developing countries in order to comply with SPS standards and hence increase or
maintain their market access. Until August 2003 (which is the last online update of the
STDF database) funding was granted to 151 different developing countries, 46 of them
are LDCs. Only three LDCs did not receive a grant, Somalia, Haiti and Timor Lesté.
The total grants amounted to more than $ 8.5 billion. Kenya is with more than $ 3 billion
by far the top receiving country of STDF grants, followed by Iran (649 million), Pakistan
(410 million) and the two LDC countries Nepal (371 million) and Bhutan (386 million).
Nevertheless, 25 LDCs rank on the end of the countries list with total grants lower than
$ 20 thousand.
In chapter 6 data on border rejections and STDF investments will be analyzed in relation
with the export performance of countries.
4 Methodology and Data
The previous sections explained the importance of standards. The remaining part of the
paper empirically analyses patterns of developing countries’ performance in agricultural
exports and possible links between export performance and standards.
The analysis is based on trade data of 73 7 developing countries taken from the PC- TAS
data base [8]. To describe the development of developing countries’ trade performance
and its relation to standards, data on export values of two commodity groups for two
time spans is collected: meat and fruits/ vegetables in the years 1993- 1995 (before the
SPS Agreement) and 2002- 2004 (after the SPS Agreement). Exports to OECD countries
are selected since these countries are seen as “standard setters” 8. The sectors of meat
and fruit/ vegetable are chosen because these markets are highly affected by standards.
For reasons of better data quality, imports of OECD countries from each developing
country are used to describe developing countries’ exports.
For the statistical analysis, four variables have been developed describing the export
performance of the individual country 9, the "average", the "ratio", the "difference" and
the "coefficient of variation". All of them are explained as follows:
In a first step, the average export values of the individual country are calculated for two
time periods 1993- 1995 and 2002- 2004. In a second step, two variables are calculated
from the average trade values: the “ratio” and the “difference”. The ratio takes into
account the average value of exports for the respective commodity group in 1993- 1995
and 2002- 2004. It describes the dynamics of export performance without taking into
consideration the absolute level. However, it must be noted that the ratio is sensitive to
the absolute volume of trade, since e.g. a doubling of exports starting from a very low
initial value is much more likely to occur. The second variable, therefore describes the
difference between the average value of exports for the respective commodity group in
1993- 1995 and 2002- 2004. It takes into account the absolute level of exports. Thus
especially large countries' relatively small percentage changes in export value are
captured better, if looking at the absolute value. Finally, the coefficient of variation is
calculated for the period 2002- 2004 to gain an idea about the stability of exports of a
country. It would be interesting to compare the variability in the two time spans, but the
variable has several missing values in the first period, if single years are not reported
and therefore would reduce the sample.
These variables will be used to group countries according to their export performance in
a cluster analysis (compare section 5). The method of cluster analysis can be used for an
exploratory, empirical classification of objects according to their similarity. The
objective of the cluster analysis in this paper is to identify patterns, or groups, of
developments in export performance across countries. The analysis is conducted for the
7 84 developing countries - more than half of all - were not included in the analysis because of a
lack of data.
8 In this respect, it would be interesting to compare the development of exports from South to
North with those from South to South, or from South to “East”, thus in countries, where standards
are not as strict. However, this was not analyzed due to lack of adequate data.
9 Missing data in PC- TAS were treated as such. An alternative would be to treat no trade records as
a trade volume of 0 for the respective pair of trading partners. In our approach any country with
missing data in all years of the first time span or less then two observations in the second time
span is excluded from the analysis. We do not hurt statistical requirements since we do not deal
with a random sample anyway (which is not required for cluster and factor analysis). However, we
slightly overestimate the average trade value and slightly underestimate the variation in trade
volumes for countries that had no records in single years of the analysis. This only affects very
small countries. Nevertheless, it has the effect, that only 17 LDCs were included in the analysis.
7