Questionnaires & Language Assistance
The census questionnaire is the primary means of collecting data in the decennial census. The questionnaire’s content – mandated by the Constitution and by statutes – has evolved over two hundred years of census taking to comprise a range of demographic and household questions.
The census "short form" is the census questionnaire that the majority of households will receive, while the "long form" is a lengthy, complex form that will be sent to one-in-six households in 2000. Presently, the census long form asks 53 questions.73 The questions range from the name, address, date-of-birth, ancestry and education level to physical-mental limitations, military service, occupation, income, plumbing, heating, and other housing characteristics.
Changes From 1990: The testing process to revise the 2000 questionnaire has been extensive. The new forms are more respondent-friendly than in 1990 and will include changes such as a larger, easier-to-read font, navigational aids to guide the respondent through the questionnaire and respondent instructions directly on the form instead of in a separate guide.74
Long-form mailback rates are lower than short-form rates, and the data acquired through follow-up are less accurate than those found through a mailback response. The long form tends to increase follow-up workload, which increases costs and impairs data quality.75
Questionnaire Subjects: The Bureau and the Office of Management and Budget reviewed in great detail the subjects to include in the 2000 census form. In an effort to shorten the form, only questions that were Congressionally mandated or required for federal agency funding were included. "Mandatory" subjects are cited in federal legislation. Without this information, the government would not be able to carry out legal duties such as distributing federal block grants. Mandatory subjects include:
Age, sex, relationship, race, Hispanic origin, marital status, place of birth, citizenship, year of entry, education, language spoken at home, veteran status, journey to work, place of work, income, grandparents as caregivers (new), support expenditures/health coverage (new).
"Required" subjects data are needed to enforce laws or meet Federal court rules. Required subjects include:
Ancestry, disability, migration, labor force status, industry, occupation, class of worker, work status last year, units in structure, value of home, monthly rent, bedrooms, plumbing facilities, kitchen facilities, telephone, house heating fuel, year moved into unit and shelter costs.
"Programmatic" topics are not mandatory, and are used to set policy or used as legal evidence. There were five programmatic questions dropped for 2000. They dealt with fertility, the labor force, and housing.
Two tests were used to evaluate the proposed content for 2000: the 1996 National Content Survey was designed to test revised question wording, formatting and sequencing and the 1996 Race and Ethnic Targeted Test examined a variety of potential changes to the questions related to race and ethnicity. On March 31, 1997, as required by law, the Bureau submitted a list of subjects planned for inclusion in Census 2000 to the Secretary of Commerce and the Congress.
The proposed 2000 Short Form includes seven subjects, where the 1990 short form included 12 subjects. The proposed 2000 long form includes 34 subjects, where the 1990 long form included 38.
|SUBJECTS PLANNED FOR INCLUSION IN CENSUS 2000|
|100-PERCENT SUBJECTS (Appear on all forms)|
|Tenure (whether home is owned or rented)|
|SAMPLE SUBJECTS (Appear on long forms only)|
Place of birth, citizenship, and year of entry
Education-school enrollment and educational attainment
Residence 5 years ago (migration)
Language spoken at home
Grandparents as caregivers*
Labor force status (current)
Place of work and journey to work
Work status last year
Industry, occupation, and class of worker
Income (previous year)
*New subject for Census 2000.
Units in structure
Number of rooms
Number of bedrooms
Plumbing and kitchen facilities
Year structure built
Year moved into unit
House heating fuel
Value of home
Monthly rent (including congregate housing)
Shelter costs (selected monthly owner costs)
Source: Department of Commerce, Bureau of the Census “Census Questionnaire:D-2.”[the long form].
|1990 CENSUS SUBJECTS DROPPED FOR CENSUS 2000|
|Children ever born (fertility)|
Year last worked (An abbreviated screener will be included with questionsabout industry, occupation, and class of worker; this will allow us toreduce respondent burden and properly define the “experienced civilianlabor force.”)
|Source of water|
The Bureau submittedthe actual questions on March 30, 1998 to the Secretary of Commerce andthe Congress.76 Also, in 2000, the Bureau will use an integrated mailingpackage for the first time. The entire mail package design (questionnaires,envelopes, and motivational slogans) will be compatible and linked to theentire marketing plan.
Multi-languagequestionnaires: The 2000 census forms will be available in six languages:English, Chinese, Korean, Spanish, Tagalog and Vietnamese. The Bureau willmail translated forms to householdsby request. Requests can be made in response to the advance notice letter(pre-census mailing), which will be sent to every mailout address notifyinghouseholds that they will be receiving the census form soon.
The Bureau doesnot plan to have Census 2000 questionnaire forms available in Braille.77Because questionnaire forms will not be available in Braille, the Boardrecommends that outreach efforts include the disability community; theBureau develop outreach and instructional material in Braille and largeprint; and an effort be made to ensure that service agencies serving populationsin need of Braille or large print receive the material.
The languageselection method was based on the probabilities of linguistic isolationamong traditionally hard to count language groups – rather than just thetotal number of people speaking thelanguage – along with the analysis of the potential increases in responserates the printing might achieve. Approximately 3% of all U.S. householdsare considered to be linguisticallyisolated.78 The five foreign languages that the Bureau is using to translatequestionnaire forms allows the Bureau to reach more than 75% of all linguisticallyisolated households(of the 3% that are considered linguistically isolated).79 Coupled with theEnglish language form, the availability of questionnaires in six languagesallows the Bureau to reach 99% of allhouseholds in the U.S.80
Linguistically Isolated Households
Covered by Five Languages
|Language||Estimated Number of Linguistically Isolated Households in 2000||As a Percentage of all Linguistically Isolated Households|
The Bureau submitted the actual questions on March 30, 1998 to the Secretary of Commerce and the Congress.76 Also, in 2000, the Bureau will use an integrated mailing package for the first time. The entire mail package design (questionnaires, envelopes, and motivational slogans) will be compatible and linked to the entire marketing plan.
