The Problematic Hoover Index

TL;DR

This paper critically examines the Hoover index, revealing its measurement artifacts, misconceptions, and proposing methods to improve its interpretation in population geography.

Contribution

It provides a comprehensive framework to understand and address the measurement issues and artifacts associated with the Hoover index in population studies.

Findings

H values are often misinterpreted as urbanization.

H remains nearly constant across diverse regions and times.

Measurement artifacts significantly influence H values.

Abstract

The Hoover index H, derived from the distribution of population density, has a long history in population geography. But it is prone to misinterpretation and serious measurement artifacts, some of which have been recognized for years. Here its problems, old and new, are put in a common framework and quantitatively dissected with thought-experiments, experimental reaggregation over a thousand-fold range, modeling, and simulation. H has long been interpreted as urbanization, which simple examples show is incorrect. Values of H near zero are taken as an ideal or primordial state, but the mechanism and distributions of settlement growth show that H can be expected in the range 0.4-0.7 and remain nearly constant across continents, centuries of time, and wide ranges of population density. Technically, the measurement of H is confounded by choices of aggregation scale, inclusion of unpopulated areas, and truncated input data, which produce wide swings in its value. Spatially isotropic modeling and simulation reproduce its trends with aggregation scale and time, nullifying many frontier- or nation-specific interpretations. I suggest methods to recognize and address its major artifacts and point out several trends with valid interpretations.