In a census, some people are not counted. In 1990 we estimated the number of people not counted by conducting a post-enumeration survey separately from the census and comparing the findings. The net undercount for 1990 is the difference between how many people were actually counted in the 1990 Census (the official count) and the estimate of how many people lived in the United States at that time (the adjusted count). The adjusted count is an estimate based on the survey and the census enumeration itself. If the adjusted count is greater than the official count, the difference is called an undercount. In a few cases, the official count is greater than the adjusted count. When that happens, we have an overcount, which we designate as a negative number. The net undercount rate is the ratio of the net undercount to the adjusted count; it is often expressed as a percent.

The data indicate that populations were undercounted at different rates. In general, Blacks, American Indians and Alaskan Natives, Asians and Pacific Islanders, and Hispanics were missed at higher rates than Whites.

To cite an actual example, in the United States overall, we estimate a net undercount of about 4.0 million people in 1990, giving us an undercount rate of approximately 1.6%. The estimate for Whites is about 1.8 million, for a rate of 0.9%. However, although fewer Blacks (1.4 million) than Whites were missed, they were missed at a higher rate, approximately 4.4%. Children were also disproportionately missed in the last census. The net undercount for children - about 3.2% - is twice the overall rate.

When comparing the undercounts or undercount rates for two geographic areas or demographic groups, one should be aware that the numbers are measurements subject to sampling error.