Multi-language questionnaires: The 2000 census forms will be available in six languages: English, Chinese, Korean, Spanish, Tagalog and Vietnamese. The Bureau will mail translated forms to households by request. Requests can be made in response to the advance notice letter (pre-census mailing), which will be sent to every mailout address notifying households that they will be receiving the census form soon.
The Bureau does not plan to have Census 2000 questionnaire forms available in Braille.77 Because questionnaire forms will not be available in Braille, the Board recommends that outreach efforts include the disability community; the Bureau develop outreach and instructional material in Braille and large print; and an effort be made to ensure that service agencies serving populations in need of Braille or large print receive the material.
The language selection method was based on the probabilities of linguistic isolation among traditionally hard to count language groups – rather than just the total number of people speaking the language – along with the analysis of the potential increases in response rates the printing might achieve. Approximately 3% of all U.S. households are considered to be linguistically isolated.78 The five foreign languages that the Bureau is using to translate questionnaire forms allows the Bureau to reach more than 75% of all linguistically isolated households (of the 3% that are considered linguistically isolated).79 Coupled with the English language form, the availability of questionnaires in six languages allows the Bureau to reach 99% of all households in the U.S.80
Linguistically Isolated Households
Covered by Five Languages
|32 Language Guides for the 1990 Census|
For 2000, the Bureau plans to add Bengali, Dutch and Urdu.
Large Households: In 1990, the census questionnaire, in addition to a household roster for 12 persons, allowed respondents to include 100 percent information for seven of the 12 persons.83 In 2000, the Bureau plans to include a roster for 12 persons, and allow 100 percent information for six persons. The Bureau cited that "69.5 percent of the coverage error came from enumerated housing units…"84 This is what is known as within-household undercount or undercoverage. The within-household undercount indicates that the census roster for many households in 1990 was incomplete. The National Academy, looking forward to 2000, concluded that "improving the quality of coverage within households – households that are increasingly complex and diverse – is crucial."85
Counting an entire household happens in one of three ways – someone in the household completes the census questionnaire, an enumerator completes the questionnaire, or (when other attempts have been exhausted) the Bureau "fills in" the questionnaire with estimates, or "imputations."86 The process of completing the 100 percent census questions for persons within households, as well as listing the persons within the household is "rostering."
The concern for the 100 percent questions and rostering arises from concerns over non-response follow-up and imputations. The 1990 questionnaire had room for detailed information on seven people in a household. In 2000, the questionnaire will have room for detailed information on six people. For all people over that limit the Bureau must contact the household to collect full information.
The majority of households in the United States have fewer than six persons. Yet the implications for large households pose a challenge to the Bureau’s efforts to collect full and accurate data on every household. According to 1990 data, the majority of people residing in households with more than six people are in Black and Hispanic households. To avoid increasing the differential undercount by a less-than-complete enumeration of large households, a dedicated effort at follow-up is necessary.
The Bureau is planning as a part of their coverage edit operation to address questionnaire quality concerns. They will "check completed questionnaires for discrepancies between the number of persons reported and the number of persons for whom information is provided, forms returned where the population count is blank and the number of persons reported is six, mailed forms with household counts of seven or more, and certain households that contain complex living arrangements."87
The Board has not been provided details regarding the follow-up and coverage edit operation for the census questionnaire for households with more than six people. However, the Board believes this operation, if executed well, can reduce the differential undercount.
Congressional Members’ Position: The Congressional Members believe the Bureau made a mistake in eliminating room for detailed information on a seventh person on the questionnaire. According to the 1990 data, the majority of people who reside in households with more than six people are in Black and Hispanic households. A less complete count of large households may contribute to the differential undercount of minority communities.
Given that one third of the hardest-to-count census tracts in 2000 are anticipated by the Bureau to be in three cities – New York, Chicago and Los Angeles88 – concern for differential undercount and the potential of within-household undercount in minority communities merits attention. During the 1990 Census, a seventh person made up 15 percent of the Hispanic population in Los Angeles, compared to only 5 percent of the White population. A seventh person made up 6 percent of both the 1990 Black and Hispanic populations in New York City. In Chicago, without space for the seventh person, detailed information on 7 percent of the Black population and 11 percent of the Hispanic population would have been left off the form.89
Presidential Members’ Position: Initially, the Census Bureau had planned for the 2000 questionnaire to include a complete set of questions for 5 persons in each household. In addition to the 5 question sets, the original form design included a roster for listing the names of up to 12 people and a question on the total number of people living at the address.
The 1990 questionnaire included 7 question sets, and the decision to reduce the count to 5 for 2000 significantly reduced the size of the form to be mailed to each household. The Presidential Members of the Board believe the Bureau made the correct decision in late 1998 when it increased the number of question sets on the form to 6.
The success of this strategy will depend, however, on the effectiveness of the Bureau’s procedures to collect complete information from larger households. This is of particular importance given that the majority of large households in the United States are in traditionally undercounted communities.
Should these follow-up efforts be effectively designed and carried out, the Presidential Members believe the size of the 2000 questionnaire strikes an appropriate balance between accommodating large households and reducing the size and complexity of the form.