Country Models, Sectors and Multiple Authors

[Wolfgang Schade]

Hi Martin and the others,

we have developed the ASTRA model (completed in 2000). It covers whole
Western Europe
divided into 4 regions (e.g. Germany and Austria are one region). It
consists of 4 modules:
macroeconomics, regional economic including population model, transport and
environment.
In a sense it is not a country model as you describe it with models for
e.g. education and
health. But our approach is to include a macro and a micro level. On the
macro level we consider
the whole economy divided into economic sectors (12 sectors e.g.
agriculture,chemicals,trade and
transport services). On the micro level we model in detail the transport
sector with freight and
passenger demand as well as the environmental consequences. Thats the point
to mention
the purpose of the model, which is to analyse the long-tem consequences (30
years) of
European transport policy.

I would say the model is not a pure system dynamics model e.g. in the sense
as described
by the rules and approach in Forresters "Principles of Systems". It is
based on existing non-
system-dynamics models and we are using as far as possible econometrically
tested equations
and apply the Vensim optimizer to calibrate it.

Since 2000 ASTRA is continuously improved into an ASTRA family of models:
# ASTRA-Italy covering only Italy with 25 economic sectors (instead of 12).
There will be a presentation
on this model at the Palermo SD conference in July (by Angelo Martino or
Davide Fiorello).
# ASTRA-E a model to analyse the trade and employment consequences of
European technology
policy
# ASTRA-T a model still under development dividing Western Europe into 14
countries plus Poland
(= 15 country models).

We are also very much interested to get funding to incorporate further
sectors (in your sense) e.g.
energy on the micro level and to extend the time horizon. Nevertheless, in
the European scientific
world funding often depends on a "fashion" and currently CGE models seem to
be more fashionable
than system dynamics (it would be interesting to hear how this is outside
Europe).

We also followed the approach of sectoral modelling supported by the
state-of-the-art models
for each module. We commenced with a sketch of the overall ASTRA model
focussing especially
on the interfaces between the existing models. However, only in the
beginning it is feasible
to speak of interfaces like in the usual software engineering sense. Later
on the modules
become interconnected more and more such that the notion of "one" interface
becomes
unrealistic. Until this stage you might develop separately the individual
modules (or sectors)
and put the updates together from time to time. By the way, we have been
modelling with
4 teams spread over Europe.
However, this does not work the whole way through (see Jim Hines 2nd
comment). Our current
approach is that always one team is the master and might in principle
change all parts of the
model. The other teams can make changes in their module, which have to be
marked and
later on updated in the master model.
I fully agree to the comments that everything becomes much easier, when you
can already
start with an integrated system dynamics model, even though it might be a
very rough first
sketch, and then continue to refine parts of the model step-by-step.

Finally, there is also a German country model called ESCOT, with a similar
approach as ASTRA. It is
also used to test transport policies. It will be presented by Burkhard
Schade at the Palermo
conference.

Regards

Wolfgang

-----------------------------------------------------------------------------------------------
Wolfgang Schade

Universität Karlsruhe
Institut für Wirtschaftspolitik und Wirtschaftsforschung (IWW)
Kollegium am Schloss, Bau IV
76128 Karlsruhe, GERMANY

Email: wolfgang.schade@iww.uni-karlsruhe.de
http://www.iww.uni-karlsruhe.de/
-----------------------------------------------------------------------------------------------

[Niall Palfreyman]

Wolfgang Schade schrieb:
> ... the ASTRA model ... covers whole Western Europe
> divided into 4 regions ... It consists of 4 modules:

I find this really interesting. If Ive understood you correctly, youve
used two orthogonal dimensions in this model: national (regional)
structure, and economic structure (agriculture, chemicals, trade,
transport, environment, etc.). What I find especially interesting is
that youve apparently picked one of these dimensions (economic) as a
split between different sectors, and this is not the first dimension
which comes to my mind as an obvious split for object-oriented
modelling. I would probably have at least started by using the regional
structure as a basis for carving the system into objects, in which case
your sectors would cut straight across the object boundaries.

Now this cutting across object boundaries is a role normally fulfilled
by the use-case model in OO-design, so Im left wondering whether your
sectors reflect a set of use-cases for your model. That is: do the
sectors represent logically coherent questions posed to the model and/or
uses of the model by human users?

Jim Hines wrote:
> Right now Im thinking an object oriented approach to modeling is
> an OK (or maybe even good) choice when creating a model of a type
> with which one is already familiar.

The aim of a use-case model in OO-modelling is to acquaint onesself with
the "type" of the model by investigating the ways in which the model is
to be used in practice. A single use-case traces through the model the
course of fulfilling one user requirement, which then leads to a set of
objects which are required to support that use-case. So given my above
comments, could it be that use-cases (and hence possibly sectors) can be
used to provide the familiarity which you are suggesting is a
prerequisite for applying OO to system dynamics?

Niall Palfreyman.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>