How Systems Evolve: Temporal Dynamics (Axiom A4)
This is Part 5 of the Computational Macrohistory series.
The Earthquake Illusion
On December 17, 2010, Mohamed Bouazizi set himself on fire in front of a government building in Sidi Bouzid, Tunisia. Within weeks, President Ben Ali had fled the country. Within months, Hosni Mubarak had fallen in Egypt. Within a year, the entire Middle East was transformed.
History seems to work this way—long periods of apparent stability punctuated by sudden, dramatic ruptures. Revolutions. Wars. Collapses. The historical record reads like a series of earthquakes: discrete events that arrive without warning and change everything.
But this is an illusion.
Earthquakes don’t come from nowhere. They emerge from tectonic pressure that accumulates continuously over decades or centuries. The earthquake itself—the sudden rupture—is simply the moment when accumulated stress exceeds the fault’s capacity to contain it. The event is discrete; the process is continuous.
The same is true for historical “earthquakes.” The French Revolution didn’t begin on July 14, 1789. It emerged from decades of fiscal deterioration, demographic pressure, elite competition, and ideological delegitimization. The storming of the Bastille was the rupture; the tectonic pressure had been building for generations.
Axiom A4 formalizes this insight. It states that historical systems evolve continuously through time, even when the events we observe appear discontinuous. Our task is to model the underlying dynamics—the slow accumulation of pressure—not just the visible ruptures.
The Illusion of Discontinuity
Why does history seem discontinuous when the underlying processes are continuous?
Narrative Bias
We tell stories about events, not processes. “The Revolution began when...” “The war started because...” “The empire fell after...” Our narrative instincts compress complex dynamics into discrete moments with clear beginnings and endings.
This isn’t wrong—it’s how human cognition works. But it obscures the continuous processes that make events possible.
Threshold Effects
Many social processes only become visible when they cross thresholds. Discontent accumulates invisibly until it erupts in protest. Fiscal stress builds silently until the state defaults. Elite competition intensifies gradually until it explodes into civil war.
Below the threshold: nothing seems to be happening. Above the threshold: everything happens at once.
The discontinuity is real at the level of observation but emerges from continuous dynamics at the level of mechanism.
Selective Memory
We remember events, not conditions. Ask someone about the 20th century and they’ll mention world wars, revolutions, the Cold War’s end. They won’t mention the continuous demographic transitions, technological accumulations, and institutional evolutions that shaped everything.
Events are memorable. Processes are invisible. This shapes how we perceive history.
The CMH Reframe
Computational Macrohistory aims to see beneath the surface—to model the continuous dynamics from which discrete events emerge. We don’t ignore events; we contextualize them as manifestations of underlying processes.
The revolution is real. But so is the decades-long deterioration that made revolution possible.
Axiom A4: The Formal Statement
Here’s how we express temporal dynamics mathematically:
The evolution of macro-historical systems can be approximated as a continuous dynamical process in time.
The formal expression:
This equation is the heart of dynamical modeling. Let me break it down piece by piece.
The State Vector: X
X(t) is the state of the system at time tt t. In CMH, this is our 25-dimensional vector:
Each component captures a different aspect of the social system:
D: Demographics (population growth, age structure, urbanization)
E: Economics (GDP, inequality, unemployment, inflation)
P: Politics (regime type, state capacity, corruption)
S: Social (ethnic fractionalization, education, trust)
Ψ: Collective psychology (legitimacy, optimism, stress)
At any moment, the system occupies a specific point in this 25-dimensional space. As time passes, this point moves—tracing a trajectory through the space.
The Rate of Change: dX/dt
is the derivative of X with respect to time—the instantaneous rate of change. It tells us how fast each variable is changing and in what direction.
If youth unemployment is rising at 2 percentage points per year, that’s captured in
The derivative transforms our static snapshot (where is the system now?) into a dynamic picture (where is the system going?).
The Evolution Operator: F
is the evolution operator—a function that tells us how the rate of change depends on the current state. This is where the “rules” of the system live.
F takes three inputs:
The current state X: Where you are affects where you’re going. High inequality today might increase political instability tomorrow. Large youth bulge now might drive unemployment later. The system’s trajectory depends on its current position.
Structural parameters θ: These are the slow-moving constraints from A2—mode of production, technological level, institutional configuration. They define the “rules of the game” within which dynamics operate.
Stochastic shocks ε(t): Random perturbations—wars, pandemics, discoveries, exceptional individuals. These inject unpredictability into the system’s evolution.
What “Locally Lipschitz” Means
The formal statement requires F to be “locally Lipschitz continuous.” In plain English: small changes in the current state produce small changes in the rate of evolution. The system doesn’t jump wildly from one trajectory to another based on infinitesimal differences.
