During my doctoral studies, I developed (together with Klaus Gründler
, ifo Munich) a new democracy index which is based on a machine learning technique for pattern recognition. The current version of my index is available for 186 countries and covers the period from 1919 to 2016. [Download
Following Robert Alan Dahl (1971), I define Political Participation
as the right of citizens to elect their political leaders and representatives. To operationalize this aspect of democracy, I use different measures of voter turnout.
Following Adam Przeworski (1991), I define the political process as competitive if individuals with different party affiliations compete in public elections for political support. To operationalize this defintion of Political Competition,
I use various election outcomes and information from the Varieties of Democracy Database
Following the Universal Declaration of Human Rights
(1948), I define that citizens enjoy the Freedom of Opinion if they can freely choose their sources of information and can express their political views even when these views are not compatible with the political views of the government. I operationalize this aspect of democracy through two expert-based indices from the Varieties of Democracy Database
A key challenge in producing a democracy index is to find an aggregation rule that transforms the observable regime characteristics in an index. The standard procedure is to weight the regime characteristics and then to apply a multiplicative or additive aggregation function (Gary Goertz, 2006). My research suggests that the standard approach creates non-randomly biased democracy indices (for details, see here
). As an alternative, I propose an aggregation method that is based on a machine learning tool for pattern recognition, known as Support Vector Machines (for details, see here
). A major advantage of my approach is that it puts the aggregation problem into a non-linear optimization problem and thus avoids simple assumptions about the functional relationship between the regime characteristics and the degree of democratization.