- Dynamic Panel Estimate (DPE):
TACT_YEMEN (Trade Assessment Computerized Tool for Yemen) is an econometric
model constructed by TAC for the European Commission in order to help Yemenite
authorities to asses the potential implications of an entry in the WTO.
Because
of Yemen’s reunification at the beginning of the 90s, no statistical information
could be used on years prior to 1990 or 1991. Most series were available up
to 2000, leaving only about 10 observations, far too little for ordinary least
square (OLS) estimates that would include more than one explanatory variable.
TAC had therefore to find other techniques enabling statistical analysis of
shorter time series.
The econometric method most often used in such circumstances
is a relatively new statistical estimation technique called Dynamic Panel Estimate
(DPE): DPE is based on OLS-type of analysis, but applied to a pool of countries
over a period of time. The technique allows to derive from such a dynamic observation
of a panel of countries the elements of causality that are common to all observations,
while allowing for differences in the level of the variables for each country.
DPE is best used either on a very large panel of countries, or on a more limited
panel of “comparable” countries. This second solution was preferred
for obvious statistical reasons, as the database of explained / explanatory
variables has to be built for all countries included in the panel. After discussion
in Yemen with the authorities, as well as after a check on available statistics,
it was decided to include, beside Yemen, Syria and Jordan, and as a “test”,
Iran. The idea behind the selection is a geographical proximity, a large dependence
on commodity exports (oil and gas for Yemen and Syria, phosphate for Jordan),
tourism earnings and workers’ remittances, and a small manufacturing base.
- Self Organizing Maps:
The picture presented here is a graphic representation of
a large population of companies. This “map” was created by using
a database of financial accounts on 32,000 French companies on which we
applied a specialized neural network, a self-organizing map (SOM). This
advanced statistical technique for data treatment and classification allows
to divide the sample into clusters and reduce the number of dimensions of
the input data. The image represents the clusters with different levels
of grey color: a white zone is an homogeneous zone (or cluster) and a black
zone is a "hole" between two or more clusters.
- Survey analysis:
TAC was appointed by the European Commission (2003-2004) to study the integration
process within ASEAN and its consequence on European companies. One the component
of the study was a survey to EU companies operating in ASEAN. About 125 companies,
from 6 different ASEAN countries, were surveyed, and, beyond the standard presentation
of the results, a limited statistical exercise was conducted in order to examine
if the answering modalities to two key sets of questions (importance of regulatory
development, obstacles faced when targeting ASEAN) can be attributed to specific
groups of respondents. Obviously, the results had be treated cautiously as the
overall sample is not quite large enough to ensure a strict statistical representation.
However, the PCA (Principal Component Analysis) suggested some interesting remarks
pertaining to the differences in perception according to the size of the company
and its geographical location within ASEAN countries.
Click to enlarge
|
Return to the top
