1
Do Food Scares Explain Supplier
Concentration?
An Analysis of EU Agri-food Imports
§
January 2010
Mélise Jaud
∗
Olivier Cadot
+
Akiko Suwa Eisenmann+
Abstract
This paper documents a decreasing trend in the geographical concentration of EU agro-
food imports. Decomposing the concentration indices into intensive and extensive
margins components, we find that the decrease in overall concentration indices results
from two diverging trends: the pattern of trade diversifies at the extensive margin (EU
countries have been sourcing their agri-food products from a wider range of suppliers),
while geographical concentration increases at the intensive-margin (EU countries have
concentrated their imports on a few major suppliers). This leads to an increasing
inequality in market shares between a small group of large suppliers and a majority of
small suppliers. We then move on to exploit a database of food alerts at the EU border
that had never been exploited before. After coding it into HS8 categories, we regress the
incidence of food alerts by product on determinants including exporter dummies as
well as HS8 product dummies. Coefficients on product dummies provide unbiased
estimates of the intrinsic vulnerability of exported products to food alerts, as measured
at the EU border. We incorporate the product risk coefficient as an explanatory variable
in a regression of geographical concentration and show that concentration is higher for
risky products.
Keywords: European Union, import concentration, sanitary risk, food,
agricultural trade
JEL classification codes: F1, O13
§
We thank the franco-suiss Egide program Germaine de Stael n°1695SQL. Also without
implicating them, we thank participants to a seminar at the Paris School of Economics, in
particular Anne-Célia Disdier and Lionel Fontagné, for helpful comments on an earlier version.
∗
Paris School of Economics
+
University of Lausanne, CEPR, CERDI and CEPREMAP.
+
INRA, LEA , Paris School of Economics.
2
1. Introduction
After a series of highly publicized food scares (bovine spongiform
encephalopathy or dioxin-contaminated chickens to name but two), public-
health concerns have started to loom large in the buying policies of EU foodstuff
distributors. These concerns have the potential to affect the evolution of EU
foodstuff imports, and therefore the access that developing countries—in
particular the poorest ones, who find it most difficult to comply with stringent
sanitary standards—enjoy on EU markets (Maskus et al. 2005). This is
particularly important for products like fruit & vegetables or fisheries products,
which represent a growing share of EU food imports and are also of particular
concern to the least developed countries (Henson 2004, 2005, 2006, Jaffee et al
2004, World Bank 2005).
The impact of sanitary concerns on industrial-country foodstuff imports has
been studied extensively, essentially by sticking standards as explanatory
variables in gravity equations (see e.g. Moenius 1999, Maskus et al. 2000,
Otzuki and Wilson 2001, Czubala and al 2007). Estimation of such models has
highlighted the trade restrictiveness of such standards (Fontagné et al 2005,
Disdier, Fontagné et al. 2008).
We differentiate ourselves from the existing literature in two ways. First, we
shift focus from gravity modelling to an analysis of the geographical
concentration of EU foodstuff imports, using conventional and non-
conventional concentration measures (similar approaches can be found in Imbs
and Wacziarg 2003 for production or in Cadot, Carrère and Strauss-Kahn 2007,
Armugo and al 2008 and Dutt, Mihov and van Zandt 2008 for exports). We
propose a decomposition of Theil’s entropy index between active and potential
suppliers with the property that variations in the index’s within- and between-
group components map directly into intensive- and extensive-margin variations.
We also propose a variant of Hummels and Klenow’s intensive and extensive
margins (which they developed for the analysis of the product-wise
concentration of exports) adapted to imports and to geographical concentration.
Our product-level analysis shows that, over the last two decades, EU foodstuff
imports have concentrated, geographically, at the intensive margin. That is, on
average, at the product-line (HS8) level, the market shares of active suppliers
have diverged. However, we also observe a trend toward diversification at the
extensive margin. That is, again at the product-line level, the EU sources its
3
foodstuffs from an increasing number of exporting countries. These two
observations appear, at first sight, to be contradictory. Using our adaptation of
Hummels and Klenow’s intensive and extensive margins, we show that the
number of suppliers used by the EU is indeed increasing, but by addition of a
fringe of small-volume exporters. Thus, EU foodstuff imports gradually evolve
toward a two-tier distribution with a small number of increasingly dominant
suppliers and a growing fringe of marginal ones.
