We provide stylized facts on the evolving shape dynamics of the US personal
income distribution from 1962 to 2000. Based on adaptive kernel density
estimation, we propose an adaptive bootstrap test for multimodality. Our
results indicate that multimodality has been a predominant feature of the US
income distribution. Both the number and location of modes change over time,
revealing rich distributional dynamics and strong heterogeneity. Decomposing
the sample by age, education, gender and race, all groups exhibit multiple
modes and follow distinct distributional patterns. For all groups and for the
population as a whole, income distribution improved over the 1960s, 1970s and
1980s, but deteriorated dramatically in the 1990s.
JEL Classification Numbers: C12, C14, C15, E25
Keywords: Adaptive kernel method, adaptive multimodality test, bootstrap, income
distribution, nonparametric density estimation, population heterogeneity