This technical condition guarantees that our differential equation has a unique solution—that given a starting point, there’s exactly one trajectory the system will follow (before we add stochastic shocks).
It’s a smoothness assumption: the social world, at the macro level, doesn’t exhibit infinite discontinuities.
Why Differential Equations?
Why model history with differential equations—tools developed for physics?
Because differential equations capture exactly what we need: continuous change where the rate of change depends on the current state.
Population growth depends on the current population. Economic growth depends on current economic conditions. Political instability depends on the current political configuration.
This feedback structure—where current state shapes future evolution—is precisely what differential equations model. The mathematics of 18th-century physics turns out to describe 21st-century social dynamics.
The Components of Change
The evolution operator F has three components, each capturing different aspects of how systems change.
Deterministic Dynamics: F(X, θ)
The deterministic component captures predictable, law-like relationships. Given the current state and structural parameters, certain changes are more or less inevitable.
Example: Demographic momentum
A population with high fertility doesn’t immediately stabilize when fertility drops. The large cohorts already born will have children themselves, creating demographic momentum that plays out over decades. This is predictable from current age structure—pure deterministic dynamics.
Example: Debt dynamics
A state with high debt and interest rates exceeding growth faces mathematically certain debt accumulation (absent default or inflation).
The trajectory follows from arithmetic.
Example: Elite overproduction
When elite aspirants grow faster than elite positions, competition intensifies. This creates predictable pressure toward political instability—a dynamic Turchin has documented across multiple civilizations.
The deterministic component includes feedback loops:
Population↑→Resource pressure↑→Wages↓→Fertility↓→Population↓
These cycles emerge from the structure of F itself—from how variables influence each other’s rates of change.
Stochastic Shocks: ε(t)
Not everything is predictable. Wars break out. Pandemics strike. Charismatic leaders emerge. Technologies are invented. These events inject randomness into the system’s trajectory.
We model this as a stochastic process ε(t)—a mathematical representation of random shocks arriving over time.
Key properties of historical shocks:
They’re not Gaussian. Unlike the bell curves of classical statistics, historical shocks have “fat tails”—extreme events occur more frequently than normal distributions predict. World wars, global pandemics, and civilizational collapses are rare but not as rare as Gaussian models suggest.
They’re not uniform. Some periods experience clustered shocks (1914-1945); others are relatively calm (1815-1914). Shock intensity varies over time.
They interact with state. The same shock affects different systems differently. A pandemic hitting a robust state is a challenge; hitting a fragile state, it’s existential. ε(t) multiplies with state-dependent vulnerability.
They can be partially endogenous. Some “shocks” emerge from system dynamics themselves. Wars often erupt when structural conditions (elite competition, resource scarcity) reach critical levels. The distinction between deterministic dynamics and stochastic shocks is partly analytical convenience.
Structural Parameters: θ
The parameter vector θ represents deep structural constraints—the factors that change so slowly they appear constant over typical analytical timeframes.
From A2, we know these include:
Mode of production (agrarian, industrial, post-industrial)
Technological level (military, communications, transportation)
Institutional configuration (state form, legal systems)
Demographic regime (fertility/mortality patterns)
θ defines the “game” being played. Different θ means different dynamics—different F functions, different equilibria, different possible trajectories.
When θ changes, we have a structural transformation. The Industrial Revolution was a shift in θ so profound that pre-industrial dynamics no longer applied. The system didn’t just move to a new state; it moved to a new dynamical regime.
Most CMH analysis holds θ constant—studying dynamics within a structural regime. But understanding transitions between regimes (when θ itself changes) is crucial for long-term historical understanding.
Visualizing Dynamics: Phase Space
How do we visualize a 25-dimensional system evolving through time? We use the concept of phase space (or state space).
What Is Phase Space?
Phase space is an abstract space where each dimension corresponds to one variable in our state vector. A point in phase space represents a complete description of the system’s state at one moment.
For a simple system with two variables (say, population P and resources R), phase space is a 2D plane. Every possible combination of P and R is a point on this plane.
For CMH’s 25-variable system, phase space is 25-dimensional. We can’t visualize it directly, but we can visualize 2D or 3D projections—slices through the full space that reveal key dynamics.
Trajectories
As time passes, the system moves through phase space, tracing a trajectory—a curve showing how the state evolves.
The differential equation
defines a vector field over phase space. At every point, it specifies which direction the system will move and how fast.
Trajectories are the integral curves of this vector field—the paths systems follow as they evolve.
Attractors
An attractor is a region of phase space toward which nearby trajectories converge. If you start near an attractor, you’ll be pulled toward it.
Fixed point attractor: A single point where the system comes to rest.
at this point—no more change. Example: a stable equilibrium where population matches resources, elite positions match elite aspirants, state revenue matches expenditure.
