2020總統大選民調資訊真實性與結果之探討

NHUIR > College of Science and Technology > Department of Information Management > Disserations and Theses > Item 987654321/28565

Please use this identifier to cite or link to this item: http://nhuir.nhu.edu.tw/handle/987654321/28565

Title:	2020總統大選民調資訊真實性與結果之探討
Other Titles:	Discussion on the Information Authenticity and Results of the 2020 Presidential Election Polls
Authors:	許全偉 XU, QUAN-WEI
Contributors:	資訊管理學系陸海文 LU, HAI-WEN
Keywords:	民調;選舉預測;權重 Public opinion survey;Election prediction;Weighting
Date:	2020
Issue Date:	2022-08-08 14:45:36 (UTC+8)
Abstract:	每當總統大選期間，各家媒體紛紛針對總統候選人進行支持度民調，加上網路世界的興起，民調往往左右了民眾的投票意向，增加了不確定性，為了準確地預測選舉結果，學者們在選舉研究工作中，以不同的理論建構不同的選舉預測模型。因此，本研究係在2020 總統大選期間，依據各媒體在大選期間所公佈民意調查資訊為主體，運用層級的方法與權重的概念，建構系統化的模型逐層分析，並客觀的分析候選人的民調支持度，探討預測的結果與實際投票結果誤差之因素。　　本研究提出之預測模型的建構程序如下：（一）建立 2020 年總統選舉的系統模型架構、（二）蒐集 2020 年大選期間的民調資訊以決定各政黨的民調集合、（三）決定2020 年預測模型中各階層權重、（四）整合 2020 年總統選舉預測模型資訊、（五）分析 2020 年總統選舉預測模型結果與 2020 年總統真實得票率之差異。　　依本研究模型預估結果與實際得票率做交叉分析，民進黨所推派總統候選人蔡英文之選舉預測模型結果換算成得票率與實際得票率相差為61.88%-57.13%=4.75%；國民黨所推派總統候選人韓國瑜之選舉預測模型結果換算成得票率與實際得票率相差為31.25%-38.61%=-7.36%，親民黨所推派總統候選人宋楚瑜之選舉預測模型結果換算成得票率與實際得票率相差為6.87%-4.26%=2.61%；其中國民黨所推派總統候選人韓國瑜之預測的結果與實際投票結果誤差極大，其可能原因有下列兩項：（一）蒐集民調資訊的時間點。（二）重大事件的衝擊及策略性投票行為。　　During every presidential election period, various media conduct opinion polls on voters' support of presidential candidates. In addition, due to the rise of Internet, polls often influence the public's voting intentions and increase uncertainty. In order to accurately predict election results, many scholars have successively engaged in election research work, constructing different election prediction models through various theoretical bases. Therefore, this study adopted the opinion poll information released by the various media during the 2020 presidential election period as the main body and applied the hierarchical method and weighting concept to construct a systematic model for layer-by-layer analysis, thereby objectively analyzing the opinion poll of the candidates and exploring the factors contributing to the errors in prediction results and actual voting results.　　The procedures for constructing the prediction model proposed in this study are as follows: (1) Establish a system model framework for the 2020 presidential election; (2) Collect opinion poll information during the 2020 election period to determine opinion polls collected by the political parties; (3) Determine the weights of the hierarchies in the 2020 prediction model; (4) Integrate the 2020 presidential election prediction model information; (5) Analyze the difference between the 2020 presidential election prediction model results and the 2020 actual presidential votes obtained. 　　Based on the model prediction results in the study and the actual votes obtained, a cross analysis was conducted. The election prediction model results of presidential candidate Tsai Ing-wen appointed by the Democratic Progressive Party were converted into the difference between the percentage of votes obtained and the percentage of actual voted obtained: 61.88%-57.13%=4.75%; the election prediction model results of presidential candidate Han Kuo-yu appointed by the Kuomingtang (KMT) were converted into the difference between the percentage of votes obtained and the percentage of actual voted obtained: 31.25%-38.61%=-7.36%; the election prediction model results of presidential candidate James Soong Chu-yu appointed by the People First Party were converted into the difference between the percentage of votes obtained and the percentage of actual voted obtained: 6.87%-4.26%=2.61%. In particular, the predicted results and the actual voting results of Han Kuo-yu appointed by the KMT showed a great disparity, possibly due to the following two reasons: (1) The time of opinion poll information collection; (2) The impacts of major events and strategic voting behaviors.
Appears in Collections:	[Department of Information Management] Disserations and Theses

Files in This Item:

File	Description	Size	Format
108NHU00396014-001.pdf		2351Kb	Adobe PDF	173	View/Open
index.html		0Kb	HTML	442	View/Open

Loading...