In recent years, the reduction in response rates to large government surveys has led to an increased importance in measuring nonresponse bias. Meanwhile, statistical agencies have been under pressure to spend limited resources efficiently. This has led to interest in the use of interviewer-generated paradata to analyze nonresponse bias. It has also led to an interest in using administrative or previously collected survey data as part of the data analysis process. This methodological brief uses survey data linked to previously collected household data to assess the agreement between interviewer-generated observations and previously collected data for nonrespondents. As part of this evaluation, we explore whether there are differences in agreement by the type of nonrespondent household, and the impact of substituting this previously collected data for variables which are currently imputed.