The story behind "Last Call"

I recently had the chance to see a documentary called Last Call, which recounts the 1970's era effort to use the new field of systems dynamics to create a computational model capable of forecasting the future of human society on planet Earth. This is the project that led to the prescient and controversial book Limits to Growth. The documentary describes some of the virulent attacks on the team and their work, and then goes on to say that the World3 standard model has, in fact, held up well when compared with the actual developments from 1970 through 2014.

Donella Meadows is one of my heroines, and I have long thought that The Limits to Growth book was brilliant. So I enjoyed the movie very much. The film-makers got interviews with systems dynamics pioneer John Forester, authors Dennis Meadows and Jorgen Randers, plus archival footage of lead author Donella Meadows and the instigator of the project, Aurelio Peccei of the Club of Rome.

However, I have to say that I'm not sure the movie would make much sense or have much impact for a person who wasn't already somewhat familiar with The Limits to Growth (LtG) work. And being a "show me the data" sort of a person, I can't for the life of me understand why the film-makers chose not to show any of the data that compare the LtG scenarios with empirical observations from the 43 years since the book was published.

Some friends and I are gearing up for a community showing of the film, and so I offer here a synopsis of the story behind the story, as I see it. My commentary is in three acts: Act I: Meadows et al create their model and write their book. Act II: Economists and organizations dependent on growth of the economy attack viscously, and the public is left with the impression that the work has been completely discredited. Act III: other researchers compare the model output with data accumulated in the 30 or 40 years since the book was published.

Act I: The model

The film tells the viewer that the model was important, and difficult, and that it was intended to forecast the future of the Earth. But it doesn't tell or show the viewer anything about how the model worked or why it was so challenging to create.

The first challenge in creating the World3 model was to figure out what aspects of the Earth to include in the model. Of the blooming, buzzing confusion that comprises the natural and built environment, what to include, what to leave out? The Limits to Growth team chose five variables: world population (along with global birth rate and death rate), industrial output per capita, agricultural output per capita, amount of non-renewable resources remaining, and persistent pollution.

Next, they had to figure out what the relations were among those various variables. They thought in terms of systems, in which a system can be defined as "a set of elements whose interconnections determine their behavior." Central to their thinking was the idea of feedback loops. As I reviewed in an earlier post, a feedback loop is a circumstance under which A causes or influences B, and B causes or influences C, and C, in turn circles around (in either a straightforward or a complex way) and influences A.

Feedback loops come in two types: reinforcing and balancing. In a reinforcing (or "positive") feedback loop, the interconnections are such that changing one element in the loop will start a sequence of changes that will result in the originally changed element being changed even more in the same direction as the initial change. Over time the system as a whole moves away from equilibrium, in one direction or another, towards a more extreme condition. Systems dominated by a reinforcing feedback loop typically exhibit exponential growth, such as money in an interest-bearing checking account, bacteria in a petri dish, or the doubling grains of rice on the squares of a chess board in the movie. " ... a positive feedback loop is self-reinforcing. The more it works, the more it gains power to work some more...Positive feedback loops are sources of growth, explosion, erosion and collapse in systems" (Meadows, 1999, p. 11).

In a balancing (or "negative") feedback loop, a change in one element is propagated around the circle until it comes back to change that element in a direction opposite to the initial change. The end result of the chain of influences is that the system as a whole tends to stabilize around an equilibrium position. If the system starts to drift or get pushed one way or the other, the feedback loop tends to pull it back from extreme states or behaviors. "Negative feedback loops are ubiquitous in systems. Nature evolves them and humans invent them as controls to keep important systems states within safe bounds" (Meadows, 1999, p.9).

The World 3 model has feedbacks leading back and forth among all of their five variables: world population, industrial output per capita, agricultural output per capita, amount of non-renewable resources remaining, and persistent pollution. The diagram at the right shows a snippet that pertains to birth and death rates. Populations of living organisms, including humans and the bacteria in the petri dish mentioned above, are controlled by a combination of a reinforcing feedback loop and balancing feedback loop. As the population increases, there are more parents who give birth to more offspring, and the population increases yet more, cycling around a reinforcing feedback loop. Population skyrockets. Eventually, though, balancing feedback loops begin to take effect: for example, more people leads to less food per capita (all else being equal) which in turn leads to higher mortality, which tends to decrease the population. The population portion of the model was connected to the rest of the model via a number of feedbacks that influenced average fertility or average mortality.

