有關選舉民意調查的數據真實性或樣本代表性所引發之爭論核心在於：選舉民調的「母體」(population)定義為何？由於選舉的投票率不是百分之百，導致「選民母體」和「投票母體」不完全一致，因此利用不同抽樣底冊進行選舉民調，其樣本結構與選民母體、投票母體的一致性，便成為調查方法上值得探討的議題。本文的研究目的有三：(1)透過回顧我國選舉民調所採用之不同抽樣底冊的演進過程，探討其中所隱含的母體涵蓋率缺失；(2)釐清選舉研究中的選民母體與投票母體的內涵，並以2016年總統選舉為例進行比較分析；(3)分析TEDS針對2016年總統選舉利用不同抽樣底冊進行民調，成功樣本與選舉母體、投票母體之間的落差及加權處理的修正效果，進而探討其在理論上或實務上的意涵。本研究的分析結果發現2016年總統選舉的選民母體與投票母體的差異不大。不過，利用不同抽樣底冊進行民意調查的樣本結構與這兩種母體並不完全一致，且無論是利用何種母體結構進行事後加權調整，對投票率、得票率和催票率的修正效果有限。因此，要解決民調結果與投票結果之間的落差，有關母體的定義或許並非關鍵，對抽樣底冊選擇、問卷題目設計、調查技術精進、訪問失敗減少等環節的檢討，可能才是治本之道。

The controversy over the authenticity of election polls or the representativeness of samples is the definition of “population” in election polls. Since voter turnout is not absolute, there is a difference between “eligible population” and “voter population”. Therefore, the demographic consistency between eligible population and voter population with different sample frames in election polls is a topic that has become crucial in survey methodology. The purpose of this study is threefold: (1) exploring the error of population coverage through reviewing the evolution of different sampling frames used in election polls in Taiwan; (2) clarifying the implication of eligible population and voter population in election studies and analyzing the 2016 presidential election campaign in Taiwan; (3) analyzing the difference of the polls conducted by TEDS for the 2016 presidential election by using different sample frames, and examining the effect of the weighting adjustment by population parameters, and exploring its theoretical and practical implications. This study found that there was a minor difference between the eligible population and the voter population in the 2016 presidential election in Taiwan. However, the polls from different sampling frames were not completely consistent with the polls from these two populations. The correction effect of the weighting adjustment by the different population parameter on voter turnout, vote gain, and encouraging voter participation is limited. Therefore, to resolve the gap between the poll results and the voting results, the definition of the population might not be critical. The selection of sample frames, the design of questionnaires, the improvement of interview skills, and the reduction of interview failures are more important factors.