Introduction

According to the results from the Census 2000 Supplementary Survey (C2SS), the foreign bornpopulation grew by 57 percent since 1990 and approximately 45 million people aged five years and older spoke a language other than English at home. Currently, there is little research investigating differences in data quality between English and non-English speaking households. To better understand the scope and depth of these differences, this paper reports results from a quantitative assessment of differences between English and non-English speaking households in the American Community Survey (ACS) using traditional data quality measures.

The ACS, a survey proposed by the Census Bureau to replace the decennial census long form, will collect social, demographic, economic, and housing data about the nation throughout the decade rather than once every ten years. The ACS will be a monthly survey, sampling approximately 250,000 addresses, and data will be collected using mail, telephone and personal visit methodologies providing varying degrees of language assistance. It is critical that high quality data be collected for all geographic areas and all population groups. The Census Bureau is interested in developing research strategies and measures of data quality that can be used to assess and improve the quality of demographic survey data obtained from people whose primary language is not English and who have little or no knowledge of English.

This research was undertaken to assess the completeness of data collected from non-English speaking households using traditional data quality measures to measure item nonresponse. The research focuses on non-English speaking households with the lowest levels of English-speaking proficiency because we expect that these households face the greatest challenges in understanding and answering survey questions.

While the quantitative measures of data quality provided in this report provide a useful and valuable assessment of data completeness, it is only a partial assessment of data quality. Other assessments from a qualitative standpoint would be necessary to provide additional insight into the quality of data obtained from non-English speaking households. For example, preliminary findings from recent focus groups and cognitive interviews indicate that how ACS interviews are conducted by Spanish-speaking interviewers and the way in which Spanish-speaking respondents interpret and respond to questions on the ACS Spanish questionnaire could potentially lead to errors in the data collected. This result would go largely undetected by item nonresponse analysis and other quantitative analyses of data quality.