鑑於主計處「台灣地區人力運用調查」資料使用的普遍性與重要性，本文針對其中「工作與否」選項是否存在調查偏誤，應用Hsiao and Sun (1999)的單邊反應模型與隨機反應模型，偵測偏誤來源並估計偏誤大小。實證結果發現，社會可欲是造成1998年［工作與否」資料中「不在做工作」虛報「在做工作」者的主因，偏誤大小約有5%的機率。此外，以與解釋「工作與否」大部分相同的變數來解釋社會可欲調查偏誤時，會發生認定上的問題，使估計結果不理想。因此，本文探討回答問卷的兩種方式對調查偏誤的影響。結果顯示，問卷非本人回答的社會可欲較高；忽略社會可欲調查偏誤將導致低估年齡、教育程度以及「結婚」對就業的影響，也可能高估高齡人口與結婚女性的就業情形。

This paper applies the one-sided response and the randomized response models of Hsiao and Sun (1999) to investigate the response bias in the question ”whether at work” in the Taiwan Manpower Utilization Survey. Approximately 5% of ”yes” responses to the ”whether at work” question in the 1998 survey were found to be attributable to the bias of social desirability. Ignoring this bias leads to an underestimation of the influence of age, education and marital status on employment. Specifically, the employment levels of the aged population and married women were overrated. The findings suggest that when surveys invite socially sensitive answers, the class of estimation models that incorporates response biases should be used.