the nontraded good for
domestic consumption. As trade barriers fall in these models, it becomes possible for some
of these goods to be produced at a low enough cost in one country so that, given the lower
5
tariff, it is optimal to produce the good in only one country and export it. These goods
which go from being nontraded, and produced in each country, to being traded, and
produced in only one country, show up as extensive margin growth in response to a decrease
in trade barriers.
In this paper we provide a simple model that can produce growth in both the
extensive and intensive margins. We do so by modifying a standard Ricardian model with a
continuum of goods, as in Dornbusch et al. (1977), by relaxing the ordering of goods in the
product space. Rather than impose an ordering based on productivity, we order the goods
according to their Standard International Trade Classification (SITC) number. This ordering
has two advantages. First, the SITC ordering is constructed to group similar items together,
which is a characteristic we will exploit in our specification of technology. Second, an SITC
aggregate (such as a 4-digit subgroup) is simply a closed interval in our product space, so it is
conceptually straightforward to map our results back to the data. A country’s productivity in
producing a particular good is random. This specification yields both intra-industry trade and
growth in the extensive margin in a model that can be calibrated. Intensive margin growth is
driven by the combination of constant elasticity of substitution preferences and an elasticity
greater than one.
Besides being able to produce the extensive margin, the model we present is also
easily calibrated using readily available data. Recent attempts at calibrating Ricardian style
models include Yi (2003), who uses the idea of revealed comparative advantage and Kraay
and Ventura (2002), who use data on wages and education to calibrate the distribution of
relative productivities . Eaton and Kortum (2002) and Bernard, Eaton, Jensen and Kortum
(2003) use data on the pattern of world to trade to calibrate a multicountry Ricardian model
of world trade. In that we also assume a distribution over relative productivities, and use
data on trade patterns, our approach is similar to that of Eaton and Kortum (2002).
Section 2 defines the measures we will use to study new good trade. Section 3
presents the evidence on the extent of trade in new goods following 5 trade liberalizations
and section 4 compares our measure to others found in the literature. In section 5 we
modify a Ricardian model to produce a model capable of delivering the growth in the
extensive margin that can be calibrated. We calibrate the model to the Mexican NAFTA
experience in section 6, and extend the model to include an intensive margin as well as an
extensive margin in section 8. We perform sensitivity analysis in section 9.
