Our Methodology

Measuring income inequality is difficult and controversial, and both academics and government statisticians have clashed over the appropriate methods for doing so. These clashes are largely about what definitions of income should be used and how income should be allotted to income groups when the data are not directly observable in the distribution of income. Let’s now discuss each in turn.

First, there are many possible definitions of income that can be used. Everyone agrees that certain types of income, such as wages from employment, should be included in income metrics. But other categories are trickier.

How, for example, should in-kind transfers from the government be counted? More than 70 million Americans receive Medicaid, and if Medicaid were not available to them, they would have to pay for insurance or health care out of pocket. As such, Medicaid clearly is an income support program for its recipients.

How, then, should it be valued? The most straightforward method is to take total government expenditures on Medicaid and divide that by the number of recipients to get an average cost of provision per recipient. Yet welfare economics studies find that this kind of in-kind provision is worth less to households than the cash equivalent.

The second source of contention is how to allot income when it’s not directly observable in the distribution of income. A notable example here is tax evasion, which, by its very nature, is largely unobserved. Regardless, economists have estimated the amount of tax evasion in the United States and around the world, but its distribution can significantly change how income is measured.

Another example is retained corporate earnings, which are cash holdings of corporations. Households that own equity in corporations technically benefit from these holdings or will benefit from them in the future. Distributing these holdings in the form of income to households, however, is not straightforward.

The Bureau of Economic Analysis stepped into these debates in 2020 by taking its Personal Income National Account and distributing it among U.S. households in deciles of the income distribution. The Personal Income data include in-kind transfers such as Medicaid but excludes some thornier types of income, including retained corporate earnings. The resulting data series, the Distribution of Personal Income, is the most comprehensive series on income inequality created by the U.S. government.

Equitable Growth opted to use this dataset to track income inequality because of its comprehensive treatment of income and its timeliness. The BEA data include wages from employment, both cash and in-kind transfers from the federal government, returns to financial assets in the form of interest and dividends, income earned by proprietors of private businesses, and income earned from renting out property. It also ends in 2023 (as of this writing in early 2025), compared to the Congressional Budget Office’s Distribution of Income series, for example, which ends in 2021.

One notable omission from this list of income sources is capital gains, or the income households earn from the appreciation of financial assets such as stocks. Capital gains are not measured in the National Accounts, so the Bureau of Economic Analysis does not include them in its distribution even though they are clearly a form of income.

Economists also know that capital gains income is skewed toward higher-income households. The BEA dataset accordingly probably understates high-income groups’ income growth at least somewhat when compared to a series such as the CBO Distribution of Income report, which adds realized capital gains (income gained from the sale of a stock or other financial asset) to income.

In terms of process, we adjust the annual BEA data for inflation using the Personal Consumption Expenditures Index. For simplicity, we use the PCE Index’s reference year, 2017.

To minimize streams of negative income, which can give misleading impressions of the changes to income over time on the component graphs we utilize, we subtract contributions for government social insurance from the total transfers received by households to reflect the net receipt of transfers by households. In other words, for a household that receives Medicaid but also pays into Social Security, their transfer income is the net amount of transfers received.

The Bureau of Economic Analysis also does not break out income for the top 1 percent by component, so we provide only a total for this group. For the most recent year of data, some data are not yet available, and rather than providing a point estimate, the agency provides a range of possible values for the top 1 percent. We therefore take the midpoint of this range of values for the top 1 percent in the most recent year of data.

Notably, the BEA dataset starts in 2000. As such, we use this as a starting point for both income and wealth even though the wealth dataset extends back further so as to make comparisons between the two simpler.

It is a bit less complex to select a dataset to measure wealth inequality. The Federal Reserve’s Distributional Financial Accounts is comprehensive, timely, and trusted. Debates over measuring wealth are tamer than those for income, although disputes do exist. For the purposes of this dashboard, it is unnecessary to delve into those debates.

We make a few adjustments to the Distributional Financial Accounts data to align with our treatment of income. First, we apply the same inflation adjustment to wealth that we do to income, using the Personal Consumption Expenditures Index. There are disagreements about what deflator should be used for wealth, and sometimes economists prefer the Gross Domestic Product deflator or an alternative. We use the PCE Index because it aligns with our income methods and because we think most people care about the purchasing power of wealth over time.

Our graphs show net worth over time.

That is, they total all assets held by households and subtract out all liabilities. We treat one liability—home mortgages—slightly differently than the rest. We subtract home mortgages from the value of real estate wealth to reflect that, for most households, these are two sides of the same coin. The actual equity that people hold in their homes is the home’s potential sale price minus what is still owed on the mortgage. This change is especially important for understanding the Great Recession of 2007–2009, when the bottom 50 percent households were underwater on their mortgages as a group.

All other liabilities documented by the Distributional Financial Accounts are consolidated and listed as “other liabilities.” The largest category by far is consumer credit. This purely negative stream of wealth should be interpreted carefully. Although the data show the growth of these other liabilities accounted for a negative 61 percent of net worth for the bottom 50 percent of the income distribution between 2000 and 2024, they decreased as a share of this group’s net worth from astronomical levels during the Great Recession, when they were nearly 6.5 times larger than their net worth, to just 76 percent of net worth in early 2024.