Behavior Coding of October 2018 Agricultural Labor Survey

Cognitive testing on the Agricultural Labor Survey in 2016 revealed that respondents were having difficulty mapping their responses to questions. This led to further inquiries into the data collection process for this survey in the computer-assisted telephone interview (CATI) mode, which is the primary source of completed interviews for this survey. Behavior coding was conducted on recorded CATI interviews and found widespread problems with both interviewer and respondent behaviors. Further efforts were undertaken to improve both the survey’s CATI questionnaire and interviewer performance in administering the survey, and new training and instructions were provided for interviewers on how to conduct standardized interviews. This report examines the extent to which standardized interviews were achieved after making these changes in time for the October 2018 Agricultural Labor Survey data collection. Several factors that impact standardized performance are examined, including characteristics of the questionnaire and questions, and the interviewer’s Data Collection Center. The findings demonstrate that efforts to increase interviewer standardization in question administration improved significantly compared to the previous iteration of the survey in April 2018. However, only a slight majority of all the questions administered by interviewers are read exactly as worded. Thus, more training is still needed for interviewers to continue improving the rate at which standardized interviews are conducted. The results highlight the need for questionnaire designers and Blaise designers to work together to make it easier for interviewers to conduct standardized interviews.