The American Community Survey has been using ungeocoded Delivery Sequence File units in areas that were either completely or partially mailout/mailback for Census 2000. In 2009, based on an indication of growth from the Population Estimates Program in areas where there was no mailout/mailback, the American Community Survey began to include ungeocoded Delivery Sequence File units in certain counties with a small percentage of non city-style addresses and from fast growing counties with more than 10 percent undercoverage.

A quick decision was needed on whether to continue allowing these records to be in the American Community Survey universe or not. We wanted to determine if the new records resulted in interviews.

• For the January -€“ May 2009 American Community Survey panels, the added ungeocoded records resulted in an overall weighted total interview ratio of 79.3 percent, which, while lower than the panels as a whole, is still a good overall ratio.
• The mail check-in ratio and Computer Assisted Personal Interviewing interview ratio were similar between the added ungeocoded records and all ungeocoded records in the January - May 2009 panels.
• The likelihood of duplication in these records appears to be low. Of the records that resulted in an interview and had an Address Characteristic Type code, 95 percent were found to be in a block where at least 90 percent of units have a city style address.
• To assess if we should not use ungeocoded addresses in some areas, we focused on counties with an interview ratio of less than 50 percent. Given the fairly small sample sizes in these counties and the good overall interview ratio, we decided to continue using ungeocoded records in all of the areas studied.