File: c:/ddc/Angel/BestIntentions/World3Again.html
Date: Wed Mar 16 22:14:54 2011
(C) OntoOO/ Dennis de Champeaux

World3 Again


Dennis Meadows and his team developed around 1971 the world simulator World3. The standard run of the World3 model 'predicts' that the world economy gets into trouble in the 21st century. Their first publication in 1972 got a lot of attention, while their subsequent books in 1992 and 2004 did not have a similar impact.

There are multiple reasons why their warnings have been ignored. Making predictions around 1970 about possible global events 60+ years later is a stretch. Science and technology have produced so many breakthroughs in the previous centuries - yielding so much more control by humanity against the challenges of Nature - that the World3 scenarios were simply dismissed and subsequently forgotten.

Ironically, we suspect that the Dennis Meadows's team contributed with their three books to the dismissal of their predictions. They outlined in all their publications sequences of interventions that would avert the troubles of their standard scenario. Since they showed a possible way out, it was realistic to assume that the full power of the world community would be able to keep things under control.

Their 2004 publication announced a last update in 2012. There will be none. Dennis Meadows does not believe anymore that realistic interventions are possible - apparently assuming the veracity of the World3 model.


In the meantime, we had the fortune to receive in 2010 from Dennis Meadows a copy of the World3 model source code. This allowed us to scrutinize deeply the model by checking its assumptions as detailed in [Meadows1974]. We also experimented with many scenarios by adjusting parameters. These parameters correspond with two distinct types of 'adjusting the future'. The first one pertains to interventions that are - in theory - under full control by humanity. The second one makes assumptions about steadily achieving technology advances to counter resource and environmental challenges.

The scenarios described in [Meadows2004] aimed at avoiding a collapse rely on both types of interventions. We doubt whether the technology advances they assume since 2002 are feasible:
- 4% reduction per year in the generation of persistent pollution
- 4% increase per year of agricultural yield
- 4% reduction per year of the use of non-renewable resources

Hence, we ran our own sequence of interventions.

Fast track description of the World3 model

The World3 model tracks key parameters every year from 1900 to 2100. Key primary parameters (simplified) are:
-- population, split up in 4 age brackets
-- capital, split up in industrial and service capital
-- agriculture, split up in arable land and land fertility among others
-- non-renewable resources
-- persistent pollution

There are about 150 secondary parameters that depend on the primary parameters. The secondary parameters determine in turn the changes of the primary parameters in a fine-grained iteration loop. About one-third of the secondary parameters represent relationships that are based on empirical data (obtained by a team of at least 17 members). The model was validated and tuned on the period of 1900-1970. Only minor adjustments were required for their 1992 publication "Beyond the Limits". No adjustments - as far as we know - were necessary for their 2004 publication.

Here two examples of causal connections:
-- Industrial output per capita determines a crowding multiplier. The latter together with the fraction of the population that is urban determines a life expectancy multiplier. This one in turn - together with a few others - sets the life expectancy, which impacts mortality, which determines the population, which ...
-- Industrial output per capita determines a multiplier regarding the usage of non-renewable natural resources. The latter - together with a few others, among which population - determines the persistent pollution generation rate. This one in turn, with a delay, sets the land fertility degradation rate, which impacts land yield, etc.

Dennis Meadows believes that the model is still valid (personal communication). We couldn't find anything, after extensive analysis of the documentation, the code and common sense that counters his belief.

The reference run

We ported the World3 source coded in STELLA to JAVA and obtained the following no-interventions reference run - omitting the 1900-2000 period.

The numbers on the Y-axis are ratios and the different graphs have been normalized to fit in this single figure.

The population peaks in 2030.
Life expectancy peaks in 2010.
Non-renewable natural resource fraction goes down steadily
Persistent pollution peaks in 2060.
Industrial output per capita peaks in 2010.
Food per capita decreases steadily - collective starvation after 2050.
Social services per capita peaks in 2020.
Fertility reaches 2.1 in 2028-9 (!) but too late.