A systems diagram, such as the one sketched at the left, can be used to map out the flows of influence as well as materials and energy, through the system. A systems map or diagram looks at first glance like the concept maps that are now used in education at all levels from middle school on up. However, the genius of the newly emerging field of systems dynamics was that it used data-informed equations to represent the arrows and boxes on the diagram, such that, for example, a reinforcing feedback loop was underlain by an equation for exponential growth over time. The equations were assembled into a computational model, which advanced through time, allowing the various feedbacks, flows, and influences to act on each other, timeslice by timeslice, across decades.

Importantly, the model included time lags or delays. For example, after a reduction in fertility, there is a time lag before the change shows up in the population data because large numbers of girls born before the fertility change are still reaching child-bearing age each year. Similarly, there is a lag between the first release of a pollutant and observable harm, and then another lag between a decision to reduce use of the pollutant and observable decrease in harmful effects. Systems with lags can behave in non-intuitive ways, and be hard to predict or control.

To connect the equations to the Earth, the LtoG group used historical data to calibrate the model. This aspect of "Last Call" was eye-opening to me. I had not fully appreciated how challenging it was to find long time series of appropriate data for the LtoG work. Very little of the data was available digitally and none was available remotely, so team members trekked to Washington D.C. to make deep dives into paper archives of long-ignored data in dusty basements of government agencies. With this data in hand, they refined and iterated the model, until the behavior of the model matched the behavior of the Earth-human system as recorded in data for the interval from 1900 through 1970.

Then they set the model loose to run into the future. The LtG team ran their World 3 model with many different combinations of input parameters. For example they held everything else constant and doubled the assumed amount of recoverable natural resources, or the productivity of agricultural land. In the book, they built their concluding narrative around three scenarios:

• The Standard Run represents a business-as-usual scenario, where the physical, economic and social relationships that pertained during the model's calibration period (1900-1970) are carried out into the future. The other scenarios are compared and contrasted to this standard. In the Standard Run, economic growth continues through the 20th century and into the 21st. But then sometime in the 21st century, balancing feedback loops kick in strongly, and the system goes into overshoot and collapse, due to a combination of diminishing resources and ecological damage from pollution.
• For the Comprehensive Technology scenario, the LtG team envisioned technological fixes to some of the early-developing stresses: pollution is reduced to 25% of its 1970's level, birth control is universally available and the productivity of farm land is doubled. Even so, the system eventually goes into overshoot and collapse, when the growth in economic activity catches up with the gains in efficiency from better pollution control and agricultural practices. Collapse is delayed until the late 21st century.
• 

  ×

  Output from Stabilized World Scenario

  Provenance: figure 46 from Meadows, D., Randers, J., & Meadows, D. (2004). Limits to Growth: The 30-year Update. White River Junction, VT: Chelsea Green Publishing.
  Reuse: Fair use claimed for purpose of commentary.
  In the Stabilized World scenario, both technological solutions and deliberate social policies are implemented, with the goal of reaching a stable equilibrium of population and economy. Birth control is widely available, and in addition a widespread preference for two children as the ideal family size has been established. Agricultural land is preserved with capital diverted from industrial use. Less spending is directed towards material goods, and more towards health.

The LtG team were able to find a set of Stabilized World assumptions that resulted in a gradual bending of all of the curves such that services per capita, food per capita, industrial output per capita, and global pollution all stabilized at livable levels by the mid-21st century. They ended their book on an optimistic note, saying that an environmentally sustainable world is possible--but only if we act soon.

Act II: The attacks

Limits to Growth was published in 30 languages and sold 30 million copies. It generally well received by natural scientists. But soon after it was published, vicious attacks on the work and the team began to appear. Most originated from economists and business people. The movie downplays Act II: The attacks, showing an amusing scene where one of the team is sorting through old newspaper articles and chuckling at how outrageous they were.

Higgs (2014) presents a detailed analysis of who the attackers were and what their motivations might have been. A few quotes from some of the more prominent attackers will give a feel for the nastiness that came to swirl around the work:

In the New York Times book review section, three economists from Harvard and Columbia wrote:

Higgs (2014, p. 53) assembles a horror show of other criticisms: "brazen impudent nonsense," "spurious scholarship," "computerized mumbo-jumbo," "fraud...fantasy... fudge..."

By the end of the century, the Limits to Growth work had been largely forgotten by the public. Those who might remember the name "Limits to Growth" or "Club of Rome" were left with the notion that the work had been long since discredited (Simmons, 2000).

Act III: The vindication

Members of the original Limits to Growth team published an update to their work (Meadows, et al, 2004) 32 years after the original book. They addressed some criticisms, and showed that population and climate change were increasing exponentially, in a way that resembled closely the scenarios of their 1972 book. However, it was easy for skeptics to dismiss these findings as self-serving, as the forecast and the validation both came from the same group.

