Do All Roads Lead to Collapse?

A Review of the book Limits to Growth - The 30-Year Update

“The metaphor is so obvious. Easter Island isolated in the Pacific Ocean — once the island got into trouble, there was no way they could get free. There was no other people from whom they could get help. In the same way that we on Planet Earth, if we ruin our own [world], we won't be able to get help.”

- Jared Diamond, Collapse: How Societies Choose to Fail or Succeed (2005)

“Our global civilization today is on an economic path is environmentally unsustainable, a path that is leading us toward economic decline and eventual collapse.”

- Lester Brown in Plan B 4.0 (2009)

It has now been 48 years since the original The Limits to Growth (LTG) was first published (1972), and the insight acquired in the interim allows us not just to review the forecasts made by the authors, but also to track how the human civilization has handled its existence on this planet, faced with the inevitable prospect of encountering limits to its expansive and extirpative biological nature (Rees, 2020).

Under these circumstances, the authors Donella Meadows, Jorgen Randers, and Dennis Meadows devised a series of scenarios in their global model (World3), arriving at the stark conclusion that “overshoot and collapse” could only be averted if the human project were to engage in deeply transformational social changes and technological progress. When such modifications were not implemented in the simulated scenarios, an abrupt collapse of both the economy and the population took place during the twenty-first century (Turner, 2014). Meadows and colleagues asserted:

Our most important statements about the likelihood of collapse do not come from blind faith in the curves generated by World3. They result from understanding the dynamic patterns of behavior that are produced by three obvious, persistent and common features of the global system: erodible limits, incessant pursuit of growth, and delays in society’s responses to approaching limits. Any system dominated by these features is prone to overshoot and collapse.” (p. xviii)

The features of the global system described as well as any substantial reframing practices were meant to come into effect already when the original LTG was published, with the inaction leading the authors to state in the 30-year Update that (2004 p. xvi):

"Consequently, we are much more pessimistic about the global future than we were in 1972. It is a sad fact that humanity has largely squandered the past 30 years in futile debates and well-intentioned, but halfhearted, responses to the global ecological challenge. We do not have another 30 years to dither. Much will have to change if the ongoing overshoot is not to be followed by collapse during the twenty-first century.”

The pessimism cited by the authors is an excellent transition into what they note in the preface as being different hopes, expectations, and worldviews by each one of the authors, and comparatively, to my own, as I dwell on the revision of this timeless book. By all means, everyone’s prospects for the future and perspectives on risk and resilience are shaped by different personality traits as well as the social context and upbringing (Sharpe, Martin & Roth, 2011; Smith et al. 2013; Wenglert & Rosén, 1999; O’Connor & Cassidy, 2006), so it should be no surprise that among the authors of such an influential book as the LTG, there are also relevant distinctions in the way they perceive the future. Indeed, Meadows et al. have made a great service to the readers by specifying in their preface the positions and where on the optimism-pessimism-realism gradient each one of them was located. In their 30-year Update version these were their positions (p. xvi):

  • Donella Meadows was the unceasing optimist. She was a caring, compassionate believer in humanity. She predicated her entire life’s work on the assumption that if she put enough of the right information in people’s hands, they would ultimately go for the wise, farsighted, the humane solution, by adopting global policies that would avert overshoot.

  • Jorgen Randers is the cynic. He believes humanity will pursue short-term goals of increased consumption, employment, and financial security to the bitter end, ignoring the increasingly clear and strong signals until it is too late. He is sad to think that society will voluntarily forsake the wonderful world that could have been.

  • Dennis Meadows sits in between. He believes actions will ultimately be taken to avoid the worst possibilities for global collapse. He expects that the world will eventually choose a relatively sustainable future, but only after severe global crises force belated action.”

With this personal portrait of the authors, the reader and I will experience an unflinching reaction to identify with one of these characterizations. To demonstrate, for my counterpart, I would consider very close similarities with the more pessimistic [also termed realist (Rees, 2019)] stance of Jorgen Randers.

The reason I’m going down this road is that I’m convinced that if a 50-Year LTG edition (or more) sees the light of day, it will be increasingly dominated by a pessimistic/realist sentiment. This, of course, will not be solely due to the absence of the optimism of Donella Meadows in any future edition, but by the sheer inescapable fact that we are moving farther away from sustainability every day, and our “answer” to this is still the:

“Support of growth-oriented policies, because [individuals] believe growth will give them an ever-increasing welfare. Governments seek growth as a remedy for just about every problem. In the rich world, growth is believed to be necessary for employment, upward mobility, and technical advance. In the poor world, growth seems to be the only way out of poverty. Many believe that growth is required to provide the resources necessary for protecting and improving the environment. Government and corporate leaders do all they can to produce more and more growth. For these reasons growth has come to be viewed as a cause for celebration. Just consider some synonyms for that word: development, progress, advance, gain, improvement, prosperity, success (p.6)”

Regardless of the conundrums we face, the purpose of this examination is not to provide solutions to our existential trepidations, but an analysis of the thought process behind the LTG, and from it attempt to answer the question posed in the title, if all roads lead to collapse. To do that, I’ll start by reviewing what can quite possibly the most prescient model ever conceived by humanity, the World3.


