There is no single number called income inequality, and the Census Bureau's own description says so by listing what it publishes: "Two of the most commonly used income distribution measures are the shares of aggregate household income received by each quintile and the Gini index. In addition to these two measures, the Census Bureau also produces estimates of the ratio of income percentiles; the Theil index, the mean logarithmic deviation of income (MLD), and the Atkinson measure." Six measures of one concept, each sensitive to a different part of the distribution. A change concentrated at the very top moves some of them more than others, so two analysts can honestly report different degrees of change from the same underlying data by choosing different measures.
The larger and less discussed lever is the definition of income itself, and it is where most public argument talks past itself. The Census Bureau's headline estimates rest on a concept it calls money income, and it flags the limitation in the report that carries them: those estimates are "based on the concept of money income, which is pretax and does not account for the value of in-kind transfers." In a footnote to its own inequality section it is blunter: "Money income is calculated pretax, meaning these inequality estimates do not reflect the direct redistributive effects of tax policy."
The Bureau therefore publishes a second series alongside it. Post-tax income is defined as "money income net of federal and state income taxes and credits, as well as payroll taxes (FICA)." Comparing the two shows how much of the measured spread is a story about what the tax system does. For income year 2024 the Bureau reported that "in 2024, after accounting for taxes and credits, aggregate shares of income in the bottom four quintiles were higher, while the share of aggregate income of the highest quintile was lower," and quantified the effect: "Inequality, as measured by the Gini index, was 8.7 percent lower when calculated using post-tax income compared to pretax income." That single comparison is the reason a headline about inequality should always say which income concept it used.
A third lever is household size. The Bureau also publishes equivalence-adjusted estimates, which scale income to account for how many people it supports. Adjusting shifts the shares: "Generally, the income shares in the lowest, second, and third quintiles are higher with equivalence-adjusted income than money income, while the reverse is true for the fourth and highest quintiles," which the Bureau attributes to "the higher concentration of single-person households and smaller household sizes at the lower end of the income distribution." Combining both levers moves the answer further still: the Bureau reported that for 2024 the Gini index was "also 10.2 percent lower using equivalence-adjusted post-tax income instead of money income."
Put together, that is three definitional choices before any data are read: which of six measures, pretax or post-tax income, and adjusted for household size or not. None of the resulting numbers is wrong and none is the number. This is not a reason to distrust the statistics; it is the reason to read the definition alongside the figure, and it is why a comparison between two published inequality figures is meaningless unless both used the same concept.
Two adjacent things this page deliberately does not do. It does not report what the level of inequality currently is, because every one of those measures moves annually and the level is only meaningful attached to a year, a measure and an income concept. And it does not address whether a typical household is doing well, which is a question about the middle of the distribution rather than its spread; the median, and why the median rather than the average, is covered under median household income.