An independent verification of the Limits to Growth scenarios came from Australia, in work by Graham M. Turner, of CSIRO Sustainable Ecosystems (Turner, 2008; Turner, 2014). Turner (2008) set out to compare historical data for 1970-2000 with the model outputs from the Limits to Growth work. He had to make some leaps to find appropriate data series for some of LtG's parameters. Global birth rate and death rate were the most straightforward. For the LtG model output parameter of "food per capita," Graham used kilocalories per capita per day. For LtG "industrial output per capita," he was able to find a suitable United Nations index of industrial production in a global aggregate. For the LtG parameter of "services per capita," Graham looked at electricity use per capita globally and at literacy rate, which strike me as relevant, but not the whole story. For the amount of non-renewable resources remaining, Graham focused on fossil fuels and presented upper and lower limits to account for the uncertainty in estimates of ultimately recoverable fossil fuel. For persistent pollution, he used atmospheric carbon dioxide level, the pollutant for which global data were most available. For each parameter, Turner normalized the observed data to the LtG model value at 1970, and then compared how the behavior of the data compared to the behavior of the model from 1970 onward.

Turner (2008) provided time series graphs comparing the observed data with three different LtG scenarios: Standard Run, Stabilized World, and Comprehensive Technology. He found that historical data for 1970-2000 compared favorably with the business-as-usual scenario that LtG called the "standard run scenario," and was a poor match for the Stabilized World and Comprehensive Technology scenarios. The graph at the right shows the food per capita comparison.

Turner (2014) updated the empirical time series up through 2010, and dug deeper into the working of the model and its assumptions. The updated data-model comparisons are shown below. Thin solid lines, extending from 1900 through 1970, indicate the time interval over which the World 3 model was calibrated. The dashed lines show the LtG Standard Run scenario from 1970 through 2100. The thick solid lines show the observed data from 1970 through 2010.

This comparison suggests that we are at a very interesting moment in history. The modeled scenario and the observed data both show rising population, industrial output per capita, food per capita, services per capita, and pollution, along with decreasing non-renewable resources. But in the model, all of these parameters (except for non-renewable resources) reach a turning point in the next few years. First industrial output per capita, then food per capita, then services per capita, and finally population, all reach a peak and then begin to decline.

Will reality continue to track the modeled scenario over these peaks and down the other side? Time will tell.

---

References & Sources:

The video Last Call:

• website for the film
• to order a copy on DVD for showing for a community group or school.

Meadows, D. H., Meadows, D., Randers, J., & Behrens, W. W., III. (1972). The Limits to Growth: Universe Books. A pdf version of the book is available here

Meadows, D. (1999). Leverage Points: Places to Intervene in a Systems, The Sustainability Institute.

Meadows, D., Randers, J., & Meadows, D. (2004). Limits to Growth: The 30-year Update. White River Junction, VT: Chelsea Green Publishing.

The definition of "system" used above ("a set of elements whose interconnections determine their behavior") comes from the tutorials on systems thinking provided by The Climate Leader.

Passell, P., Robert, M., & Ross, L. (1972). Review of Limits to Growth.New York Times Book Review . April 2 issue, pages 1, 10, 12-13.

Higgs, K. (2014). Collision Course: Endless Growth on a Finite Planet. Cambridge, MA: MIT Press. Higg's book was originally a PhD thesis, and is a pretty heavy lift. A summary of her argument can be found in a review by Herman Daly.

Simmons, M. R. (2000). Revisiting the Limits to Growth: Could the Club of Rome have been correct, after all? GreatChange.org. Investment banker and peak oil commentator Matt Simons recounts how he had picked up the notion that Club of Rome and Limits to Growth were long ago discredited."

Turner, G. M. (2008). A comparison of "The Limits to Growth" with 30 years of reality. Global Environmental Change, 18(3), 397-411. This paper does a really good job of summarizing the LtG work, as well as comparing LtG model outputs with data.

Turner, G. M. (2014). Is global collapse imminent? An updated comparison of The Limits to Growth with historical data, MSSI Research Paper no. 4. Melbourne Sustainable Society Institute, University of Melbourne.

A video interview with Graham Turner, discussing his work on comparing the Limits to Growth model output with new data, is available here

Another comparison of LtG with empirical data was done by Hall, C. A. S., & Day, J. W., Jr. (2009). Revisiting The Limits to Growth after peak oil. American Scientist, 97(May-June), 230-237. They also found good agreement, but they did not present their data in a very compelling way, looking only at a single time slice (2008), rather than at behavior over time. However, there is a priceless quote in their paper (p. 235): "... its predictions have not been invalidated, and in fact seem quite on target. We are not aware of any model made by economists that is as accurate over such a long time span."