2005年10月

配對橫剖面資料的樣本代表性及估計有效性：以1981－2000年人力運用調查為例
Representability and Estimation Effectiveness of Longitudinally Linked Cross-sections: An Assessment Based on the 1981-2000 Manpower Utilization Surveys

研究論文 Article

林季平
Ji-Ping Lin

73-111

以1981-2000年QLMUS、NOMUS、及MUS資料為基礎，本研究主要目在（1）探討配對導致的樣本流失對樣本選擇的影響和影響配對樣本流失的主要因子，及（2）以某些統計量為例，評估QLMUS相對於NOMUS及MUS的代表性及其估計之有效性。主要發現如下：（1）配對樣本較少選擇年輕族群及高教育程度者；（2）樣本配對比率呈現顯著的空間差異，都市化程度對樣本配對比率呈現負向效果，且都會區的核心地區的配對率明顯的低於其周邊的地區，該現象之直接或間接成因主要係遷徙所致；（3）由1981-90年QLMUS、NOMUS、及MUS估出的年齡別勞動參與率沒有顯著差異，但由1991-2000年估出的結果卻存在顯著差異；（4）由各年的QLMUS、NOMUS、及MUS估出的年齡別失業率皆有顯著差異，但由各年的QLMUS、NOMUS、及MUS估出的年齡別工作異動率及平均工資水準沒有顯著差異。

Based on the QLMUS as well as the NOMUS and MUS in 1981-2000, the research aims (1) to characterize QLMUS in terms of sample selection and attrition due to match and (2) to assess the representability and estimation effectiveness of QLMUS in comparison to MUS and NOMUS. Main findings: First, attrition due to match does have a salient effect on the selectivity of QLMUS in which the age of samples is associated with a concave effect and education with a negative effect on match rate. Second, the match rate of QLMUS is seen to exhibit distinct spatial disparities, with the highly urbanized areas having lower match rate. This phenomenon is highly likely due to geographic migration. Third, on the basis of labor force participation rate (LFPR), unemployment rate (UR), turnover rate (TR), and wage, LFPR estimated from the MUQLS, NOMUS, and MUS basically are compatible at all for the period of 1981-90, the estimates of UR from the three data sets are mostly not compatible, while the estimates of TR and wage from the MUQLS, NOMUS, and MUS are compatible at all.

樣本流失, 樣本選擇, 人力運用調查, 橫剖面資料, 擬－追蹤調查資料
Sample Attrition, Sample Selection, Manpower Utilization Survey, Cross-sectional Data, Quasi-longitudinal Data

經濟學, 社會學
Economics, Sociology

https://doi.org/10.7014/SRMA.2005100003

全文下載