Scaling Laws

When you plot certain quantities on a log-log scale and get a straight line, you've found a power law — a relationship where doubling one thing multiplies another by a fixed factor, regardless of absolute size. Power laws show up in earthquake magnitudes, city sizes, word frequencies, wealth distributions, and the growth rate of human civilization. They're interesting not because any single instance is surprising, but because they keep appearing in systems that have nothing else in common. Either there's a deep structural reason, or we're all pattern-matching the same statistical artifact. Probably both.

The Human Trajectory: Growth That Grows

David Roodman's analysis of global economic output from 10,000 BCE to 2019 — explored in much greater mathematical depth in Superexponential Growth — reveals something stronger than exponential growth. Exponential growth has a fixed doubling time. Human economic growth has had a shrinking doubling time — the bigger the economy gets, the faster it doubles. Plot GWP on a log scale against time on a log scale (years before some reference date), and you get a remarkably straight line. The agricultural and industrial revolutions, usually treated as world-historical ruptures, shrink to gentle ripples on this power-law trend.¹

The mathematics of this are unforgiving. If you fit the best model and project forward, you get infinite output in finite time — around 2047, give or take 16 years. This isn't a prediction; it's a reductio. The model fits the past with extraordinary precision and implies an impossible future. Roodman explored this using stochastic calculus (borrowing from financial mathematics — the same Itô calculus behind the Black-Scholes formula) and found that adding randomness to the model doesn't soften the paradox. In the best-fit stochastic model, the probability of the economy not eventually reaching infinity is one in 10⁶⁹. The explosion is baked into the mathematics of any system whose growth rate increases with its size.¹

The resolution, of course, is that the model must eventually break. Physical constraints, resource limits, or some qualitative transformation will intervene before infinity arrives. But the honest insight is that we don't know what breaks it or when. The model has been fitting well for twelve thousand years. It correctly treats the industrial revolution not as a break with the past but as a continuation of the same trend. The stochastic model is even "surprised" by the industrial revolution — if you fit only pre-1820 data and project forward, 1820 GWP sits at the 95th percentile — but it's not that surprised. The industrial revolution was unlikely but not impossible under the prior trend.

What this means for thinking about the future: the range of possible outcomes is genuinely enormous. A system that has been superexponentially accelerating for millennia might explode (some economic transformation that eclipses the industrial revolution as thoroughly as it eclipsed agriculture), implode (environmental thresholds, runaway AI, civilizational collapse), or do something we can't currently conceptualize. The power law doesn't tell you which. It tells you the stakes are high and the present is a leverage point — which is exactly what Emergence and systems thinking predict about complex systems near instabilities.

The Inescapable Casino: Wealth and Phase Transitions

The most striking application of statistical physics to economics is the affine wealth model developed by Bruce Boghosian and colleagues. Start with a population of agents, each with equal wealth, engaging in pairwise transactions decided by fair coin flips. The poorer agent risks a fraction of their wealth; the richer gains or loses that amount with equal probability. This seems perfectly fair. It isn't.²

Run this "yard sale model" for enough rounds and one agent ends up with everything. Every time. The mechanism is a subtle asymmetry in multiplicative processes: even with a fair coin, the most likely outcome of many rounds is a net transfer from poor to rich. The expected value of each transaction favors the poorer player, but the most probable long-run outcome is the opposite. (This is the same arithmetic trick as the casino game where you win 20% on heads and lose 17% on tails: positive expected value, but the most likely outcome of many rounds is losing money.) The system exhibits symmetry breaking — starting from perfect equality, the first random fluctuation creates an imbalance that amplifies itself.²

The model becomes realistic when you add two parameters: redistribution (a flat pull toward the mean, analogous to wealth taxes and subsidies) and wealth-attained advantage (a bias favoring the richer party, representing systemic economic advantages). The resulting three-parameter "affine wealth model" matches empirical U.S. and European wealth distributions to within fractions of a percent across three decades.²

The really interesting part is the phase transition. When wealth-attained advantage exceeds redistribution, the system undergoes a transition to a "wealth-condensed" state — partial oligarchy, where an infinitesimal fraction of the population holds a finite fraction of total wealth. This is mathematically analogous to a physical phase transition, like water freezing. And it may have actually happened: the shock therapy economics imposed on former Soviet states in the 1990s simultaneously decreased redistribution and increased wealth-attained advantage, potentially pushing those countries across the phase boundary into oligarchy essentially overnight. At least ten of the fifteen former Soviet republics remain oligarchies today.²

The philosophical implications cut deep. The overall shape of the wealth distribution is explained to better than 0.33 percent by a model that completely ignores individual talent, effort, or cleverness. The "virtue" commonly attributed to wealth and the stigma attributed to poverty are, statistically speaking, unjustified. The free market's natural equilibrium isn't a bell curve of outcomes — it's complete oligarchy, moderated only by active redistribution. This isn't a political argument; it's a mathematical one, and it replicates across every country where the data has been tested.

The Gap Between Simulation and Understanding

One theme that surfaces repeatedly in complex systems is the tension between models that simulate well and models that explain well. The climate science community has grappled with this explicitly: Earth system models can reproduce climate behavior with impressive fidelity, but they're so complex that nobody fully understands why they produce the results they do. Isaac Held, in a widely cited 2005 paper, argued that the gap between simulation and understanding was widening — and that bridging it required a hierarchy of models, from the simplest (a dynamical core solving fluid equations alone) to the most complex.³

The analogy Held used is biological: molecular biology made rapid progress because nature provided a hierarchy of organisms of increasing complexity (bacteria, fruit fly, mouse, human) where insights at one level often transfer to the next. Climate science, and complex systems generally, need the same thing — not a single model that captures everything, but a ladder of models where each rung is simple enough to understand and complex enough to be informative.³

This connects to Emergence in an important way. Emergent behavior is, almost by definition, behavior that's easier to simulate than to explain. The Game of Life is Turing-complete but its rules fit on a napkin. NCA can produce any texture but nobody can explain why the learned rules work. The affine wealth model reproduces wealth distributions with three parameters but doesn't tell you which individual gets rich. In each case, the simulation captures the phenomenon while understanding lags behind. The hierarchy-of-models approach suggests that the way forward isn't to make the simulation more detailed but to find the right simplified version that isolates the mechanism.