Limit cycle attractor: A closed loop that the system orbits forever. The system never rests but repeats the same pattern endlessly. Example: Turchin’s secular cycles—expansion, stagflation, crisis, depression, expansion again.
Strange attractor: A complex, fractal structure associated with chaotic systems. Trajectories stay within a bounded region but never repeat exactly. Example: possibly the long-term behavior of civilizational dynamics.
Basins of Attraction
Each attractor has a basin of attraction—the set of initial conditions that eventually lead to that attractor.
If phase space has multiple attractors, different starting points lead to different long-term outcomes. The boundaries between basins are critical—small differences in initial conditions can lead to completely different futures.
This is where historical contingency lives. Two similar societies might be in different basins, heading toward entirely different fates.
Repellors
The opposite of attractors: regions that push trajectories away. Unstable equilibria are repellors—even tiny perturbations send the system careening elsewhere.
A society balanced on a knife’s edge (marginal stability) sits near a repellor. Any shock—any perturbation—initiates divergence toward one attractor or another.
Stability and Instability
Central to dynamical analysis is the question of stability: what happens after a small perturbation?
The Intuition
Imagine a ball in a bowl. Push it slightly, and it rolls back to the bottom. The bottom of the bowl is a stable equilibrium—perturbations are self-correcting.
Now imagine a ball balanced on top of a hill. Push it slightly, and it rolls away, faster and faster. The hilltop is an unstable equilibrium—perturbations amplify.
Social systems work similarly. Some configurations are robust: small shocks are absorbed, and the system returns to its previous state. Others are fragile: small shocks trigger cascading changes.
The Jacobian Matrix
Mathematically, we assess stability by linearizing the dynamics around an equilibrium point. The Jacobian matrix J captures how small deviations evolve:
This matrix of partial derivatives tells us, for each variable, how its rate of change responds to small changes in every other variable.
Eigenvalues and Stability
The eigenvalues of J determine stability:
All eigenvalues have negative real parts: Stable equilibrium. Perturbations decay exponentially. The system returns to equilibrium.
Any eigenvalue has positive real part: Unstable equilibrium. Perturbations grow exponentially. The system diverges.
Eigenvalues with zero real parts: Marginal stability. The system is on a knife’s edge—structural changes (bifurcations) may be imminent.
Complex eigenvalues: Oscillatory dynamics. The system spirals toward or away from equilibrium rather than approaching directly.
Marginal Stability and Vulnerability
The most dangerous configuration is marginal stability—when eigenvalues are close to zero. The system isn’t actively unstable, but it has lost its resilience. Small shocks that would normally be absorbed can now trigger large responses.
This describes many pre-revolutionary situations. The ancien régime in 1788 was marginally stable: functioning, but with no reserve capacity. A bad harvest—a shock that France had weathered many times—became the trigger for systemic collapse.
Bifurcations: When Systems Change Qualitatively
Sometimes systems don’t just move to a new state—they change their fundamental character. These qualitative transformations are called bifurcations.
What Is a Bifurcation?
A bifurcation occurs when a small change in parameters causes a qualitative change in the system’s dynamics—the number or type of equilibria changes, a stable equilibrium becomes unstable, or new patterns emerge.
Think of it as a phase transition in the system’s behavior, not just its state.
Saddle-Node Bifurcation
Two equilibria—one stable, one unstable—collide and annihilate each other. Before bifurcation: the system has a stable resting point. After bifurcation: that resting point no longer exists. The system must transition somewhere else, often dramatically.
Historical analogy: A regime loses its stable equilibrium. The configuration that previously “worked”—balancing elite interests, popular demands, fiscal constraints—no longer exists. The system is forced into a new configuration.
Hopf Bifurcation
A stable equilibrium becomes unstable, and a stable limit cycle emerges. The system transitions from rest to oscillation.
Historical analogy: A society that previously maintained steady state begins cycling—periods of expansion followed by crisis, followed by recovery, followed by expansion again. Turchin’s secular cycles may represent dynamics on a limit cycle born from Hopf bifurcation.
Cascading Bifurcations and Chaos
Systems can undergo sequences of bifurcations, each increasing complexity. Period-doubling cascades—where cycles become progressively more complex—are one route to chaos.
In chaotic regimes, the system remains bounded but never repeats. Prediction becomes impossible beyond short horizons. This connects to A5 (endogenous indeterminacy), which we’ll explore next week.
Tipping Points as Bifurcations
The popular concept of “tipping points” corresponds roughly to bifurcations. A tipping point is a threshold beyond which change becomes self-reinforcing and often irreversible.
