Systems: Overshoot and Collapse

ELI5/TLDR

When you remove the brakes from a system that naturally accelerates—predator control, healthcare that lowers death rates, industrial expansion—the population shoots up. But the ground under your feet starts eroding. By the time you feel the floor disappearing, it’s usually too late. This is overshoot and collapse, and it’s hiding in the structure of most growing systems.

The Full Story

The Kaibab Plateau: Where Predators Were the Emergency Brake

Meadows opens with the Kaibab deer case, which feels like a biology textbook story until you realize it’s a masterclass in feedback loops. In equilibrium, deer births perfectly balanced predator kills. The system wasn’t stable because it was static. It was stable because predators acted as a thermostat. Too many deer meant well-fed predators. Fewer deer meant hungry predators and births again. A negative feedback loop, continuously adjusting.

Then humans removed the predators, thinking we were helping. The positive birth feedback loop ran unchecked for 20 years before it became visible. Exponential growth looked gentle at first—more births per year, yes, but still just gentle doubling. The explosive part only shows up when you’re already committed. By the time the deer population exploded, they were eating the grassland faster than it could regenerate. The carrying capacity wasn’t a hard wall. It was a sponge that dissolved.

Here’s the insight Meadows emphasizes: the system didn’t collapse because there wasn’t enough grass. It collapsed because the grass couldn’t regenerate while being eaten. A deeper positive feedback loop kicked in—the more depleted the vegetation, the longer it took to regrow, the more pressure to clear more land. The terrain itself got damaged.

The Structure of Overshoot and Collapse

Meadows then steps back and describes four generic patterns for how growth interacts with limits:

Infinite World: The carrying capacity grows as fast as demand. Think: we keep finding new oil fields, new agricultural techniques. This works forever if your time horizon is short enough. Once resources tighten, this pattern breaks.

Sigmoid (S-curve): Growth is exponential at first, then slows as you approach the ceiling. The population has immediate feedback about how close it is to saturation. Bacteria in a petri dish behave this way—they sense nutrient scarcity directly and slow their reproduction. It’s smooth, stable, boring.

Oscillation: The system gets a signal it’s overextended, but not quite in time. It overshoots, then falls back, overshoots again. This can dampen down to equilibrium or cycle forever. Think: fish populations in a lake that surge one year and crash the next, then surge again. No permanent damage yet, just rhythmic imbalance.

Overshoot and Collapse: This is when three things align. First, a positive growth loop running unchecked. Second, the negative feedback signal arrives late—after the overshoot. Third, and most dangerous, the carrying capacity itself erodes irreversibly. By the time the system realizes it’s in trouble, it’s eaten its own substrate.

The Tsembaga: A Ritual Thermostat That Actually Worked

Meadows then describes Roy Rappaport’s anthropological study of the Tsembaga people in New Guinea. They practice slash-and-burn agriculture on a rainforest slope—an extraordinarily difficult system to balance. A naive model predicts they should have collapsed centuries ago.

But they didn’t. Why? Rappaport hypothesized that their pig festivals served not just ceremonial purposes, but ecological regulation. The Tsembaga kept pigs alongside their human population. Pigs reproduce faster than people. Every 8-10 years, pig numbers spike high enough that the women of the tribe start complaining—pigs break into gardens, eat stored food, become a burden. The men gather and declare it’s time for a festival. They kill 80% of the pigs. Immediately after the feast, they declare war on neighboring tribes, and about 12-20% of young males die.

When Meadows and her team modeled this system, they got regular 10-11 year cycles. The pig population shot up faster than human population, serving as a leading indicator. The festivals and wars kept both populations oscillating well below the true carrying capacity of the land. The soil fertility never crashed because the system had a built-in thermostat.

Then Meadows explores what happens when that thermostat is removed. If a health clinic lowers death rates, births stay constant, and the net growth rate strengthens. The cycles compress and eventually fail—the system enters full collapse mode. When the Australian government abolished inter-tribal warfare (“civilized people don’t do that”), the human population grew without oscillation and the system eventually exceeded its carrying capacity.

The Tsembaga weren’t wise in a way we’d celebrate. But their system was stable in a way ours is not.

The World Model: Three Feedback Loops Running a World

Meadows then moves to World3, the model underlying Limits to Growth. It’s a 250-equation model of planetary systems. She shows a simplified diagram to convey the main loops:

Population growth: birth rate minus death rate, both influenced by industrial development (richer → fewer kids), healthcare, food availability, pollution.
Industrial capital growth: output minus depreciation, where surplus output gets reinvested to build more factories, tractors, roads.
Resource depletion and pollution generation as side effects of industrial output.

The key insight: there are two dominant positive feedback loops. More people → more births. More industrial capacity → more output → more capital investment → more capacity. Both are running at world averages of ~2% and ~7% per year. Exponential growth.

