Article
Reference
Trade Diversification, Income, and Growth: What Do We Know?
CARRERE, Céline, STRAUSS‐KAHN, Vanessa, CADOT, Olivier
Abstract
This paper surveys the empirical literature on export and import diversification and its linkages
with growth. We review widely used measures of diversification and the evidence about their
evolution focusing on how export diversification relates to trade liberalization and economic
development. We also discuss the linkages between trade diversification and productivity at
the firm and industry level, highlighting new advances on the linkages between import
diversification and productivity.
CARRERE, Céline, STRAUSS‐KAHN, Vanessa, CADOT, Olivier. Trade Diversification, Income,
and Growth: What Do We Know? Journal of Economic Surveys, 2013, vol. 27, no. 4, p.
790-812
DOI : 10.1111/j.1467-6419.2011.00719.x
Available at:
http://archive-ouverte.unige.ch/unige:46655
Disclaimer: layout of this document may differ from the published version.
1 / 1
1
Trade Diversification, Income, and
Growth: What Do We Know?
April 2010
Céline Carrère+
Vanessa Strauss-Kahn§
Olivier Cadot∗
Keywords: Export diversification, International trade, Growth and
employment
JEL classification codes: F1, O11
+ CERDI-CNRS, Université d’Auvergne.
§ ESCP-EUROPE and CEPR.
∗ The World Bank and HEC Lausanne, CERDI, CEPR and CEPREMAP.
2
1. Introduction
Policy interest in export diversification is not new but, as noted by Jose
Salazar-Xirinachs in his comment on Cimoli, Dosi and Stiglitz (2009), for
over two decades it was mired in an ideologically-loaded debate about the
role of the State. Old-time industrial policy having died of its own excesses,
the debate over what, if anything, the government should do to promote
export growth was contained to the fringe of the economics profession.
Mainstream economists were happy to believe that whatever market
failures there were out there, government failures were worse, and that
anyway most governments in developing countries lacked the means to do
anything. But by an ironic twist of history, years of (Washington-consensus
inspired) fiscal and monetary discipline have put a number of developing-
country governments back in a position to do something for export
promotion, having recovered room of maneuver in terms of both external
balance and budget position. So the question is back.
With limited guidance from theory, the economics profession’s answer to
the return of the industrial-policy debate has been to go back to descriptive
statistics (as opposed to the investigation of causal chains). The result is a
wealth of new stylized facts. For instance, surprising patterns of export
entrepreneurship have emerged from the use of increasingly disaggregated
data. Decompositions of export growth into intensive and extensive
margins have revealed interesting patterns, and so has the study of export
survival.
One area where theory has proved useful is in the exploration of the links
between productivity and trade. So-called “new-new” trade models
(featuring firm heterogeneity) have highlighted complex relationships
between trade diversification and productivity, with causation running one
way at the firm level and the other way around (or both ways) at the
aggregate level.
Even at the aggregate level, new issues have appeared. First, Imbs and
Wacziarg (2003) uncovered a curious pattern of diversification and re-
concentration in production, prompting researchers to explore whether the
same was true of trade. Second, a wave of recent empirical work has
questioned traditional views on the “natural-resource curse”, challenging
the notion that diversification out of primary resources is a prerequisite for
growth.
Thus, our current understanding of the trade diversification/
productivity/growth nexus draws on several theoretical and empirical
3
literatures, all well developed and growing rapidly. It is easy to get lost in
the issues, and the present paper’s objective is to sort them out and take
stock of elements of answers to the basic questions.
Among those questions, the first are simply factual ones—how export
diversification is measured and what are the basic stylized facts about trade
export diversification, across time and countries, which we explore in
Section 2 and 3 respectively. The third one is about diversification’s drivers,
and is tackled in Section 4. In Section 5, we turn to the relationship between
diversification and growth. Section 6 focuses on the import side; we review
the evidence on import diversification and productivity and extend the
discussion to labor-market issues. In Section 7, we consider some tentative
policy implications and conclusions.
2. Measuring diversification
2.1 Overall indices
Although much of the talk is about trade diversification, quantitative
measures, most of them borrowed from the income-distribution literature,
are about concentration. We will review these measures taking the example
of export diversification (which has anyway been the focus of most papers)
keeping in mind that they apply equally well to imports. All concentration
indices basically measure inequality between export shares; these shares, in
turn, can be defined at any level of aggregation. Of course, the finer the
disaggregation, the better the measure.
The most frequently used concentration indices are the ones used in the
income-distribution literature: Herfindahl, Gini, and Theil. These indices
are formalized in the technical Appendix 8.1.1. All three indices can be
easily programmed but are also available as packages in Stata. Authors
have used one or several of these measures. Across the board, results are
not dependent on the index chosen.
The Theil index has decomposability properties that make it especially
useful. It can indeed be calculated for groups of individuals (export lines)
and decomposed additively into within-groups and between-groups
components (that is, the within- and between-groups components add up to
4
the overall index).
1
It is thus possible to distinguish an increased
concentration (diversification) that occurs mainly within groups from one
that occurred mainly across groups. We will see in the next section a useful
application of this property in our context.
2.2 Intensive and extensive margins
Export concentration measured at the intensive margin reflects inequality
between the shares of active export lines.2 Conversely, diversification at the
intensive margin during a period
t
0
to
t
1
means convergence in export
shares among goods that were exported at
t
0
. Concentration at the
extensive margin is a subtler concept. At the simplest, it can be taken to
mean a small number of active export lines. Then, diversification at the
extensive margin means a rising number of active export lines. This is a
widely used notion of the extensive margin (in differential form), and the
decomposition of Theil’s index can be usefully mapped into the intensive
and extensive margins thus defined.
Suppose that, for a given country and year, we partition the 5’000 or so
lines making up the HS6 nomenclature into two groups: group one is made
of active export lines for this country and year, and group “zero” is made of
inactive export lines (i.e., export lines for which there are no exports). This
partition can be used to construct within-groups and between-groups
components of the overall Theil index. As shown in the technical appendix
8.2, by distinguishing the Theil sub-index for the group of inactive line from
the Theil sub-index for the group of active lines, changes in
concentration/diversification within and between groups can be set apart.
More importantly, it can be shown that given this partition, changes in the
within-groups Theil index measure changes at the intensive margin
whereas changes in the between-groups Theil index measure changes at the
extensive margin. In sum, Theil’s decomposition makes it possible to
decompose changes in overall concentration into extensive-margin and
intensive-margin changes.
3
This is a particularly important feature as
changes at the intensive margin or extensive margin reflect very different
1
Appendix 8.1.2 presents the Theil index decomposition.
2
An active line corresponds to a non-zero export line of the HS6 nomenclature
(about 5000 lines) for a given year.
3
This mapping between the Theil decomposition and the margins was first
proposed by Cadot et al. (2009).