Climate tipping points, social tipping points, political tipping points—all describe parameter regions where small changes trigger qualitative transformation. The mathematics of bifurcation gives this intuition rigorous form.
Application: Modeling the Arab Spring Dynamics
Let’s apply A4 to our ongoing case study.
Tunisia as a Dynamical System
Consider Tunisia from 2000 to 2011 as a trajectory through phase space. Key state variables:
Youth unemployment (E4): Rising from ~25% to ~30%
Inequality perception: Increasing
Regime legitimacy (Ψ1): Declining
Repressive capacity: Moderate and stagnant
Internet penetration (S3): Rising rapidly
The Trajectory: 2000-2010
Through the 2000s, Tunisia’s trajectory showed continuous deterioration on multiple dimensions:
This wasn’t dramatic—year to year, changes were incremental. But the cumulative effect was movement toward a critical region of phase space.
Approaching the Bifurcation
By late 2010, Tunisia was near a bifurcation point. The eigenvalues of its social Jacobian were approaching zero—the system had lost stability but hadn’t yet collapsed.
In this marginally stable state, the system was exquisitely sensitive to perturbation. Shocks that would have been absorbed in 2000 could trigger cascades in 2010.
The Shock: December 17, 2010
Bouazizi’s self-immolation was a shock ε\varepsilon ε that struck a system poised at bifurcation. The perturbation didn’t decay—it amplified.
From this perturbed state, the trajectory diverged rapidly. Protests spread, repression failed to contain them, the military refused to fire on civilians, and Ben Ali fled.
The system crossed the bifurcation point and transitioned to a new regime.
Counterfactual: Saudi Arabia
Saudi Arabia experienced the same regional shock wave—news of Tunisia and Egypt, calls for protest, similar demographics.
But Saudi Arabia’s position in phase space was different:
Higher legitimacy (religious authority, tribal ties)
Higher repressive capacity (well-funded security services)
Higher distributive capacity (oil wealth → patronage)
The system was further from bifurcation. Its eigenvalues were more negative—perturbations were damped rather than amplified.
Same shock. Different dynamical position. Different outcome.
The Lesson
A4 reframes our understanding of the Arab Spring. The “cause” wasn’t the trigger (Bouazizi) or the mechanism (social media). The cause was the dynamical trajectory that brought Tunisia to a bifurcation point where any spark could ignite conflagration.
Saudi Arabia wasn’t saved by better policy responses in 2011. It was saved by a different trajectory through the 2000s—one that kept the system away from critical regions of phase space.
The Limits of Dynamical Modeling
A4 provides powerful tools, but it comes with significant limitations.
Dimensionality
Twenty-five dimensions is manageable mathematically but impossible to visualize. We work with projections, simplifications, and reduced-order models. Important dynamics might live in dimensions we’re not watching.
Unknown Functional Form
We’ve written F(X,θ,ε) as if we know what F looks like. We don’t. The functional form must be estimated from data—and historical data is sparse, noisy, and potentially biased.
Different assumptions about F lead to different dynamics, different predictions, different policy implications. Model uncertainty is irreducible.
Parameter Estimation
Even with a correct functional form, parameters must be estimated. How strongly does inequality affect instability? How fast does legitimacy decay? These quantitative questions require quantitative answers that historical data may not support.
Chaos and Prediction Horizons
Some dynamical systems exhibit chaos—deterministic but unpredictable. Small errors in initial conditions amplify exponentially, making long-term prediction impossible regardless of model quality.
This isn’t a failure of our models; it’s a feature of the systems themselves. A5 will explore this in depth.
The Map Is Not the Territory
Differential equations are approximations. Real social systems involve discrete individuals making discrete decisions. The continuous approximation works at macro scales but breaks down if we zoom in too far.
A4 is a modeling choice, not a metaphysical claim. We use continuous dynamics because they’re mathematically tractable and empirically useful—not because we believe society is literally a continuous fluid.
Looking Ahead
A4 gives us the mathematical framework to model how systems evolve through time—continuous dynamics punctuated by shocks, trajectories through phase space, stability analysis, and bifurcations.
But there’s a deep problem lurking beneath the surface.
Even if we knew F perfectly—even if we had exact measurements of current state—prediction might still be impossible. Some systems exhibit deterministic chaos: governed by precise laws yet fundamentally unpredictable.
This isn’t ignorance. It’s not insufficient data or inadequate models. It’s a property of the systems themselves—intrinsic indeterminacy emerging from deterministic rules.
That’s Axiom A5: Endogenous Indeterminacy. And it forces us to confront the deepest limits of historical prediction.
Next week: “The Chaos Within: Endogenous Indeterminacy (Axiom A5)”
The Computational Macrohistory Bulletin is a weekly publication exploring the mathematical foundations of quantitative history. Subscribe for updates.