This report documents results of an experiment conducted during Census 2000, in which1990-style census short forms (preserving 1990 question wording, categories, order, and format) were mailed to an experimental sample of 10,500 households. A control panel of about 25,000 households received Census 2000 questionnaires. A simplified version of the pre-edits used in Census 2000 production were applied to the data. Missing data were not imputed or allocated.

Comparisons of the two panels show that questionnaire changes made in Census 2000 substantially improved the completeness of race and Hispanic origin reporting in mail questionnaires. It also affected race reporting, especially by Hispanics. Reports of two or more races more than doubled in response to the “mark one or more” instruction. The fraction of Hispanics reporting as White was higher by about 10 percentage points, and reporting as Some other race was lower by the same amount, in Census 2000-style forms.

Although the same fraction reported as Hispanic in both forms, the 2000-style questionnaires elicited fewer reports of specific Hispanic groups, and more reports of general Hispanic identity (e.g., Hispanic, Latino, Spanish) than the 1990-style questionnaires.

This report also includes two supplemental analyses conducted after the original report was issued in 2002. Supplemental analysis 1 assesses the effect on race reporting of spelling out “American” in one or more race categories. In 2000-style short forms, which spelled out “American” in the “American Indian and Alaska Native” category, twice as many people wrote “American” in one of the race write-in spaces compared to the 1990-style short forms, which did not spell it out anywhere. Supplemental analysis 2 assesses response rate differences between a matrix (2000-style) and person-space (1990-style) form. There is no overall difference in return rates between the panels, but the person-space 2000-style panel obtained a significantly higher return rate in the LCA (low coverage area) stratum. Item nonresponse rates differ significantly between the two forms, although many design features in addition to the structure of the questionnaire affected these differences. For five short form items, the 2000-style form performed better, and for two (age and name) it performed worse.