NEWS


inequality 0.2.0

This release responds to feedback from Frank Cowell and Emmanuel Flachaire (personal communication, 1 May 2026) on the v0.1.0 release. Two gaps were flagged: confidence intervals were available only for the Gini, and the package rejected non-positive values for the Gini and the top shares. A follow-up internal audit produced several smaller fixes that ship together.

Confidence intervals on every inequality measure

Bootstrap confidence intervals are now available on every inequality function via ci = TRUE. Each function gains ci, R, and level arguments matching the existing Gini API. Results are returned in ci_lower, ci_upper, and se fields on the output object and shown by the print method.

Functions extended: iq_theil(), iq_atkinson(), iq_sgini(), iq_palma(), iq_hoover(), iq_kolm(), iq_percentile_ratio(), iq_polarisation(), iq_shares(), iq_concentration(), iq_kakwani(), and iq_poverty().

The bootstrap uses probability-proportional resampling, so survey weights flow through to the variance, not just the point estimate.

iq_compare() runs one bootstrap loop, propagates CIs to every row

When ci = TRUE, iq_compare() now runs a single resample loop and attaches ci_lower and ci_upper columns to every row of the table. The old gini_ci field is removed. The table now also covers S-Gini, Kolm, and Wolfson (12 measures, up from 9).

Negative values are now supported via negatives = "keep"

Functions that are mathematically defined for distributions containing negative values now accept negatives = c("error", "keep"), with "error" as the default for back-compatibility. With negatives = "keep":

iq_kolm() already worked for negative values and is unchanged.

iq_atkinson() and iq_theil() continue to require strictly positive values: they involve log(x) or x^(1 - epsilon) for which the formula is mathematically undefined at zero or below. The error message now documents this explicitly.

Bug fixes

New features (audit follow-up)

Acknowledgements

Thanks to Frank Cowell and Emmanuel Flachaire for the careful read and the two-line list of gaps.

inequality 0.1.0 (2026-04-20)

Inequality indices

Distribution and decomposition

Poverty

Fiscal

Utilities