台灣勞工失業期間之研究至今尚無定論，究竟勞工個人特性（如年齡、教育、性別等）與其失業期間有無關聯，也是眾說紛紜。本文認為現有文獻之實證結論與理論或常理違背的可能主要原因就是資料挑選與保留工資的遺漏。本文串聯主計處「人力資源暨運用調查」五月及六月之資料，再以兩階段法處理保留工資變數，發現個人特性在失業期間的實證上都非常顯著。在刻意遺漏保留工資的實證上，發現所有的個人特性全都不顯著，說明保留工資的遺漏所造成估計上的偏誤問題。此外，本文亦估計含有不可觀察異質變異數之失業期間模型，個人特性的顯著性與符號未發生任何改變。

Whether workers' characteristics influence their unemployment duration is inconclusive in the existing literature in Taiwan. This study traces those persons unemployed in May and resampled in June in the data of the Manpower Utilization Survey conducted by the Directorate-General, Budget, Accounting, and Statistics (DGBAS). It is found that conclusions of insignificance and opposite signs of workers' characteristics in the existing literature are likely due to data used and omission of reservation wage. It is shown that with reservation wage, all explanatory variables of workers' characteristics are significant. On the contrary, without reservation wage, all workers' characteristics are insignificant. It implies a severe omitted variable problem. Results of estimates under unobserved heteroscedasticity are consistent with the model without unobserved heteroscedasticity.