人力運用擬：追蹤調查資料庫的產過程、應用現況、及未來發展

Abstract

為了克服研究資料之限制，本文作者以1979-2001年的「人力運用調查」原始資料為基礎，產生所謂的1979-2001年「人力運用擬-追蹤調查」（MUQLS）資料庫及1979-2001年「無重疊樣本的人力運用調查合併資料」（NOMUPD）。MUQLS資料庫及NOMUPD資料庫目前皆以SAS v.8.1的格式儲存，其中MUQLS資料庫具追蹤調查特質，計包括53萬餘筆樣本，而NOMUPD資料庫是另類的人力運用調查，特點是己將重復樣本自前後期的「人力運用調查」移除，但本質上還是橫剖面的資料，計包括了79萬餘筆樣本。我們建立這些資料庫的主要學術貢獻是：相較於一般追蹤調查，MUQLS資料庫占有極大的成本優勢，並能幫助我們克服許多以前難以進行的研究，而NOMUPD資料庫則具有減少樣本選擇偏誤的特質。本文除了談MUQLS資料庫及NOMUPD資料庫的產生緣起、對學術研究重要性及潛在貢獻、資料庫的產生過程和資料庫特質之外，亦以一個簡單實例，說明MUQLS資料庫及NOMUPD資料庫對克服研究資料限制的潛在幫助。最後，我們提出對未來進一步發展的看法及建議，主要包括（1）建立定期擴充MUQLS資料庫及NOMUPD資料庫的標準化機制、（2）以開發MUQLS資料庫的技術為基礎，建立「廣義人力資源擬-追蹤調查」（GQLHRS）資料庫、（3）建立友善的網路界面藉以加快資料傳播速度及減少資料運用阻礙、及（4）開放原始程式碼及建立使用者論壇和回饋機制。

In order to overcome data constraint of research, the authors construct two huge databases, Manpower Utilization Quasi-longitudinal Survey (MUQLS) as well as Non-overlapped Manpower Utilization Pooled Data (NOMUPD), based on the 1979-2001 Manpower Utilization Surveys (MUS) of Taiwan. The construction of MUQLS and NOMUPD is mainly implemented through SAS v8.1. MUQLS is essentially longitudinal, including more than 530 thousand individuals, while NOMUPD remains cross-sectional in nature, including around 790 thousand individuals. Out major contributions in constructing these databases are that MUQLS is not only fairly cost-efficient relative to ordinary longitudinal surveys, but also helps us overcome many research constraints; moreover, NOMUPD helps us reduce the conventional danger of selection bias when using the pooled MUS data. In addition to sharing our experiences in constructing MUQLS and NOMUPD, this article also uses a simple example to demonstrate the potential importance of MUQLS and NOMIUPD in academic researches. In the end, some suggestions are proposed for future developments, including (1) establishment of standardized procedures in extending the MUQLS and NOMUPD, (2) construction of Generalized Quasi-longitudinal Human Resources Survey (GQLHRS) based on the existing techniques in constructing MUQLS, (3) setup of user-friendly interface on the internet to help the sharing of data, and (4) opening-up source codes to public access and establishment of user forum and feedback mechanism.