바로가기 및 건너띄기 링크
본문 바로가기
주메뉴 바로가기


발간자료

연구보고서

홈 발간자료 연구보고서
인쇄

SNS공유하기

해당 컨텐츠 트위터에 추가하기 해당 컨텐츠 페이스북에 추가하기

Methodology for the Improvement of Panel Data Quality-CAPI(Computer-Assisted Personal Interview) Effects-
  • 저자 최효미, 이상호, 성재민, 김기민, 배기준
  • 출판일 2013.03.14
  • 판매가 비매품
  • 재고 재고없음
  • 페이지수 114 페이지
  • ISBN 978-89-7356-931-1
  • 목차
    Executive Summary ix

    Chapter 1 Introduction 1

    Chapter 2 Empirical study on the effect of Computer Assisted Personal
    Interviewing (CAPI)-With a Focus on the Korean Labor and Income Panel
    Study(KLIPS) CAPI Experiment Data 4

    1. Introduction 4
    2. Outline of the KLIPS Test Run Design 6
    3. Empirical Analysis of the Effects of CAPI 9
    4. Summary and conclusion 54

    Chapter 3 The Effect of CAPI Introduction on KLIPS-Focusing on Data Integrity
    Between Survey Waves 57

    1. Introduction 57
    2. Transition to CAPI 58
    3. Effect of CAPI introduction on survey performance 62
    4. Analysis of item response and missing errors 71
    5. Effect on changes to response propensity 79
    6. Summary and conclusion 85

    Chapter 4 Event History Calendar Program Development andAnticipated
    Effects-Main Functions and Characteristics of the KLI CAPI-EHC 87

    1. Introduction 87
    2. EHC effects as described in previous studies 88
    3. Main functions and characteristics of the KLI CAPI-EHC 92
    4. Summary and conclusion 100

