The Nonresponse Followup operation assigned enumerators to interview all addresses that had not yet provided information for the 2010 Census in the prior Mailout/Mailback and Update/Leave operations. These interviews were completed on paper questionnaires and collected such information as housing unit status, number of people living at the address, and demographic data about all residents of the household. All completed questionnaires were returned to the Local Census Office for check-in and then shipped to a data capture center.

The Nonresponse Followup quality assurance program ensured enumerators understood and followed interviewing procedures through enumerator observations, reviews of completed materials, and a separate reinterview operation. Crew Leaders completed enumerator observations just after training and reviewed completed materials as they were received. Office clerks also reviewed all forms for completeness before case check-in. Finally, the reinterview operation assigned a separate staff of enumerators to recontact a sample of respondents and verify select data from the original interview. The paper reinterview forms were shipped to data capture centers, and all Nonresponse Followup and reinterview data were electronically delivered to the Nonresponse Followup Matching, Review, and Coding System for comparison and final reinterview outcome coding.

Overall, the Nonresponse Followup Reinterview program was successful in deterring and identifying falsification by the enumerators. We selected 1,894,664 (or 5.9 percent) of the 31,991,588 Nonresponse Followup cases that were eligible for reinterview, with 1,525,297 of them (or 4.8 percent of the eligible universe) selected for the random reinterview type. The random reinterview selection rate was higher than the expected four percent because it was not a simple random sample but was instead a systematic sample stratified by enumerator and starting with one of the first three cases for each enumerator. This design was used in an effort to reinterview all enumerators at least once.

Of all reinterview cases selected, 1,632,798 (or 86.2 percent) received a final outcome of Pass, which means the Nonresponse Followup data were collected with no critical mistakes or intentional falsification. Another 68,043 reinterview cases (or 3.6 percent) found unintentional mistakes, and 12,912 (or 0.7 percent) found falsification. This falsification was found for 1,419 enumerators, which is 0.27 percent of all Nonresponse Followup enumerators. The targeted reinterview samples (Outlier, Supplemental, and Hard Fail) discovered higher rates of mistakes and falsification than the random sample, so we suggest further research on adjusting the reinterview sampling algorithm to more efficiently use reinterview resources.

We did not plan or implement Headquarters monitoring of Local Census Office reinterview coding trends during the operation, but this analysis shows some offices that appeared not to have implemented proper data collection or reinterview coding procedures. Future operations should implement tools to monitor and address these possible office-level issues during data collection.

The results of enumerator observations are incomplete because only half of the observation checklists were delivered in time for data capture. Of the forms received, we saw that 86.5 v percent of enumerators passed the observation and 0.8 percent failed. The remaining enumerators had other or no outcomes marked, likely because they left the operation before an observation could be completed. An automated observation checklist would prevent these data issues and ensure all enumerators are observed in a timely manner.