2003年11月

機率與非機率抽樣對於調查結果之影響——以2004年我國總統大選候選人之民意調查支持度為例
The Effect of Probability and Non-Probability Sampling on Survey Result--Using Support Rating of Candidates' Polls of Taiwanese Presidential Election in 2004 as on Example

研究論文 Article

初文卿
Wen-Ching Chu

61-111

一般電話訪問的抽樣設計方式有任意成人法與戶中選樣法兩種， 前者屬非機率性抽樣，後者則爲機率抽樣。一般硏究大多針對改善任意成人法的非機率性缺失，以調整樣本結構的缺點，但也有以實證調査的結果，比對任意成人法與戶中選樣法在樣本結構上的差異，不過大多忽略不同抽樣方法在實際執行檢驗項目上所可能產生的效應。本文即以2004年我國總統大選爲主題，同時進行任意成人法與戶中選樣法，以探査不同抽樣在單一檢驗項目——支持度的效應。本硏究在控制各種變項後，發現兩項調査呈現迥異的結果，亦即兩種調査不但有支持度的差異，甚至在領先趨勢上也呈現相逆的情形。在比對交叉分析後，發現在18個層別中，有7個層別呈現趨勢相逆的現象，而且也發現任意成人法對現任者陳呂配支持度有膨脹的效應，而戶中選樣法對現任者陳呂配則有收縮的效應，顯示在轉換成支持度後，呈現對某組候選人較爲有利或不利的調査結果。因此不同的抽樣仍有超越其本身差異的效應存在，値得調査機構在運用與解讀分析上，做爲審慎參考的依據。

As to the sampling scheme of telephone interview, the popular methods are noselecting and within-household respondent selecting. The former belongs to a kind of the non-probability sampling, the later the probability sampling. Most studies of related fields modify the shortage of non-probability sampling in noselecting method and adapt its' sample frame. However, these studies neglect the inconsistency caused by different sampling methods. By checking polls of Taiwanese presidential election in 2004, this paper contrasts the survey results of noselecting and within-household respondent selecting, focusing on candidates' support ratings and leading tendency. By analyzing the interrelation, this study reveals inconsistency in 7 out of 18 strata. It indicates that noselecting method causes an effect of inflation in "current position" bias toward Chen-Lu pair, while within-household respondent selecting has an opposite effect. This research concludes different sampling methods do have impact on survey results.

支持度, 任意成人法, 戶中選樣法, 機率抽樣, 非機率抽樣, 樣本結構
support rating, noselecting, within-household respondent selecting, probability sampling, non-probability sampling, sample frame

政治學
Political Science

https://doi.org/10.7014/SR.2003110003

全文下載