Glass Barriers in Income Distributions

We introduce glass barriers: income levels at which the density reaches a local minimum, creating structural gaps that impede upward mobility. A glass barrier at percentile p * is simultaneously a local maximum of the Lorenz curvature L ′′ (p * ), a local maximum of asymptotic quantile estimation variance, and a local minimum of the Fisher information for the income quantile-a new information-theoretic characterisation. We propose a nonparametric estimator with bootstrap inference and validate it in a Monte Carlo study: detection rates exceed 90% for moderate and sharp barriers at n ≥ 1,000 with bootstrap coverage approaching nominal at survey-scale sample sizes.