Forecasting accuracy + Transportation

[Wolfgang Schade]

Dear all,

Our institute (IWW) together with three partners is currently
developing a Strategic Assessment Tool based on System Dynamics
that deals with transportation issues. Also included are submodules
for macroeconomic, regional economic and environment. The aim of
the project (called ASTRA) is to analyse the outcome of different
transport policies in terms of e.g. macroeconomic indicators (e.g. GDP)
or environmental indicators (e.g. NOx emissions). Surely we are
not aiming at a point forecast but at an analysis of trends and
of maybe unexpected behaviour of these variables.

You can find more information on the website of the project:

http://www.iww.uni-karlsruhe.de/ASTRA/

On this site it will soon be possible to download already
finalised deliverables of the project.


Also I would recommend to look into the proceedings of the
System Dynamics Conference in Quebec. There you should find a
paper from Karsten Kuchenbecker from IWW named "Transport
=46orecast and System Dynamics". For more information you
may contact him by email:

karsten.kuchenbecker@iww.uni-karlsruhe.de


Last I would like to make some comments to the discussion on
forecasting accuracy. First you should define clearly the
intention what to forecast.
If one aspires forecasts for a certain value of a variable
this is possible with System Dynamics Models (SDM). BUT this is
only the case when you work with clearly definable relationships
like the laws of nature. E.g. Hartmut Bossel gives an example
to model the movement of a pendulum with system dynamics.
But if there are soft variables integrated into the model one
should not expect to get accurate point forecasts from the SDM.
Usually this is the case for socio-economic models. They include
a lot of soft variables (e.g. quality of life) and soft
relationships (e.g. the relation between regional environmental
quality and settlement behaviour). In this case we still
can expect useful output of a SDM in the sense of getting an
idea of the development of the model variables over time. Combined
with sensitivity analysis e.g. for different policies
SDM are a quite useful tool to identify policies which
in the long-term should fulfil the social, economic and/or
environmental targets of society.
The next step is to avoid that people do use the mere numbers
at a certain point of time as the result of the model but instead
they should focus on the development of the variables over time.
E.g. this happened in some way to "The Limits to Growth" where some
people are arguing that we reached e.g. 1998 now and the
values of the variables in the model are different from
reality. But in my opinion we are still moving on the line
of the forecasted development of the most variables.

Best regards

Wolfgang Schade

### ---------------------------------------------------------- ###
Wolfgang Schade


Universitaet Karlsruhe
Institut fuer Wirtschaftspolitik und Wirtschaftsforschung (IWW)
Kollegium am Schlo=DF, Bau IV
76128 Karlsruhe, Germany

eMail: wolfgang.schade@iww.uni-karlsruhe.de
fon: +49 (0) 721 608-4371 fax: +49 (0) 721 34613
### ---------------------------------------------------------- ###=20