Recording Response Codability Live: An Investigation of Its Usefulness and Possible Interference
可過錄的回答，是指受訪者第一次就從訪員提供的選項擇一而答。回答的可過錄性（RC）常用以評量問卷與訪問品質。但傳統上 RC必須在訪問結束後額外蒐集。本文認為，RC 應可由訪員在訪問時正確記錄，且不會因此干擾訪問。本研究於 2017 與 2018 年分別進行一次電話調查，訪員於訪問時記錄 RC；2017 年訪問的正確 RC 資料，則於調查結束後由專人獨立蒐集。本文利用這些資料從三方面探討這項主張。研究一分析 2017 年的資料，發現訪員記錄正確率達 93.4％；而且記錄的正確性，隨訪員的記錄經驗增多而提高。研究二探討訪員 RC記錄的可用性，結果發現訪員的 RC 記錄，如同正確的 RC 資料一般，都可有效預測明確態度與默許兩種回答歷程。研究三利用 2018 年調查的訪員評量資料，藉以觀察：多做這項記錄的工作，是否影響訪問的進行。分析結果顯示，訪員不當追問的比例，跟是否做記錄無關，故訪問時記錄 RC 並不影響訪問程序。最後，本文就如何訓練訪員於現場記錄 RC 提出建議。
A codable response is an initial response which matches one of the response options. Among the various interviewer's and respondent's behaviors that survey researchers collect, the information of whether an initial response is codable (response codability, RC) is the most useful in evaluating survey questionnaires and data quality. Yet, because RC is always additionally coded after surveys, along with other behaviors, the information is rarely used in social science research. Instead, social scientists rely on a closely related index of the response behavior, response latency (RL, the length of time taken to answer a question), to tap respondents' underlying mental processes when answering a question. The use of RL brings challenges, however, because respondents may answer before the question is finished, give an extremely lengthy response, or engage in conversations unrelated to the question. Researchers' strategies of dealing with the challenges in turn bring uncertainty to their findings. Such challenges are not likely to emerge when using RC, because RC considers only whether the response is codable or not. This becomes an advantage for applying RC to social science research.
This paper suggests that useful RC information can be collected during survey interviews. It posits that interviewers can accurately record RC live without interfering with the interviewing, because they are used to evaluating the adequacy of every answer in each questionanswer sequence. In addition, to show the usefulness of RC, it argues that RC is associated with response processes in ways parallel to RL. The propositions are assessed in three studies, using two surveys conducted in 2017 and 2018. Interviewers recorded RC live in both surveys. RC information for the 2017 survey was additionally independently collected for verification of the interviewers' RC records.
Study 1 examines hypotheses concerning accuracy of the interviewers' RC records for the 2017 survey. The accuracy rate was 93.4%. Results of three-level logistic regression models further support the hypotheses by showing that the accuracy increased as interviewers recorded for more interviews. Study 2 assesses the usefulness of the interviewers' records for social science research by using the interviewers' RC records and the verification records. Hypotheses on associations of RC with two response processes (attitude accessibility and acquiescence) were formed based on research using RL. That is, a codable response is associated with higher likelihoods of both strong attitudes and acquiescence. Analytical results of the interviewers' records were similar to those of accurate information, and both were able to predict the two processes. Study 3 roughly assesses whether the RC recording task interfered with interviewing, by using incidences of inappropriate probing in the 2018 survey. A cross-tabulation suggests that frequencies of inappropriate probing were related to question types, but not related to the task. In conclusion, this paper contributes to social science research by showing that reliable RC information can be made available by interviewers recording it live. The paper ends with discussions of the consistency of findings concerning the two processes with prior research, of guidelines for interviewer training, and of limitations.