The negative loops that might stop this are weak or slow. Education and family planning can reduce fertility, but with a 20-50 year delay. Food production can increase with better technology, but pollution generation and resource depletion are unambiguous: one-way erosion.

When Meadows runs the World3 model with no policy changes, it produces a classic overshoot and collapse: growth peaks around 2050, then a gradual decline as resource constraints throttle industrial output, which then throttles everything else. Population continues rising for a while (more delays in the system), then falls as death rates climb. Food per capita rises for decades, then drops. Pollution keeps rising even as industrial output falls (it’s a lagging indicator).

The collapse is more gradual than the Kaibab crash because humans have choices—they can shift resource allocation, change fiscal priorities, innovate. But the core dynamic is identical: growth until you hit something irreversible, then contraction.

Possible Futures

But Meadows is careful to note that overshoot and collapse is not the only story a system can tell. If you remove all limits (assume resources grow, pollution abates automatically, food yields increase), the model produces exponential growth forever. Unrealistic, but possible in the equations.

If you assume limits have cost—it takes industrial capital to reduce pollution or restore resources—then you get a rise-and-fall pattern in per-capita output. Resources last longer, but at the cost of never achieving American-diet living standards for everyone.

The final run Meadows shows is her favorite: a stable equilibrium. She simply assumes people voluntarily choose smaller families and that once per-capita industrial output reaches a European level, growth stops. No crash needed. The system smooths into equilibrium: population at 6 billion, food abundant, pollution minimal, resources depleting slowly. It requires direct intervention in the two positive growth loops before they hit limits, not after.

The point: the model doesn’t determine the future. It reveals what happens under different assumptions about how we respond to signals.

Key Takeaways

Stocks and flows: A level (deer population, soil nutrients, oil reserves) changes when inflows exceed outflows. Equilibrium is when they balance, not when the level stops moving.
Positive feedback: More → more. Birth loops, capital investment loops. These create exponential growth. They’re not inherently bad; they’re just unbalanced.
Negative feedback: More → less. Scarcity → slower growth. These restore balance. But they only work if they get a signal fast enough.
Delays: Every level in a system is a delay. Deer can’t appear instantly; food takes time to regenerate. These delays cause overshoot.
Carrying capacity: Not a hard ceiling. It’s the sustainable number when supply and demand balance. It can erode (degraded soil regenerates slowly) or grow (innovation).
Overshoot requires three conditions: (1) an unchecked positive growth loop, (2) a delayed negative response, (3) irreversible damage to the carrying capacity.
Signal latency: If the system doesn’t know it’s overextended until after the damage is visible, it’s too late. The worst collapses have this hidden.
Policy intervention: You can flatten any exponential curve by interrupting the positive loops before they hit limits, not after.

Claude’s Take

This is the foundational systems-thinking lecture, and it deserves the reputation. Meadows doesn’t just describe overshoot; she shows you how to see it. The three parallel case studies—Kaibab, Tsembaga, World3—are pedagogically brilliant. Same dynamic structure, wildly different contexts. It drives home that overshoot isn’t a biology problem or an economics problem. It’s a structure problem.

The Tsembaga example is particularly unnerving because it works. We tend to think of wisdom traditions as quaint or inefficient, but here’s an entire society that stabilized itself through cultural ritual for thousands of years, and we broke it by imposing our idea of what civilization should look like (no war, but yes healthcare without the balancing mechanism). That’s not a knock on progress, but it’s a humbling reminder that you can’t just remove parts of a feedback loop and expect the system to stay stable.

The World3 model is the part people focus on, and for good reason—it’s terrifying. But Meadows is careful to show that the collapse isn’t determined. It’s what happens if you do nothing. It’s what happens if you try to solve it after you’re already overshooting. The stable equilibrium run is almost apologetic—yes, you’d have to deliberately choose to stop growth, and yes, that requires coordination and sacrifice. But it works. The system doesn’t naturally find equilibrium. You have to deliberately interrupt the positive loops.

What Meadows doesn’t address much (and this is fair for a systems talk) is the political economy of actually doing this. Voluntary family planning in a wealthy society requires collective action and cultural values that are notoriously hard to sustain. Stopping industrial growth means someone’s profits don’t multiply. The model shows what’s possible. The world shows what’s probable.

Score: 8/10. This is top-tier foundational work. Meadows is the clearest thinker on this material, and her teaching is direct without being simplistic. The three case studies lock the concepts in. The only reason it’s not higher is that it’s a lecture—it requires sustained attention and comfort with abstraction, and it doesn’t offer immediate action steps beyond “yes, the growth loops are unsustainable, and yes, you have to deliberately choose something different.” But for understanding how systems behave, this is essential.