이 연구에서는 한국청소년패널조사 중2패널 1∼6차년도, 초4패널 1∼5차년도 데이터에서의 종단적 최초표본이탈에 영향을 미치는 표본특성요인에 대해서 경험적으로 분석해보고자 하였다. 이를 위해 선행연구분석을 통해 성별, 학업성적, 부(남자보호자)학력, 모(여자보호자)학력, 부모(보호자)취업여부, 한부모가구여부, 가구원수, 월평균가구소득, 자가거주여부, 거주지규모의 독립변수가 청소년최초표본 이탈, 부모(보호자)최초표본이탈, 청소년부모(보호자)동시최초표본이탈의 종속변수에 미치는 영향을 연구모형으로 설정한 후, 생존분석기법(survival analysis) 중 콕스의 비례위험 회귀분석 모형(cox proportional hazard model)을 적용해 경험적으로 분석하였다. 분석결과, 중2패널에서는 여성에 비해 남성일수록, 학업성적이 낮을수록, 모(여자보호자)학력이 높을수록, 월평균가구소득(log)이 높을수록 최초표본이탈 위험이 높은 것으로 나타났고, 초4패널에서는 모(여자보호자)학력이 높을수록, 양부모가구에 비해 한부모가구일수록, 자가거주에 비해 자가거주가 아닐수록, 거주지가 중소도시에 비해 대도시일수록 최초표본이탈위험이 높은 것으로 나타났다. 중2패널과 초4패널의 최초표본이탈위험에 공통적으로 영향을 미치는 것으로 나타난 변수는 모(여자보호자)학력 변수뿐이었고, 중2패널에는 남성더미(여성기준), 학업성적, 월평균가구소득(log) 변수가, 초4패널에는 한부모가구더미(양부모가구기준), 자가거주아님더미(자가거주기준), 거주지대도시더미(중소도시기준) 변수가 각각 따로 최초표본이탈위험에 영향을 미치는 것으로 나타났다. 이러한 분석결과는 향후 아동 및 청소년 대상의 패널데이터를 구축할 때 표본이탈을 최소화하기 위한 표본관리방법 강화에 도움을 줄 수 있을 것이다.
In this study, an attempt was made to empirically analyze the sample characteristics that affect the longitudinal initial sample attrition in data taken from the Korea Youth Panel Survey. The data was taken from the 1st through to the 6th annual datasets of the second grade junior high school youth panel data and the 1st through the 5th annual datasets of the fourth grade elementary school youth panel data. The independent variables consisted of gender, academic performance, father(male guardian)’s level of education, mother(female guardian)’s level of education, parental(guardian’s) employment status, single parent households, number of household members, monthly average household income, whether the primary property is owner occupied or rented, and size of the primary property. The dependent variables consisted of the adolescent’s longitudinal initial sample attrition, the parent(guardian)’s longitudinal initial sample attrition, and the adolescent- parent(guardian)’s concurrent longitudinal initial sample attrition. The cox proportional hazard model was applied for the purposes of analysis. The results of this study were as follows: The higher the proportion of males compared to females, the lower the level of academic performance. Similarly, the higher the level of the mother(female guardian)"s level of education. Finally it was found that the higher the monthly average household income, the higher the risk of longitudinal initial sample attrition was in the analysis of the 1st through to the 6th annual datasets of the second grade junior high school youth panel data. The following relationships led to a higher risk of longitudinal initial sample attrition in the analysis of 1st through the 5th annual datasets of the fourth grade elementary school youth panel. The higher the mother(female guardian)"s level of education, the higher the single parent households compared to the two parent households, the higher the proportion of primary properties being rented rather than owner occupied, and the higher proportion of metropolitan city residence compared to small-medium city residence –all led to higher risks of attrition. The only variable that had a common effect on the risk of longitudinal initial sample attrition of the second grade junior high school youth panel and fourth grade elementary school youth panel was the variable of the mother(female guardian)’s level of education. The variables such as the proportion of males compared to females, the level of academic performance, and the monthly average household income affected only the risk of longitudinal initial sample attrition of the second grade junior high school youth panel. The variables such as single parent households compared to more conventional two parent households, ownership status of the primary property and the proportion of metropolitan city residence compared to small-medium city residence affected only the risk of longitudinal initial sample attrition in the fourth grade elementary school youth. panel. The results of this analysis should help to strengthen sample management methods to minimize sample attrition when creating panel data for children and adolescents in the future.