Overview
Three complementary sources on ergodicity economics (EE), the mathematical framework that distinguishes between ensemble averages and time averages to reveal when standard expected-value reasoning misleads individual decision-makers.
The Core Distinction
| Concept | Definition | Meaning |
|---|---|---|
| Ensemble Average | Arithmetic mean across many parallel outcomes | What happens on average across a population at one point in time |
| Time Average | Geometric mean along a single trajectory | What happens to one individual over repeated exposure |
| Ergodic | Ensemble average = time average | Population statistics predict individual outcomes |
| Non-Ergodic | Ensemble average ≠ time average | Population statistics are misleading for individuals |
Why It Matters
Most of economics since Samuelson (1947) implicitly assumes ergodicity — that ensemble averages are valid proxies for individual outcomes. EE shows this is a “kunstgriff” (mathematical trick) imported from Boltzmann’s closed-system physics to open-system economics.
Practical consequence: An observable with multiplicative growth dynamics can have a positive arithmetic mean (ensemble gain) but a negative geometric mean (individual loss). The market shows positive returns but most investors lose money.
The Russian Roulette Test (Dellanna)
The most intuitive illustration: Russian Roulette has an expected value of 0 — you’re dead. The population outcome (many people playing once) diverges catastrophically from the lifetime outcome (one person playing many times).
Key insight: Irreversibility absorbs future gains. Any activity where losses can be permanent (bankruptcy, injury, burnout, broken trust) is non-ergodic. Maximize growth that conserves survival.
Puzzles Resolved by EE (Gadenne)
EE resolves multiple finance anomalies without invoking psychological explanations:
- St. Petersburg Paradox — Time average finite despite infinite ensemble expectation
- Equity Premium Puzzle — Normal fluctuation rate of noisy growth process
- Cooperation Puzzle — Variance reduction without external altruism assumptions
- Inequality Puzzle — Winner-takes-all from high variance (ensemble gain, time-average loss)
- Insurance Paradox — Price ranges where both parties increase their time-average growth
- “Irrational” Risk Aversion — Rational under non-ergodic conditions with finite wealth
Growth Optimality
EE’s default behavioral model: people are first-order growth optimizers based on their growth dynamics.
- Additive dynamic (salary, no bankruptcy risk): Rational to lever up exposure
- Multiplicative dynamic (investments, skin in the game): Rational to be cautious
- Kelly Formula: Special case of EE where growth is multiplicative and transformation is logarithmic
Academic Foundation (Kirstein)
Kirstein’s PhD thesis (TU Dresden, 2019) provides the formal mathematical treatment, tracing EE from Boltzmann’s ergodic hypothesis through Samuelson’s axiom to Peters’ modern resolution. Developed at the London Mathematical Laboratory (LML) with Alexander Adamou and Ole Peters.
Related
- thinking-in-systems — Meadows’ systems thinking provides complementary framework for understanding non-linear dynamics
- high-output-management — Grove’s management framework where understanding growth dynamics matters for organizational leverage