Trade Elasticities

• Date: 01 May 2017
• Author: Jean Imbs,Isabelle Mejean
• Published date: 01 May 2017
• DOI: http://doi.org/10.1111/roie.12270

Trade Elasticities

Jean Imbs and Isabelle Mejean*

Abstract

Conventional aggregate trade elasticity estimates hardly vary across countries. We introduce an aggregate

elasticity that is implied by theory: It is the value that equates the welfare gains from trade as implied by

one- and multi-sector versions of the model in Arkolakis et al. (American Economic Review,102

(2012):94–130). These estimates are predicated on sector-level values for trade elasticites, which we

provide at three-digit levels for 28 developed and developing countries. The values for this aggregate

elasticity vary greatly across countries, and they do so because of countries’ patterns of production and

because a given sector-level elasticity displays considerable cross-country heterogeneity.

1. Introduction

Estimates of the price elasticity of imports obtained form aggregate data are close to

zero and hardly display any differences across countries. For instance, Houthakker

and Magee (1969) estimate price elasticities of imports for 15 developed economies

and find no two estimates are significantly different from each other. Little has

changed since then: the aggregate price elasticity of imports is a small number, close

to zero from below, and it tends to be so for most countries. This conflicts with the

view that the resilience of domestic production varies across sectors, depending for

instance on the goods’ quality and substitutability, or on the productivity of the firms

producing in each sector. Under such a view, it is natural and intuitive to expect that

the aggregate price elasticity of imports ought to vary significantly across countries,

depending on the overall specialization of their production across sectors. For instance,

the aggregate elasticity should be different in a developing country specialized in com-

modities and in a mature diversified economy that produces high-quality goods and

services. Yet, to our knowledge, there are no estimates of aggregate elasticities that

display any meaningful cross-country heterogeneity.

Of course, this fact c ould come from difficulties i n estimating such a paramet er in

the aggregate: because of issues of measurement, of endogeneity, or of aggregation. In

this paper, we propose to infer the aggregate trade elasticity on the basis of a model.

We build from the well known result in Arkolakis et al. (2012) that the welfare gains

from trade depend on two parameters: the price elasticity of trade and the degree of

openness to foreign trade. Estimates of the welfare gains ought to be the same irre-

spective of whether they are computed on sector-level or aggregate data. We use this

equivalence to introduce an “aggregate” trade elasticity such that the welfare gains

from trade implied by the one-sector version of Arkolakis et al. (2012) are equal to

* Imbs: Paris School of Economics, 106–112 boulevard de l’Hopital, Paris, France 75013. E-mail:

jeanimbs@gmail.com. Mejean: Department of Economics, Ecole Polytechnique. 91128 Palaiseau Cedex,

France. Thanks are due to an anonymous referee, to Brent Neiman, and to Andrei Levchenko, for their

discussions of an early version of this paper. The authors are grateful to audiences at the September

2010 San Francisco Fed Pacific Basin Research Conference and the January 2011 American Economic

Association Meetings. All errors are those of the authors.

V

C2016 John Wiley & Sons Ltd

Review of International Economics, 25(2), 383–402, 2017

DOI:10.1111/roie.12270

the welfare gains implied by sector-level estimates of trade elasticities. By definition,

the aggregate trade elasticity is given by an adequately weighted average of sector-

specific trade elasticities. The weights depend on the composition of sector expendi-

tures and on the distribution of openness at sector level. As such, the aggregate trade

elasticity we introduce depends directly on the specialization of countries’ production

and trade.

The exercise enables us to decompose the sources of cross-country differences in

aggregate trade elasticities, into differences in the sizes of sectors, their openness and

their trade elasticity. We find that aggregate trade elasticities are close to the USA in

most developed countries, with the exceptions of the UK, Australia and Canada,

where they are substantially larger in absolute value. Most emerging markets have

much larger elasticities (in absolute value) than the USA, e.g. China, Turkey, Chile or

Slovakia. In both cases, the differences come from high values of trade elasticities at

sector level. Sector-level heterogeneity in trade elasticities has first-order consequences

on the aggregate trade performance of countries.

We conduct the exercise in a panel of 28 developed and developing countries. The

approach builds on three-digit sector-level estimates of trade elasticities for all coun-

tries in the panel. The micro estimates are obtained following the methodology in

Imbs and Mejean (2015), where we focused on the USA. The estimation is based on a

structural model and thus immune to the conventional endogeneity issues that plague

elasticity estimates.

1

The resulting estimates of the aggregate elasticity e

j

range from

23.4 to 29.9, with the lowest values in Cyprus, Chile and China. The cross-country

average equals 25.9. Developed economies have estimates around 25, 24.9 in the

USA, although Canada, the UK, Australia and Greece have substantially larger values

(in absolute value), closer to 26, or even 29 in Greece. The values of e

j

are clearly sig-

nificantly different from classic trade elasticity estimates obtained from aggregate

data.

We decompose these international differences into three components: (i) the disper-

sion in sector-level elasticities across countries, (ii) differences in sectoral openness to

trade, and (iii) differences in the sectoral allocation of expenditures. Trade in most

economies in Western Europe is about as elastic as in the USA (24.9) and the differ-

ences are minimal across all three components of our decomposition: France,

Germany, Hong Kong, Japan have similar aggregate elasticities and also similar

sector-level patterns. Some exceptions are Norway, Sweden, the UK, Australia, Can-

ada and Greece, that all have substantially larger elasticities in absolute value. In all

these cases, the differences reflect the fact that sectoral trade is more elastic than in

the USA, especially in large and open sectors. This seems to correspond to relatively

specialized developed economies—either in commodities, or perhaps in the financial

sector. An exception is Austria, whose estimated elasticity is close to the USA (24.8).

On average, sector-level elasticities in Austria are substantially higher than in other

developed countries, but this is offset by the fact that those sectors with relatively low

elasticities are in fact the big sectors of Austria, both in terms of domestic expendi-

tures and of trade—i.e. the aggregate trade elasticity is low.

Most developing economies have aggregate elasticities that are larger than in the

Organisation for Economic Co-operation and Development (OECD). In most cases,

this happens because sector-level elasticities are larger in absolute value. For instance,

the Chinese trade elasticity is 26:9 and this happens because in China, large and open

sectors are relatively more elastic than in the USA. The same is true of Chile, Turkey

or Slovakia. An interesting exception is Malaysia, whose elasticity is 23.4. This hap-

pens not because trade is inelastic at sector level, but rather because the bulk of final

384 Jean Imbs and Isabelle Mejean

V

C2016 John Wiley & Sons Ltd