Revisiting the thermal and superthermal two-class distribution of incomes: A critical perspective

TL;DR

This paper critically examines the empirical evidence for the thermal and superthermal two-class income distribution, arguing that alternative models like the exponential with a log-normal tail may fit the data better and questioning the robustness of previous claims.

Contribution

It provides a formal critique of prior graphical analyses, demonstrating that the thermal and superthermal mixture is not definitively superior to other models like the exponential with a log-normal tail.

Findings

Exponential distribution with a power-law tail fits IRS income data better than log-normal.

Thermal and superthermal mixture's fit can be improved with a log-normal component.

Previous graphical evidence does not conclusively support the thermal and superthermal model.

Abstract

This paper offers a two-pronged critique of the empirical investigation of the income distribution performed by physicists over the past decade. Their finding rely on the graphical analysis of the observed distribution of normalized incomes. Two central observations lead to the conclusion that the majority of incomes are exponentially distributed, but neither each individual piece of evidence nor their concurrent observation robustly proves that the thermal and superthermal mixture fits the observed distribution of incomes better than reasonable alternatives. A formal analysis using popular measures of fit shows that while an exponential distribution with a power-law tail provides a better fit of the IRS income data than the log-normal distribution (often assumed by economists), the thermal and superthermal mixture's fit can be improved upon further by adding a log-normal component. The economic implications of the thermal and superthermal distribution of incomes, and the expanded mixture are explored in the paper.