[Paper Review] Statistical Mechanics of Money, Income, and Wealth: A Short Survey

Adrian A. Drǎgulescu, Victor M. Yakovenko|RePEc: Research Papers in Economics|Nov 9, 2002

Complex Systems and Time Series Analysis20 citations

TL;DR

This paper applies statistical mechanics principles to model income and wealth distributions, showing that individual income follows a Boltzmann-Gibbs exponential distribution (equilibrium state) with a Gini coefficient of 1/2, while family income—derived from the convolution of two such distributions—yields a Gini coefficient of 3/8, both matching empirical U.S. and global data remarkably well.

ABSTRACT

In this short paper, we overview and extend the results of our papers cond-mat/0001432, cond-mat/0008305, and cond-mat/0103544, where we use an analogy with statistical physics to describe probability distributions of money, income, and wealth in society. By making a detailed quantitative comparison with the available statistical data, we show that these distributions are described by simple exponential and power-law functions.

Motivation & Objective

• To model the statistical distribution of money, income, and wealth using principles from equilibrium statistical mechanics.
• To test whether the Boltzmann-Gibbs distribution, derived from energy conservation in physics, applies to economic exchange with money conservation.
• To compare theoretical predictions with real-world income and wealth data, particularly from the U.S. IRS and Census data.
• To quantify income inequality using the Gini coefficient and assess whether it aligns with the maximum entropy equilibrium state predicted by the model.
• To investigate the role of power-law tails (referred to as 'Bose condensate') in shaping inequality in high-income populations.

Proposed method

• Model money exchange between agents as a stochastic process with local money conservation, analogous to energy conservation in statistical physics.
• Derive the equilibrium probability distribution of money as $ P(m) = e^{-m/T}/T $, where $ T $ is the average money per agent (effective 'money temperature').
• Use the convolution of two exponential income distributions to model family income: $ P_2(r) = \frac{r}{R^2} e^{-r/R} $, assuming uncorrelated individual incomes.
• Apply the Lorenz curve and Gini coefficient to quantify inequality, comparing theoretical results with empirical data.
• Fit empirical data (IRS and Census) to exponential and power-law functions, identifying a small fraction of high-income individuals as the 'Bose condensate' with weight $ f $.
• Use the delta-function approximation $ f\delta(1-x) $ to model the contribution of the power-law tail to the Lorenz curve.

Experimental results

Research questions

• RQ1 Does the Boltzmann-Gibbs distribution accurately describe the equilibrium distribution of individual income in a society?
• RQ2 How does the income distribution of families emerge from the statistical mechanics of two-earner households?
• RQ3 To what extent do empirical data on income and wealth distributions match the predictions of the statistical mechanics model?
• RQ4 What is the quantitative role of the power-law tail (the 'Bose condensate') in shaping overall income inequality?
• RQ5 Is the observed Gini coefficient in developed economies consistent with the maximum entropy equilibrium predicted by the model?

Key findings

• The distribution of individual income in the U.S. (1997) follows an exponential Boltzmann-Gibbs law $ P_1(r) = e^{-r/R}/R $ for incomes below $100,000/year, with $ R $ equal to the average income.
• For high incomes above $100,000/year, the distribution transitions to a power law, with the fraction of population in this tail being less than 3%.
• The Gini coefficient for individual income is $ G_1 = 1/2 $, which matches IRS data for the past 20 years.
• The family income distribution, derived as the convolution of two exponential distributions, follows $ P_2(r) = \frac{r}{R^2} e^{-r/R} $, and predicts a Gini coefficient of $ G_2 = 3/8 = 37.5\% $, in excellent agreement with U.S. Census data from 1947–1994.
• The 'Bose condensate' fraction $ f $, representing the share of total income in the power-law tail, is 16% for individual income in 1997 and increases over time, with similar values found for wealth distribution.
• Developed market economies (e.g., West Europe, North America) exhibit Gini coefficients close to 37.5%, indicating alignment with the theoretical equilibrium state of maximum entropy, while inequality is higher in other regions, especially post-communist countries.