September 20, 2001

Recent Census Data Significantly Understate The Increase in Income Disparities


PDF of report

If you cannot access the files through the links, right-click on the underlined text, click "Save Link As," download to your directory, and open the document in Adobe Acrobat Reader.

On September 25, the Census Bureau will release its annual report on income and poverty trends among families and households in the United States for calendar year 2000. These data are likely to show that poverty declined and household income increased. The new Census report may also show that across the income spectrum, income grew at roughly the same rate for low, middle- and high-income families, which is the pattern shown by the Census data from 1993 to 1999.

Conclusions on recent trends in income disparities should not, however, be based on these Census data. Virtually all analysts agree that a new Congressional Budget Office study is preferable to the Census data in this regard. The CBO data show that recent gains in income have been heavily concentrated among high-income households, and that income disparities have widened considerably.


What Data Should Be Used for Measuring Income Disparities?

There are two primary sources of data on household income — the Census data and Internal Revenue Service (IRS) data. Each has strengths and weaknesses.

The IRS regularly issues data showing annual income trends in adjusted gross income. These data, which are based on income tax returns, are particularly useful in that they include comprehensive income information for people at the top of the income spectrum. Furthermore, these data are taken from a system in which reporting of income is subject to audit and penalties if income is found to have been reported incorrectly. The IRS data do not, however, cover the entire population; they do not cover households that lack adjusted gross income, have negative income, or have incomes so low that they do not file tax returns.

In contrast, the Census data are based upon a voluntary population-based survey. They cover virtually the entire population. However, the standard data on income that the Census Bureau publishes do not include capital gains income, which is a large source of income for high-income individuals. The Census Bureau also places an upper limit on the amount of certain types of income it counts for any individual, disregarding income above these amounts; this is done for confidentiality and other reasons. For example, the highest salary that can be recorded is $999,999. Anyone with a salary above this amount is recorded as having a salary of $999,999. As a result of these and other factors, the Census data significantly understate the income of those at the top of the income spectrum as well as recent increases in income among this group.

A third source of information on income and tax trends is a data series that the Congressional Budget Office maintains. CBO combines the best features of both the IRS and Census data to create what most experts would agree is the single best data series on household income.(1) The CBO data series is based heavily on the IRS data for its information about income at the top of the spectrum, while making several improvements in the IRS data, in part by combining it with Census data.


What Are the Recent Trends in Income Disparities?

Table 1 compares recent trends in real income growth (i.e., income growth after adjustment for inflation) among low, middle- and high-income Americans in the data that CBO published to the trends shown in the Census data.(2) The table shows the change in average income between selected years for each fifth of the population.(3) The first row displays the percentage change in average income between 1993, the last year in which income fell due to the most recent recession, and 1999, the last year for which Census data are now available. (Data for 2000 will be released on September 25.) The next category displays both Census and CBO data for the percentage change in income between 1993 and 1997, the last year for which CBO data are available.

The Census data show fairly similar income growth since 1993 for each fifth of households. Between 1993 and 1997, income growth was 7.8 percent among the poorest fifth of U.S. households and 9.2 percent among the highest income fifth of households, while income growth for the broad middle class was slightly less. In sharp contrast, the CBO data for the 1993 to 1997 period show income growth among the poorest fifth of the population to have been only 3.8 percent, while income grew 20.6 percent among the highest income fifth — more than five times as great a percentage. Among the top one percent of the population, the CBO data show real income grew an astonishing 51 percent.

Comparison of Income Growth Rates Between Census and CBO For Selected Years

Lowest Fifth

Second Fifth

Middle Fifth

Fourth Fifth

Highest Fifth

Top 5%

Top 1%



























Indeed, according to the CBO data, income disparities grew more sharply between 1995 and 1997 than in any other two-year period since 1979 (the first year included in the CBO study). This suggests the possibility that the growth in income disparities may have accelerated in the latter half of 1990s. Specifically, from 1995 to 1997, the top one percent of the population racked up a gain of 33.6 percent in average before-tax income, far outpacing the 3.6 percent gain for the middle fifth of the population and the 0.2 percent gain for the bottom fifth.

Census data for 1998 and 1999 indicate that all income groups have gained somewhat since 1997. The CBO study concludes, however, that at least through 1999, those gains continued to be concentrated at the top. CBO reports: "Although CBO's database on taxes and income ends in 1997, information from tax returns suggests that the recent rapid rise in the share of income going to the top of the distribution and in the share of individual income taxes those households pay has continued in 1998 and 1999."

For a thorough examination of the CBO study and why it is preferable to other data sources, see the Center on Budget and Policy Priorities' May 2001 publication, "Pathbreaking CBO Study Shows Dramatic Increases in Income Disparities in 1980s and 1990s."

End notes:

1. Two other, somewhat more technical, differences between the CBO data and Census data should be noted. First, CBO places households into income categories after making an adjustment for the number of people in the household. Take two households with incomes of $20,000. Household A has four people in it; Household B has one person in it. Recognizing that it is harder for four people to live on $20,000 than it is for one person, CBO's method considers the household with one person to be better off, and to have higher adjusted household income, than the household with four people. The Census Bureau makes no adjustments for family size, so that a household of four with an income of $20,000 can be placed in a higher income category than a single-person household with income of $19,500.

Second, CBO places an equal number of people into each income fifth. The Census Bureau places equal numbers of households into each income fifth. Since many low-income households consist of elderly individuals, under the Census approach the bottom fifth of the population contains a smaller number of people than each of the other fifths of the population do. As a result, the Census approach understates the amount of income going to the bottom fifth of the population.

2. Historical Effective Tax Rates, 1979-1997. Congressional Budget Office, May 2001.

3. As noted earlier, CBO arrays people by income and divides them into fifths, while the Census Bureau bases the fifths on households.