    Chapter 5 General conclusion 102

    References 106
  • 요약
    Compared to cross-sectional studies, longitudinal studies present the possibility of a greater range of non-sampling errors. Because of this, the entire process covering study design, fieldwork, and data processing must be carefully planned in advance. Leading longitudinal studies in other countries have conducted various methodological studies for the improvement of the data that is collected and released. In contrast, it is difficult to find instances of methodological studies in the field of longitudinal research in Korea. This study is the first to study mode effects in Korean longitudinal studies. In essence, it concerns the empirical analysis of the question, “How did the introduction of CAPI (computer-assisted personal interviewing) affect the data quality of certain longitudinal studies?” This question is examined using KLI (Korea Labor Institute)s KLIPS (Korean Labor and Income Panel Study), which is among the first of Koreas longitudinal studies to have introduced the CAPI mode. ◈ The Introduction of CAPI mode in Korean longitudinal studies In 2005, the KLoSA (Korean Longitudinal Study of Ageing) and WPS (Workplace Panel Study) conducted by KLI became the first Korean longitudinal studies to adopt the CAPI mode. The project was pursued by KLI in order to hasten fieldwork while reducing interviewer and respondent error, thus improving overall data quality. During the 5 following years, CAPI has emerged as the mode of choice among Korean longitudinal studies, with 9 studies currently employing CAPI systems in their fieldwork. However, despite this rapid spread in adoption, there have been no follow-up or experimental research to empirically determine the effects of the CAPI mode. ◈ The KLIPS Wave 10 Test Run In 2007, KLI conducted an experimental test run on 560 households selected from the KLIPS sample residing in the Daejeon-Chungcheong regions to ascertain the effects of a change of interview mode to CAPI. The objective was to test whether there were any significant changes in the response behaviors of the test group (CAPI) and the comparison group (PAPI). The analysis was conducted solely on the basis of this carefully planned experimental data, using both pre- and post- supervisor review versions of the data. The main findings from this exercise are as follows: First, the interview mode had no significant effect on response rates. Of the households selected for the experiment, 212 PAPI and 218 CAPI households returned successful interviews, with 58 partial successes. These ‘partial successes’ were mainly due to lack of preparation time for CAPI interviews. Second, we found no evidence of interview-shortening effects of CAPI. However, interviewers reported the impression of shorter interviews. In addition, both respondent and interviewer attitudes toward CAPI were mostly neutral or positive. Third, CAPI was found to greatly reduce the incidence of routing (or ‘skip’) errors, albeit this effect was somewhat diminished when post-supervisor review data was examined. This suggests that CAPI does a better job than PAPI at correcting errors in the field, which, if exploited properly, holds the potential for reductions in both interview time and fieldwork costs. Fourth, although we found evidence of statistically significant changes in the distributions of certain variables, we also found that these effects were not global. This suggests that PAPI interviews may be run parallel to CAPI interviews during a transition period without having to worry about damaging the quality of the data. However, this implication is tentative, especially in the case of some variables (such as life satisfaction) where statistically significant effects were detected. These instances warrant closer examination in future studies. ◈ The effect of a transition to CAPI on KLIPSs inter-wave data integrity Following the successful test run in wave 10, KLIPS made the full transition to CAPI in wave 11. This development motivates a different research question - What are the effects of the transition to CAPI on the inter-wave consistency of the KLIPS data - especially sample attrition, overall success in fieldwork, item non-response and non-response errors, and propensities when responding to ‘sensitive’ questions? Some findings that were not discussed in the analysis of the wave 10 experiment data are unveiled in the inter-wave analysis. First, we identified positive effects of CAPI in relation to the overall fieldwork success, such as respondents attitudes and understanding of the questionnaire, higher incidences of self-completion and successful person-to-person interviews, and shorter interview times. Although small in magnitude, an increase in the number of strong refusals has led to protracted fieldwork and lower unit response rates. However, this is attributable to the long history (and thus, repeated waves) of KLIPS, leading to ‘panel fatigue’, rather than to the effect of CAPI. This ‘panel fatigue’ effect may have exhausted the potential for better response rates. Meanwhile, CAPI was found to substantially reduce errors in key variables throughout the fieldwork (routing errors), data inputting, and data processing stages. Thus, item non-response and non-response errors which would have otherwise been detected in the later cleaning stages were found to have been prevented in the fieldwork stage. However, errors only identifiable through cross-checking and longitudinal data cleaning were not found to be reduced through CAPI. Also, no evidence was found that CAPI improved response rates for questions that are socially sensitive or cumbersome to answer. One potentially worrisome finding was that following the transition to CAPI, there was a greater incidence of heaping in consumption and saving-related questions. These results suggest that, while the overall transition to CAPI went smoothly for KLIPS, a closer look must be taken into issues such as changes in response patterns in some variables and the effects of ‘panel fatigue’ as the study is repeatedly administered year after year. ◈ Anticipated effects of the KLI CAPI-EHC EHC (Event history calendar) is a survey tool used mostly to survey major life events such as education, vocational training, employment, marriage and residence. EHC increases readability to reduce errors in recalling time points at which past events occurred, and thus improves datareliability; but the design and data entry process of traditional EHC modules are complicated, and requires substantial time and money to develop and operate. These weaknesses are expected, however, to be significantly alleviated with the introduction of CAPI, as CAPI surveys character-istically require a lot of time and money in the early development process, but involve lower cost burdens in later years. Also, since CAPI data input and output occurs simultaneously with fieldwork, the input/output burden is that much lower. In particular, since error checks occur on a real-time basis during fieldwork in the CAPI-EHC, we are able to collect more consistent life event data. The KLI CAPI-EHC program was developed with a focus on ensuring fast operation speeds, easy links with pre-collected information or other CAPI programs, and an easy-to-configure EHC program that can be customized to the questionnaire and survey design. The KLI CAPI-EHC program allows the 7 basic variables required for calendar composition to be specified in the Blaise questionnaire file, and is designed so that it can be quickly utilized in surveys just by designating the program title, calendar format and design and external linked files within the configuration file. The KLI CAPI-EHC system is a very practical and useful program that has been developed on the basis of know-how accumulated by KLI panel researchers in the course of conducting numerous panel surveys, and which brings together the strengths of CAPI and EHC. This CAPI-EHC program has not, however, been used in the field so far, and is therefore limited in terms of verification of the program’s stability and effectiveness. ◈ General summary and motivation for further research The introduction of CAPI into a longitudinal study environment was found to strengthen internal consistency, improve functionality for interviewers, while reducing simple errors such as routing (ie, ‘skip’) errors. Meanwhile, in the case of response rates, interview length, operational costs, data cleaning and processing overhead, it is too early to attempt a definitive assessment. While this study is meaningful in that it is the first to analyze CAPI mode effects in Korean longitudinal studies, it cannot cover the myriad issues which arise in survey methodology. The authors hope this study will serve as a starting point for the future pursuit of systematic and long-term research on survey methodology in Korea.
공공누리 마크 한국노동연구원 연구 발간물 보호 저작물 '공공누리' 출처 표시, 상업적 이용금지, 변경금지 조건에 따라 이용하실 수 있습니다.
최효미 님의 다른 보고서 : 14 건
연구보고서 게시판
번호 서명 저자 출판일 원문
14 패널자료 품질개선 연구(XI) 장인성, 이지은, 김기민, 김우영, 이상혁, 이지희, 채민희, 최효미 2021.12.30 패널자료 품질개선 연구(XI) 바로보기
13 지역노동시장과 여성 고용에 대한 연구 고영우,최효미,조장희 2018.12.28 지역노동시장과 여성 고용에 대한 연구 바로보기
12 패널자료 품질개선 연구(Ⅶ) 김유빈,박민규,변종석,최효미,권혁진,김우영,성재민,이지은,신선옥,이혜정,김기홍 2017.12.29 패널자료 품질개선 연구(Ⅶ) 바로보기
11 Methodology for the Improvement of Panel Data Quality-CAPI(Computer-Assisted Personal Interview) Effects- 최효미, 이상호, 성재민, 김기민, 배기준 2013.03.14 Methodology for the Improvement of Panel Data Quality-CAPI(Computer-Assisted Personal Interview) Effects- 바로보기
10 패널자료 품질개선 연구(II) 윤윤규·홍민기·이상호·배기준·최효미·김기민·장숙랑·조성일·천희란 2012.12.28 패널자료 품질개선 연구(II) 바로보기
123