The capacity to predict the future might be the most coveted skill in the history of our species, yet, no human has ever been endowed with such foreknowledge except in mythology (Fry, 2017). However, that isn’t to say that our shared talent in achieving a sort of mental time travel by resorting to projections and simulations of the mind hasn’t given our species an evolutionary advantage (Suddendorf & Corballis, 2007). With that in mind, the task that Meadows and colleagues set themselves into, back in the 70s was of prodigious relevance, and even more so, when we consider that the trends have mostly been proven accurate (Turner, 2014; 2008).

To explain, Meadows and colleagues resorted to their computer model called World3, which simulated how key variables such as population, industrial output per capita, pollution, agricultural systems, food per capita, services per capita, consumption goods per person, non-renewable resources, life expectancy, human ecological footprint and the human welfare index would interact in a system dynamics approach.

Marking the 40th anniversary of the LTG, co-author Jorgen Randers wrote in 2052: A Global Forecast for the next 40 years (2012) that “when people think about the future, world population usually comes first.” I will be honoring this premise and setting population as the starting point of my analysis, not just do its utmost relevance and often superficial and untended inspection and consideration (Kopnina & Washington, 2016; Campbell, 2012; Coole, 2012), but also due to my academic involvement with the subject (Abegão, 2018; 2019; 2020), which grants me the critical capacity to scrutinize past, present and future knowledge and perceptions regarding the growth of the human population, while supplementing (what I hope to be) valuable inputs.

For starters, one has to concede the refreshing and honest tone of the LTG regarding the role of population [by today’s standards (O’Sullivan, 2018)] on achieving any meaningful sustainability. A straightforward examination of the Index results in more than 100 pages committed to the topic of population. To demonstrate, the authors write in their preface (xi) that:

“In our scenarios, the expansion of population and physical capital gradually forces humanity to divert more and more capital to cope with the problems arising from a combination of constraints.”

They supplement this statement with the following a couple of pages forward (p.2):

“Growth in the globe’s population and material economy confronts humanity with this possibility [catastrophic overshoot and possible collapse].”

However, I would like to take this opportunity to review the material and forecasts dedicated to population in LTG, to showcase how reality has racked up a significant advantage compared with the model. With that in mind, the first item on the list would be to ponder on how the projections have changed since the original LTG (1972) and 30 years later on the updated version (2004).

When reviewing the original LTG one is hard-pressed to find a single specific mention of a future population estimate. It is beyond me if this was because the first official United Nations population projection was published almost a decade after the book had been published (Pison, 2019), or if the authors cautiously avoided that so that the book would age better and avoid ex post facto acrimonious inspections. Notwithstanding, World3 has still managed to escort reality for almost 50 years, even though I slightly diverge from their current forecasts.

For starters, when we examine figure 1 we see that the forecast was for the number of people added to the planet each year [yearly change (births – deaths)] to continue its trend of decline after the year 2000. I’m not criticizing the decision to trust this trend at the time, it was completely justified. In detail, when one examines figure 2, in 1990 the yearly change reached its peak, and for the following ten years, this trend was maintained, until the year 2000 when the yearly change was roughly 12 million fewer people compared with 1990 values.

Figure 1: World Annual Population Increase. Retrieved from Limits to Growth, 2004 page. 30
Figure 2: Population Data from 1960 to 2020. Retrieved from Worldometers, 2020

The trend proved to be short-lived. After the year 2000, the yearly change never went down again, and it even rose above 80 million afterward. This has of course cast a shadow on the expectations of many that the human population would stabilize and decline as soon as 2050 (Bricker & Ibbitson, 2019), something that has been rightfully contested (McKeown, 2019a; 2019b). If nothing else, the malleability inherent to official population projections, as well as the overconfidence in fertility declines has meant that population projections end up being revised upwards almost every year (O’Sullivan, 2016). Indeed this still appears to be the situation we find ourselves in, as the official forecasts already put the population by the end of the century, 400 million above initially predicted in 1981, and most likely to rise (Pison, 2019). However, what truly concerns me was not that the 30-Year Update forecast on population didn’t stand the test of time, but that almost a decade after its release, Randers wrote (2012) that according to his forecast the population would peak in 2040 at 8.1 billion people and decline afterward for the rest of the century. His reasoning, he explains, is because “I think fertility trends will continue downwards at the stupendous rate that has occurred over the past 40 years.”