Second, we shift focus from standards, which affect trade flows ex-ante, to
alerts, which affect them ex-post. What we call here a “food alert” is the
notification of a contaminated foodstuff shipment at the external border of an
EU member state. Food alerts have the power to alter buyer perceptions of the
quality of particular suppliers. They could either lead to reinforced
concentration if buyers react by eliminating fringe suppliers perceived as
dubious by analogy with the culprit, or, alternatively, to the destruction of
dominant positions if they affect dominant suppliers. This is what we explore,
using an original database constructed from the European Commission’s Rapid
Alert System for Food and Feed (RASSF) database. Technically, the EU
Commission classifies contaminated-shipment notifications into two types:
“informations”, which lead to the destruction or re-routing of the concerned
shipment, or “alerts” stricto sensu which lead to the destruction or re-routing of
all shipments from the same exporting country at all EU borders. Since 2001, all
informations (about 19’000 of them) have been recorded in a detailed database,
which has never been used. We coded that database into HS8 product categories
to make it compatible with trade data, generating a population of events each
defined at the (product × exporter × year) level.
The RASFF database shows substantial heterogeneity in the incidence of food
alerts across exporting countries. This implies that a raw count of alerts by
product cannot give a correct proxy for product-specific sanitary risk. For
instance, a product imported overwhelmingly from a country with weak quality
standards would appear as risky even though other exporters might have
managed to make the product safe. In addition, the incidence of notifications is
likely to be correlated with the frequency of controls. Those controls may not be
purely random: they may reflect a particular exporter’s past performance or
hidden protectionism. Thus, regressing concentration indices on the frequency
of notifications at the product level would say nothing without controlling for
other factors.
4
In order to get an unbiased estimate of product-specific sanitary risk, we rely on
a two-step procedure. In the first step, we estimate product-specific sanitary risk
with a regression of the count of food alerts at EU borders over the sample
period, using an original database described in the next section. The unit of
observation is a product × exporter pair where alerts are summed over all years
in the sample period. The regressors are exporting-country characteristics and
product dummies. Estimated coefficients on those product dummies give us an
estimated measure of product risk. In a second step, we regress the evolution of
geographical concentration indices on our measure of product risk and time
dummies.
Overall, we find that except for fisheries products no chapter stands out as
having particularly high risk levels. Incorporating our constructed measure of
product risk as an explanatory variable in a regression of geographical
concentration confirms that product riskiness affects sourcing concentration.
Product riskiness leads to reinforced concentration at the intensive margin and
reinforced diversification at the extensive margin. Thus, the distribution of EU
suppliers for riskiest agrofood imports is converging towards a pattern of
increasingly dominant suppliers with a growing fringe of small-scale ones.
The paper proceeds as follows. The next Section analyses the trend in the
geographical concentration of EU agro-food imports both at the intensive and
extensive margin. We then outline the EU "Food Alerts" Database in Section 3,
contrast it with previous data collection efforts, and present some descriptive
results. Section 3 then explores the impact of product riskiness on the patterns
of concentration. Section 4 concludes.
5
2. Agri-food supplier concentration
2.1 Overall diversification
2.1.1 The data
We use EUROSTAT agri-food import data covering EU-12 member states
1
(France, Belgium-Luxembourg, the Netherlands, Germany, Italy, Ireland,
United Kingdom, Denmark, Greece, Portugal and Spain) between 1988 and
2005 at the HS8 level (the highest level of disaggregation available, as Eurostat
does not make 10-digit data available to researchers). Agri-food products,
excluding beverages and animal feed, are in chapters 1 to 21 of the HS system,
which represent 3’073 potential export lines. With 146 partner countries
(exporters) including 122 developing countries, we have a four-dimensional
panel where the unit of observation is a product imported by an EU member
state from an extra-EU partner in a given year.
2
For some calculations, however,
we aggregate import data across EU member countries, reducing the panel’s
dimension to three (product × exporter × year).
At the HS8 level, reclassifications are frequent. Five types of reclassification can
be distinguished: (i) creation of a new code corresponding to a new product; (ii)
creation of several new codes by splitting a former one; (iii) creation of a new
code by merging several former ones; (iv) creation of new codes resulting from a
change in the coding system (HS harmonizations in 1988, 1996 and 2002); and
finally (v) termination of old codes. Of the 3’073 HS8 codes available in our
dataset, only 37.7% are unaffected by reclassification between 1988 and 2005.
Of the remainder (62.3%), 1.6% are new products (type i), and 0.7% are
terminated codes (type v). This leaves 60% of reclassifications of “continued”
1
We use this restrictive definition for consistency of time series, as EUROSTAT does not provide
data on member states before their accession.
2
We drop intra-EU trade on the ground of the mutual recognition of standard. The principle
ensures that a product lawfully produced in one Member State is acceptable without adaptation
in another Member State, provided that both states pursue the same general objectives in health