The authors of the World3 model warn repeatedly that the model is not accurate enough to generate actual dates; they only claim that the qualitative behavior is to be considered. In addition, the model has multiple parameters to modify assumptions and to explore the consequences of interventions.

Doubling of available non-renewable natural resources

The model contains a key assumption about the initially available non-renewable natural resources in 1900. It is easy to adjust this assumption to acknowledge current frantic development of energy and fertilizer resources. Curiously enough, doubling the initial resources has some impact but the qualitative behavior beyond doubling does not change. Here the model's behavior with a doubling of initial resources.

Peak population is postponed 10 years.
Persistent pollution goes up dramatically.
Life expectation peaks in 2020 (instead of 2010) but goes down anyway due to decreasing food.

Worldwide max 2-child policy

Demografics is - in theory - under control of humanity. Lets assume the addition of a worldwide max 2-child policy starting in 2011.

The graph does not show changes, but the underlying numbers show differences. Peak population in 2040 goes down from 8B to 7.6B. Famine is postponed for about a decade. Life expectancy decreases slower.

Limiting industrial output

Dennis Meadows et al pioneered the intervention of a worldwide limitation of the industrial production. We selected to maximize the production at the 2011 level.

Industrial output does not increase any longer. Pollution is decreasing substantially. Still, even this very strong limitation on global economic activities is not sufficient to avoid a collapse; life expectancy decreases steadily and a massive famine hits after 2060.

Pollution control

The subsequent additional intervention makes the assumption that persistent pollution can be reduced after 2011 at a modest, hopefully realistic rate.

It is an act of faith that pollution can be controlled as shown. Famine is avoided. Still life expectancy goes down after 2020. The larger population in 2100 has eroded the natural resources further.

Resource use reduction

Assuming in addition that the non-renewable resources consumption rate can be reduced by better recycling, etc. after 2010.

We have at this point:
- Assumed doubling of initial non-renewable natural resources
- Global 2-child policy after 2010
- Restriction of industrial production at the 2011 level
- Assumed pollution control improves after 2010
- Assumed reduction of the non-renewable resources consumption rate after 2010
Even with all these hypothetical technology interventions, we still observe the model's prediction of a painful increasing mortality, caused by decreasing food/per capita.

Worldwide max 1-child policy

Here the impact of a (social-fiction) worldwide 1-child policy starting in 2011.

Only with this final exceptional intervention - in addition to the earlier questionable interventions - does the model predict maintaining life expectancy, a main achievement of the 20th century by keeping food/per capita stable. The price is a decreasing world population; not by famines but due to 'voluntary' fertility control.

The model is too optimistic in that it ignores local differences, wars, local/global conflicts, climate change, structural unemployment and - perhaps most importantly - it ignores how humanity in the 22nd century and beyond is supposed to survive on a ravaged, plundered planet.

The World3 model has been attacked - not based on its architecture and not based on its causal mechanisms - but regarding its findings. Extensive analysis of the World3 model has not revealed a flaw. The attacks appear to be irrational denials fueled by faulty analogy reasoning and by excessive optimism: the collapse scenarios should not happen therefore they cannot happen.

Humanity has been described as climbing a resource consumption ladder and breaking off lower rungs. It is just faith that higher rungs exist that can carry an imprudent, humanity.

Sure, we have shown also a sequence of interventions that avoid a catastrophic collapse. However, we want to be the first to admit that these interventions are quite implausible to pull off in our geo-political constellation - while options are shrinking with delays.

The chapter What If addresses briefly how things may actually get out of hand.


[Meadows1974] Meadows, D. et al, "Dynamics of Growth in a Finite World", Wright-Allen Press, ISBN 0-9600294-4-3, 1974.

[Meadows2004] Meadows, D., J. Randers, & D. Meadows, "Limits to Growth, The 30-Year Update", Chelsea Green Publishing Co., ISBN 1-931498-58-X, 2004.

Back to index