Dynamic Modelling

Dr. Gayl D. Ness
Member of AUICK International Advisory Committee / Professor Emeritus, University of Michigan, USA

Note: This discussion draws on a 2000 AUICK book, Five Cities: Modelling Asian Urban Population Environment Dynamics, Singapore: Oxford University Press, pp 54ff.

Introduction

The nine AUICK Associate Cities (AACs) addressing their position on the Millennium Development Goals (MDGs) are using some form of dynamic modelling to assess their current conditions and help plan for the future. This makes it important to understand what dynamic modelling is and what are its advantages and limitations.

What are dynamic models and how do they work?

Models are simplified versions of reality. They are useful tools for understanding, because they allow us to cut through the great complexity of any condition or process to focus on a manageable number of variables. They also allow us to examine possible futures, called scenarios in the language of modelling. Models describe relationships between different conditions. Knowing these relationships allows us to ask what will happen if these relationships persist into the future. If we do not actually know what the relationships are we can make assumptions (or guesses) about them and use those assumptions to examine future scenarios.

Models are often called dynamic, meaning that what happens in one part of a model, with one or two conditions, will affect other conditions in the model. They often involve what are called feedback loops. For example, in population modelling, births constitute a positive feedback loop. The more births we have the more people will be available to give birth and thus the more births we will have in the future. Death, on the other had, is a negative feedback loop. The more deaths we have the fewer people will be available to contribute to the population by giving birth.

Illustration of a dynamic model

To illustrate, we can model population growth by looking at four rates: birth and death, and in- and out-migration. If we know these rates from the past, we can project them into the future to ask what the size of the population will be under those rates. If we do not know past rates, we can make assumptions about them and still project them into the future. In either case, we cannot know for certain what the future rates will be, but we can make assumptions (or guesses) about them and calculate the outcome of those assumptions. The United Nations has been doing this with world and regional population estimates for near half a century. The different scenarios the UN has run for future world population levels have been fairly accurate on the global level. To be sure, more recent projections are being revised downward, but on the whole the projections have been remarkably close to what actually came to pass.

This illustrates some important aspects of dynamic modelling. The model is very simple; it uses only four rates and four variables: births, deaths, in- and out-migration. Many things affect births, including things like the age of marriage and the proportion of the population that does not marry, how many children are wanted, and the living standards of a population. The same can be said for deaths and migration. Many conditions affect these movements. Trying to encompass all of these complex conditions would make it almost impossible to think about future population sizes. The model allows us to cut through this complexity and to deal with a manageable number of conditions or variables.

It also illustrates the importance of assumptions. Modelling requires making assumptions about the future. If we assume that the death rate will decline, we will find increasing population numbers in the future scenarios we run. If we assume birth rates will decline, future scenarios will show a decline in the rate of population growth. If we assume the birth rate will be very low (below replacement level, as it is now in Japan and most of Europe), future scenarios will show an actual decline in the population. The most important aspect of these assumptions is that they are explicit and thus highly visible. People, whether expert or not, may disagree with any assumptions. Thus assumptions can be changed to examine the implications of the change. Most people carry models in their heads. Usually the assumptions behind those models are not explicit; they are not stated; they are not visible; and they cannot be challenged. Dynamic modelling makes the assumptions explicit and visible.

Benefits of using dynamic models

It is the visibility and the challengeability of assumptions that makes dynamic modelling both powerful and useful.

Let me give another example from the Five Cities study, using the case of Kobe. We modelled the numbers of vehicles in the past with projections of their increase to the year 2020. Knowing the ratio of automobiles to specific air pollutants ? suspended particulate matter (SPM) in particular ? we could calculate the increase in SPM, assuming the growth of vehicles and the emissions technology that existed in 1995. This showed rising levels of SPM, from which we could calculate that this would imply an additional 10,000 deaths to the year 2020 that would result from the increased SPM. Kobe?fs engineers are quite certain that the emissions technology will improve, and the increase of vehicles will not actually result in our scenario?fs outcome of pollution and mortality, but it was very useful to run the scenario with those assumptions to point out to Kobe?fs leaders one possible implications of an increase in vehicles.

The ability to run different future scenarios with different assumptions gives dynamic modelling much of its power and utility. Moreover, today there are a number of computer programs for dynamic modelling. The AACs are using a program called STELLA. This implies great speed in calculations. It also implies that the running different future scenarios with different assumptions is a very simple task, which is especially useful for urban administrators. The assumptions are all clearly stated in the computer model. If a particular future outcome seems highly improbable, its underling assumptions can be examined and changed to produce an outcome that seems more probable. On the other hand, however, if the assumptions appear reasonable, administrators can see a probable outcome of today?fs conditions. They can then create policies or programs to correct current conditions so that the model?fs projected outcome will not come to pass.

How can dynamic modelling assist urban administrators?

In effect dynamic modelling gives urban administrators the opportunity to look into the future and to make educated or informed guesses about what the future might hold. More importantly, it allows groups of administrators and scientists to work together to look into the future and to design policies and programs to make the future a better one. This is what is happening among the AACs. They are building teams of urban administrators and university scientists whose collaboration will be facilitated by dynamic modelling. This hope is that this will help to improve the quality of life of the people in those cities. It is also the hope that this will provide a useful model for other cities and national governments to follow.

Editor's Note: This article was contributed by Dr. Gayl D. Ness specifically to this newsletter to provide more concrete ideas on the dynamic modelling theory after the workshop.