Aristotle University Of Thessaloniki (AUTH)
Μέθοδοι πρόβλεψης εκλογικών αποτελεσμάτων: συγκριτική μελέτη εργαλείων του διαδικτύου
Abstract
dc:descriptionThe subject of this thesis, titled "Methods of Forecasting Election Results. Comparative Study of Internet Tools", is the study of the existing methods of analysis used to predict election results. The work aims at a comparative study of the methods used in electoral research for analyzing the behavior of voters, namely the forecasting of the electoral attitude. The study focuses on three electoral forecasting methods that are used in Greece and extends to the analysis of an innovative method deriving from the data analysis of Voting Advice Applications (helpmevote): 1) exit polls, 2) polls on voting intention by weighting of data, 3) time series analysis with respect to the voting intention for estimating the electoral influence of parties. Apart from the description of the main techniques, the effectiveness of these methods is analyzed in terms of proximity of their forecasts to the actual election results. Simultaneously, issues that create problems on the validity of the methods are explored, such as the weighting of data, addressing unspecified vote and many technical issues concerning the application of these methods, particularly in the current Greek electoral environment, where over the last years structural changes are observed in the party system, as well as in the electoral behavior of voters and the attitude of the respondents in opinion polls, making the currently used voting forecasting methodologies problematic. Additionally, the creation of a more efficient predictive model is attempted, using internet tools and particularly the Voting Advice Application. In the framework of the research data from the application helpmevote of the 2010 regional elections were used. In the first part of the analysis, three main questions are explored: 1) identifying groups of voters who have similar views with candidates in their region and the correlation of these groups with specific topics and electoral preferences, 2) the prediction of voting choice of the undecided or those who refuse to specify their voting intention, based on their degree of correspondence with candidates, parties and issues, 3) the percentage to which the factor of issues is involved in the final voting choice and the creation of a predictive model based on these. In the second part of the analysis the construction of a linear model is attempted in order to predict the voting intentions, using the placement of voters in relation to issues and the degree of their correspondence with the candidates, as main variables. As independent variables of the function are selected the stated voting intention and in search of an alternative model the predicted voting intention as calculated from the first part of the analysis. The methods of analysis proposed and applied are Discriminant Analysis, Fuzzy Clustering, Stacked Data Analysis and Binary Logistic Regression.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2014
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Panagiotidou, Georgia
- Παναγιωτίδου, Γεωργία
Subjects
dc:subject × 40- Μοντέλα πρόβλεψη εκλογικού αποτελέσματος
- Μέθοδοι εκλογικής ανάλυσης
- Ανάλυση εκλογικής συμπεριφοράς
- Θεματική ψήφος
- Δημοσκοπήσεις εξόδου
- Δημοσκοπήσεις πρόθεσης ψήφου
- Εκτίμηση εκλογικής επιρροής κόμματος
- Στάθμιση δεδομένων
- Προηγούμενη ψήφος
- Κομματική ταύτιση
- Ελληνικές περιφερειακές εκλογές 2010
- Ηλεκτρονικός σύμβουλος ψήφου
- Αδιευκρίνιστη ψήφος
- Ανάλυση κατά στοιβάδες
- Forecasting models of Election Results
- Methods of electoral analysis
- Analysis of electoral behavior
- Issue voting
- Exit polls
- Opinion polls on voting intention
- Estimation of electoral party influence
- Data weighting
- Past vote
- Party identification
- Greek regional elections 2010
- Voting Advice Application VAA
- helpmevote
- Unspecified vote
- Discriminant analysis
- Fuzzy clustering
- Stacked Data Analysis
- Binary Logistic Regression
- Φυσικές Επιστήμες
- Μαθηματικά
- Κοινωνικές Επιστήμες
- Πολιτικές Επιστήμες
- Natural Sciences
- Mathematics
- Social Sciences
- Political Science
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/35433