Results of the 2003 National Census Test of Race and Hispanic Questions
Elizabeth Martin, Dave Sheppard, Michael Bentley, and Claudette Bennett
KEY WORDS: mail questionnaire, experiment, examples, question wording effects
ABSTRACT
The 2003 National Census Test was a mailout/mailback experimental test of alternative versions of questions on Hispanic origin and race. Its objectives were to reduce item nonresponse, reduce the number of race responses that do not fall into categories defined by the Office of Management and Budget (OMB), and improve reporting of detailed Hispanic origins.
Results from the test indicate that:
- Adding the term “origin” and other changes to the question on Hispanic origin improved detailed reporting and reduced generic responses (e.g., “Hispanic”, “Latino,” “Spanish”).
- Adding examples tended to decrease reporting of generic Hispanic origin and race responses.
- Dropping the “Some other race” category reduced the fraction of Hispanics who gave race responses in this category, but also led more of them to skip the question.
- Modified instructions led fewer Hispanics to report as “Some other race”or to skip the question, but led to more missing data for the Hispanic origin question and (for non-Hispanics) the question on race.
- Panels with wording changes to the Hispanic origin question and no “Some Other Race” response category, or with wording changes to the Hispanic origin question, no “Some Other Race” response category, and examples, achieved the greatest reductions in the number of race and Hispanic origin responses that are missing or do not fall into OMB race categories.
CITATION: Elizabeth Martin, Dave Sheppard, Michael Bentley, and Claudette Bennett. 2004. Results of the 2003 National Census Test of Race and Hispanic Questions. 2010 Census Test Memoranda Series, 2003 National Census Test Analysis, No. 3. U. S. Census Bureau. September 7, 2004.
