In 2010, the American Community Survey (ACS) began releasing all 5-year estimates for census tracts, census block groups, and small governmental units without any data quality filtering. However, concerns were raised prior to that about the reliability of many of the estimates likely to be included in the data products. Based on a requested by Director Steve H. Murdock in the fall of 2008, a series of assessments were conducted to look at the possible quality or reliability of 5-year ACS data. The only 5-year ACS estimates available in 2008 were from the Multiyear Estimate Study (MYES) conducted in 2006, which included data for 34 of the ACS test counties and covered the period from 1999 through 2005. The assumption is that the patterns of quality seen in these data will be consistent with the quality of the first 5-year estimates released in 2010. This report summarizes results of two of the assessments undertaken that were presented to the director on November 20, 2008.

The ACS has been described as the replacement for the Census Long Form. The ACS will ultimately release multiyear estimates for the same small areas that the Long Form did. The quality of the Census Long Form estimates is therefore an important benchmark for comparing projected 5-year reliability measures for the ACS to comparable Census 2000 Long Form measures.

The error in survey estimates typically includes two components: sampling and nonsampling error. This assessment is limited to aspects of sampling error as measured by the coefficient of variation (CV). The CV is a relative measure of reliability defined as the ratio of the standard error of an estimate to the estimate itself. For example, a 10 percent CV on an estimate of a 20 percent population characteristic suggests a standard error equal to 2 percent.

This assessment will address three questions surrounding the issue of the reliability of ACS estimates for small areas.

• Question 1: Is the reliability of ACS multiyear estimates adequate?
• Question 2: Does the reliability of ACS estimates vary by size of area and topic?
• Question 3: How does the ACS reliability compare to the Census 2000 Long Form?

To answer the first two questions, we quantified the reliability of ACS 5-year estimates for small areas according to specific rules based on the CV, resulting in a classification of "€œHigh", "Medium", or "€œLow". Results were summarized for a subset of estimates from the Data Profiles by size of area. Similarly, results were summarized by group of estimates usually referred to as subject or topic areas. Results are displayed for subjects such as race (major groups) and housing tenure.