Please keep in mind that Randers was the most pessimistic among all of the authors of the LTG. However, in regards to population, he appears to be even more optimistic (or borderline unrealistic) than the most quixotic of ‘experts’ on the field of demography (Bricker & Ibbitson, 2019; Rosling et al. 2019). We already stand close to 7.8 billion (Population Reference Bureau, 2020), and if the yearly change trend has given us any indication is that we will most likely continue to grow by over 80 million people per year [COVID-19 will likely increase this value even more, due to shortages of contraceptives affecting reproductive health (Godin, 2020)]. It remains to be seen, but Randers’ ‘non-peak’ of 8.1 billion might be more likely to occur somewhere around the middle of this decade than in 2040 and continue to climb to 10.8 billion by 2100 [assuming of course that fertility declines are maintained, which is itself a considerable gamble to make (Bish, 2020)], until of course our current overshoot turns into collapse.


One of the most remarkable aspects of the World3 model is the fact that it reveals how the variables included (one of which is population) are likely to be expressed when combined all together. In other words, World3 incorporates the momentum of population growth, the build-up in pollution, the escalation of industrial output, the variations in investment among different sectors, the surge of the ecological footprint or any inconstancies in food production, among others, and sketches them all with many active feedback loops which end up creating different scenarios. As an illustration, changes in population may cause alterations in the economy, with less industrial output and services in turn affecting food per capita and health services, eventually leading to an increase in death rates. As the authors state (xi):

“When industry declines, society can no longer sustain greater output in the other economic sectors: food, services, and other consumption. When those sectors quit growing, population growth also ceases (it will also revert).”

It is important to realize that not just a contraction of industrial output can end up creating a scenario of collapse. A lack of non-renewable resources, increased pollution, decreasing investments, rising depreciation, population growth, or consumption of the natural capital of the Earth can all lead us down that road. To be sure, figure 3 reveals all of the scenarios of World3 relative to population, and how other variables end up shaping the size of our human aggregate during this century.

As one can discern, in only two out of the nine scenarios on display does the population curve not suffer a sudden crash and decline. Under these circumstances, and not to mince any words here, billions of people are foreseen to face a premature death.

Regrettably, as the authors affirm:

“It should come as no surprise that the most likely mode of behavior of the model world is overshoot and collapse.”
Figure 3: Alternative scenarios for population in World3 model. Retrieved from Limits to Growth, 2004. Page 14.

To better illustrate what happens in their recurrent behavior mode of overshoot and collapse, the graph provided by the late Bradford Hatcher (2019) perfectly encapsulates what happens to the human population as it breaches its carrying capacity. To point out, carrying capacity is the limit that allows a population to reach a maximum size and be sustained in an environment with food, habitat, water, and any other necessary resources (Hui, 2006). Together with this, it is critical to point out that the maximum population does not necessarily mean optimum population size, a relevant distinction that has been raised by population ecologists regarding the sustainable size of the human population (Daily & Ehrlich, 1994).

Figure 4: Graph portraying the relation between population growth, the overshoot of carrying capacity and the degradation of carrying capacity during overshoot. Retrieved from Hatcher, 2019

According to the graph, as population grows past its carrying capacity it enters overshoot (Catton, 1982), leaving behind any meaningful state of sustainability. While in overshoot two modes of behavior can take place, and both of them are modeled by World3. The first state is overshoot followed by oscillation and the second is overshoot and collapse. As the authors describe, ceteris paribus, collapse becomes the norm. Come what may, there is still one aspect of this graph which I deem to be of critical importance.

In the original Limits to Growth version, Meadows and colleagues (1972, p. 92) explained that as the population goes beyond its carrying capacity it may face a sudden collapse, due to reasons such as insufficient resources or increasing pollution, among others. Another issue that is confronted throughout LTG is what I perceived to be addressed as the erosion of limits by the authors. To clarify, Meadows and colleagues consider that when a system surpasses a given threshold, feedback loops will take over and create a cycle of erosion, which ultimately leads to further environmental damage and collapse (Butzer, 2012). Furthermore, if the feedback loops maintain their activity, loss of planetary bio habitability will ensue, as in the case of an eventual ‘Hothouse Earth,’ rendering the planet uninhabitable for humanity and many other non-human species (Steffen et al. 2018).

In any event, as a system enters into overshoot, the carrying capacity also starts to diminish, increasing the deficit between a growing overshoot and a declining carrying capacity, as it can be observed happening in figure 4. For this reason, the authors write:

“Any population that grows past its carrying capacity, overshooting the limit, will not long sustain itself. And while any population is above the carrying capacity, it will deteriorate the support capacity of the system it depends upon. If regeneration of the environment is possible, the deterioration will be temporary. If regeneration is not possible, or if it takes place only over centuries, the deterioration will be effectively permanent.”

