Achieving Accuracy in the 2000 Census

Census 2000 Education Kit

Achieving Accuracy in the 2000 Census

Undercounting has a long history

Before U.S. entry into World War II, the Census Bureau estimated the number of men who might enlist in the war effort, based on its 1940 population count. The projection turned out to be about 3 percent short for all males, and 13 percent below the actual figures for black men. That was the first year accuracy was measured scientifically.

Voluntary response rates have fallen sharply

In 1970, the non-response rate was 15 percent; in 1980 it was 25 percent; in 1990, fully one-third of all households failed to return their census forms.

Hundreds of thousands of temporary workers (census enumerators) were dispatched in 1990 to find the non-respondents. The job was very difficult:
1. Fewer people were home during the day because of the rise of two-earner households. Some households were visited six times.
2. Shrinking household size meant many more households needed visits. In gated communities or where apartment buildings have locked entries, enumerators had trouble even reaching the households.
3. In some low-income areas, residents' high mistrust of government and census-taker's fear of crime may have reduced the count's accuracy.
Still, the Census Bureau determined it had missed about 8.4 million people and double-counted 4.4 million others. The net undercount of 4 million represented 1.6 percent of the total population.
The undercount was far worse for minorities: for blacks, about 4.4 percent; and for Latinos, nearly 5 percent, or about one person in 20.
The 1990 census was the first to be less accurate than the previous one. From 1940 through 1980, the net undercount declined with each census, despite the falling rate of initial responses.

A census for the 21st century

Congress mandated a redesign for the 2000 census, and the Census Bureau unveiled its plan in February 1996.

Goals of Census 2000 are: to improve accuracy, hold down costs, and eliminate the persistent, disproportionate undercounts of minorities and the poor.
New partnerships with local officials will use their experience and knowledge (of addresses, for example) to improve the count's accuracy and efficiency.
A $100 million paid advertising campaign will seek to build public awareness of community benefits from programs that hinge on census data. In years past the Census Bureau has relied on TV and radio public service announcements, which received only token air time.
A notice that census forms are coming will be mailed to all households.
The form itself will be simpler (with only seven questions) and easier to fill out than previous forms. A longer 52-question form will go to one-sixth of American households to provide in-depth data.
Forms will be placed in public places - grocery stores, libraries, post offices, churches, etc. - to further increase participation.
New form-processing technology will permit the use of people-friendly forms. The technology will also do better at weeding out duplicate responses.
Follow-up mailings will remind people to send their forms in after Census Day, April 1. Duplicate forms may be mailed to all households.
Follow-up visits will run late April into the summer to reach remaining nonresponders.
Enumerators may revisit housing units classified as vacant, to search for possible residents, as well as households that did not report complete information on the census form. The Census Bureau is considering other "coverage improvement programs" in the hardest to count neighborhoods.

Scientific sampling may improve the accuracy of the initial count

In January 1999, the Supreme Court ruled that the Census Act bars the use of sampling methods to compile the state population totals used to apportion seats in the U.S. House of Representatives among the 50 states.
The Census Bureau will make an extraordinary effort to count all households by mail, telephone, or follow-up visits. The result will be used as the basis of congressional apportionment.
After the initial count, the Bureau will send its best enumerators out to re-interview a large number of chosen households in all states and the District of Columbia, starting over with new address lists. This Post Enumeration Survey (PES) will check the quality of the first count and can be used to correct undercounts and overcounts. The corrected, or "adjusted," numbers may be used for non-apportionment purposes such as drawing legislative district boundaries (redistricting) and allocating federal funds.
Congress and the Census Bureau must resolve a disagreement over whether scientific sampling can be used to measure and correct expected undercounts and overcounts in the direct counting process.