Modele regresyjne w badaniu zachowań wyborczych
- Year of publication: 2012
- Source: Show
- Pages: 114-142
- DOI Address: https://doi.org/10.15804/athena.2012.36.06
- PDF: apsp/36/apsp3606.pdf
Regression Models in Research on Voting Behavior
REGRESSION ANALYSIS IS a very useful research method in electoral studies. In this paper I examine two types of regression models that are of great importance in studying voting behavior and electoral systems, that is, logistic regression models and models for panel data. Both logistic regression and panel data models allow political scientists to deal with problems, where classical linear regression models, estimated using ordinary least squares (OLS), fail. In the first part of the paper I provide concise theoretical explanations of the panel data and logistic regression models and also the motivation to make use of them in research. If we examine panel data and we use classical OLS method of estimation, we probably violate some important assumptions of classical linear regression model. Fortunately, we can use special methods of estimations for panel data models that allow us to relax some classical linear regression model assumptions. One of them is least squares dummy variables (LSDV) method. As for logistic regression, this method is also very useful in electoral studies. We use logistic regression model if we examine binary dependent variables that often appear in social sciences. In the second part of the paper I apply the methods mentioned above to study voting turnout in Polish 2011 elections and to examine the probability of voters’ overrepresentation in Japanese constituencies.