人力資源調查方法改進芻議

人力資源調查前身勞動力調查自民國五十二年起由臺灣省勞動力調查統計研究發展小組開始按季辦理，後還經省勞動力調查研究所等單位接辦，至民國六十七年正式移由主計處辦理，迄今已逾三十七年，對於建立全國勞動市場基本情勢資料功不可沒。其間在調查方法及內容方面亦有所變革，惟隨社會及經濟情況變遷，其中容有若干值得檢討改進之處，俾能充分反映現今勞動市場瞬息萬變之特性。本文主要從人力資源調查之抽樣方法、問項內容等方面深入探討，發掘問題，並以實際資料驗證，從而參酌其他圈家作法，並提出「當母體更新時，第一段樣本單位應訂定換層準則，不能僅因某一項特徵值超越分層點時，即更換層別，以減少抽樣變異」、「副母體樣本配置宜改採均等配置，再配合母體特徵變異及地區特殊考量調整之，以兼顧地區別資料之穩定性」、「第一段樣本村里宜改為PPS方式抽取，不但樣本村里調查戶數可依機率模型控制在調查員所能負荷之水準，同時亦可獲得不偏且變異較小之估計量」、「強化非勞動力中之失望勞動者及就業者中工時不足之不完全就業者資料蒐集，俾利隱藏性失業月指標之建立，以使勞動市場變化充分反映景氣之榮枯。」、「人力資源調查增加詢問兼職者之兼職數，藉以消弭與受僱員工薪資調查間就業人數與工作職位數間之差異，以為日後調查比較改進之基礎」等結論。

關鍵字

人力資源調查；樣本輪換；勞動力；隱藏性失業

並列摘要

The Labor Resource Survey have been conduct for 37 years by different agencies, Labor Force Survey, Statistics Research and Development Group begun to conduct this survey in 1963, several agencies succeeded it till DGBAS took over the survey in 1978. The result of survey has been the primary source of information on labor market ever since. Although the survey have ever been improved in sampling techniques over last two decade, but the Nation's society and economy underwent many changes and current sampling method still existed some problems, so survey design need to be reviewed again and make the data derived from the survey become more sensitive to the labor market vanes.This article discusses current sampling technologies and the contents of survey, several issues of survey design were proved to affect the quality of the data. The article also introduces the solution of those issues base on basic theories and what many countries have done before. The conclusions we have as follows:．A change in single characteristic should not justify removing a PSU from a stratum to another, a change in several population characteristics is required before a PSU is removed instead, some rules for the restratification should be established to improve reliability of estimates.．Assigning sample sizes to the sample of sub-population with equal allocation will extremely contribute to decrease of variances for those small sizes of sub-populations. The number of allocated sample also can be readjusted depending on homogeneity of characteristics within each sub-population.．Selection of sample PSU is referred to as sampling with probability proportional to size, the number of sample households in PSU can be limited within workload of an interviewer, the estimates will be unbiased and estimated variance will be smaller than that from sampling with equal probabilities.．Improving monthly data collection of discouraged worker outside the labor force and under-employment in the labor force will help establishment of index for hidden unemployment.．Employed persons are asked whether they have more than one job and how many jobs they have, it may help to reconcile employment estimates from the Labor Resource Survey with those from the Employees' Earnings Survey.