Skip Header

Lynn Weidman, Kathleen T. Ashenfelter
Component ID: #ti1612913974

As part of its efforts to improve the usability of American Community Survey (ACS) data products, the Census Bureau examined two related issues associated with the presentation of information on sampling error for ACS estimates. The first issue concerned the determination of data user preference for the measure of sampling error, margin of error or confidence interval, to be displayed with estimates in ACS data products. To explore this issue, we distributed a Web survey requesting information about each user'€™s background in statistics, their use of ACS data products, and their preference for displaying sampling error. The sampling frame was the ACS Alert email list and subjects were recruited to take the survey via an email. The eligible response rate for the full list was 9.7 percent (324 out of 3,331 possible respondents). The results of the survey show that of those respondents who expressed a preference and do not work for the Census Bureau, 61.2 percent favored the confidence interval and 33.6 favored the margin of error. The second issue was the need for revised definitions of the terms "€œstandard error," "€œconfidence interval," and "€œmargin of error" for use in American Factfinder (AFF). For this purpose, we examined underlying concepts for these terms as they appear in textbooks and throughout statistical literature in an effort to develop alternative definitions that were clearer and more succinct. It is anticipated that improved definitions will contribute to a better understanding of ACS data products in AFF.

Back to Header