Gini

Gini most commonly refers to the Gini coefficient, a statistical measure of income or wealth inequality developed by the Italian statistician Corrado Gini in 1912. The coefficient summarises the dispersion of a distribution and is based on the Lorenz curve, which plots the cumulative share of income against the cumulative share of people. Values range from 0 to 1, where 0 indicates perfect equality and 1 indicates maximal inequality (in practice, much lower values are observed). The Gini coefficient is widely used to compare inequality across countries and over time, often alongside other economic indicators.

In economics, the Gini coefficient can be calculated for any set of values, such as income, consumption, or wealth, and can be decomposed to analyze contributions by subgroups with more advanced methods. It is a widely used, though sometimes debated, measure due to sensitivities to population structure, income definition, and the distribution’s tail behavior. Critics note that different distributions can yield the same coefficient and that the measure does not reveal where in the distribution inequality is concentrated.

Gini is also used in machine learning to denote Gini impurity, a criterion for evaluating splits in decision trees. Gini impurity at a node is 1 minus the sum of the squared class probabilities, and lower values indicate purer nodes. This impurity measure helps determine which feature and threshold to split on to improve classification accuracy, and it is commonly associated with the CART algorithm.

The term Gini thus encompasses both the economic inequality measure and a machine learning concept, both named after Corrado Gini.