The PAIP Score: A Propensity-Adjusted Interviewer Performance Indicator

Evaluating interviewers based on their response rates is complicated in most surveys. By random chance, interviewers may call cases that are more or less difficult to interview. In addition, interviewer response rates can only imperfectly be computed because of the contributions of other interviewers’ prior contacts with those cases (calling a case after a contact from an expert interviewer may pose different difficulties than calling a case after a contact from an inexperienced interviewer). This paper proposes and evaluates an interviewer performance indicator that attempts to repair these weaknesses. The proposed indicator is computed using a three-step algorithm. First, for each active case, available paradata are used to estimate the propensity that the next contact with the case will generate an interview. Second, if the interviewer assigned the case obtains a successful interview on the next contact, the interviewer receives a score of 1 minus the estimated response propensity for the contact; a non-successful contact by the interviewer results in a score of 0 minus the estimated response propensity for the contact. Finally, for each interviewer, the contact-level scores are averaged over all contacts, resulting in a propensity-adjusted interviewer performance (PAIP) score. Addressing an important drawback of previous interviewer performance measures discussed in the literature, this performance indicator gives large credit to the interviewer who obtains success on very difficult cases, and only a small penalty given failure with such cases. The indicator gives only small credit to success on very easy cases and larger penalties given failure with easy cases. This paper illustrates computation of the PAIP score using two different surveys (one face-to-face, one telephone), and assesses the validity of the indicator as a new metric for evaluating the performance of interviewers.