If we were to use the metric of Ecological Footprint (EF) (Wackernagel & Rees, 1988) as a proxy for our collective impact, humanity would already be in a state of overshoot, using as many ecological resources as if we inhabited 1.75 Earths. According to the ecological footprint metric (figure 5), this would mean a present overshoot of 75% (Global Footprint Network, 2020; Tamburino, 2020). The EF is compared with another concept, called biocapacity, which can be described as the readiness to produce biological materials used by people and to absorb waste material generated by humans (Tamburino, 2020). When EF is above biocapacity overshoot occurs. Both are on display in the graph below.

Figure 5: Comparison of Ecological Footprint, Biocapacity and current Ecological Deficit. Retrieved from Global Footprint Network, 2018.

Indeed as one can observe, the EF rose above biocapacity for the first time around the time the original Limits to Growth was being published, and we have remained in overshoot ever since (Galli et al. 2014). Although this may be true, one might be inclined to argue that humanity has managed to keep its ecological footprint somewhat constant for the last few decades, seeing that only oscillations remain and EF appears to have reached its peak (Global Footprint Network, 2018).

I remain skeptical of this alleged plateau, partly because on a finite planet that continues to pursue economic and population growth as its dominant mantra, a stable EF (even though we remain in overshoot) seems nonsensical. If one considers the fact that ever since the early 70s until today, roughly 4 billion humans were added to this planet, all the while people have been getting wealthier (global GDP rising 100-fold since 1800) (Rees, 2017; Kharas, 2017; Kharas & Hamel, 2018), it becomes extremely challenging to concede that Technology has managed to stem this tide all by itself. My reasoning is based on the I=PAT formula (Ehrlich & Holdren, 1971), where Impact is derived from Population x Affluence x Technology. If both Population and Affluence have increased, for Impact to remain stable that would translate to prodigious achievements in Technology without more material needs, a panacea which has been dubbed as ‘Absolute Decoupling,’ with absolutely no evidence to support it (Jackson & Victor, 2019; Parrique et al. 2019). If my logic falters, one might find solace in the emerging literature delineating the limitations of the EF concept (Blomqvist et al. 2013; Richardson, 2018).

The Unbearable Concept of Limits

Limits to Growth has not flinched when considering the relevance of population dynamics to the process of reaching the biophysical frontiers of our planet. However, population stands as only one of the two variables which constitute what William Rees calls our ‘human enterprise’ (2005; 2009). The other crucial variable is the growth of the physical economy, and that is exactly what will be examined next.

The human population has become increasingly dependent on the functioning of the flows which underpin our entire economic system. In effect, the interconnectedness has become so great that each separate variable becomes indispensable to the fabric of the entire system, with the physicist Yaneer Bar-Yam having argued that our globalized industrial civilization has become too vulnerable to disruptions that cascade down the system and bring about collapse (quoted in Mackenzie, 2008; Servigne & Stevens, 2020).

This couldn’t become clearer than in Limits to Growth, where the authors included in World3, the world industrial growth, and feedback loops connected to its expansion and contraction. For starters, Meadows and colleagues have divided the physical economy into a chart portrayed in figure 6. In detail, they considered the industrial capital to be actual hardware that sustains the economy – machines and factories – generating all of the products (consumer and investment goods) on which society hinges upon. These goods made by the industrial capital become known as industrial output, and although it isn’t referenced in figure 6, they are entirely dependent on the co-factors of production such as labour, energy, management materials, land, water, the services of ecosystems and biogeochemical flows.

The industrial output is then divided into the service capital, the agricultural capital, and the resource-obtaining capital. Moreover, some industrial output ends up in the form of consumer goods, which encompasses clothing, cars, houses, refrigerators, etc. The amount of consumer goods per person is a pivotal indicator of a population’s material well-being. Lastly, there is a portion of the industrial output that goes back into the system in the form of industrial capital, to offset the gradual depreciation of the hardware that undergirds the whole economic system. When all of these flows of physical capital, products, and services are put together they acquire a monetary value, defined by Gross Domestic Product (GDP).

Figure 6: Flows of Physical Capital in the Economy of World3. Retrieved from Limits to Growth, p.37.

With these flows made clear we can now move on to understand the primary interconnections between capital and population, while grasping the feedback loops that can either potentiate their exponential growth or induce their contraction, respectively with the signs (+) or (-).

With this in mind, some of the ways population and capital are affecting each other are described in figures 7 and 8. For starters, as mentioned above, industrial capital generates industrial output, which involves many products, among other things, those that are agricultural inputs, including fertilizers, pesticides, irrigation material… To point out, agricultural inputs will need to be increased if food per person plunges below the desired level. Furthermore, food production is also affected by pollution, derived from both industrial and agricultural activity (leading to the erosion of limits, or loss of carrying capacity, as previously discussed). It is important to realize that both food per capita and pollution determine the mortality of a population. These interconnections are on disp