This thesis is rooted in the research tradition known as coalition politics, where governments, political parties and political institutions are the central focus. The main emphasis here is on government instability and the question of why governments in modern parliamentary democracies often come to an end before the next regular election. In five distinct but interrelated papers, the thesis explores the issue of early government termination and how it is affected by public support, economic developments and the functioning of the state apparatus. The studies included in this thesis generally take a quantitative approach and make use of a dataset that contains 640 governments in 29 European democracies. Their joint goal is to improve our understanding of when early termination happens by introducing and testing new explanatory factors as well as by improving how previously identified factors are modelled. The first paper focuses on Central and Eastern Europe. It shows that the stability of governments in that region is affected by slightly different factors than those that impact on governments in Western Europe. In particular, ideological factors and political institutions are found to be less important in Central and Eastern Europe while the formal power basis of the government and the country’s economic performance matter more. In the second paper, co-authored with Professor Torbjörn Bergman, the state is brought into government stability research. The paper shows that countries with a lower quality of governance and a less efficient public sector have less stable governments. This is mainly because government parties struggle to achieve their policy goals when the state apparatus is inefficient and corrupt. Paper 3, co-written with Associate Professor Johan Hellström, looks at how different types of governments respond to economic challenges. In particular, this paper demonstrates that the same changes in economic circumstances (e.g. increases in unemployment or inflation) have different effects on cabinet stability depending on which type of government is in charge. Single party governments are better equipped to deal with economic changes, because they are better positioned to devise new policy responses without having to compromise with other parties. Coalition governments, in contrast, become significantly more likely to terminate early when the economy takes a turn for the worse. Finally, over the course of two papers I first explore new techniques for analysing polling data and then use them to empirically test whether governments sometimes choose termination as a way to cope with bad poll numbers. Most of the existing techniques for pooling polls and forecasting elections were explicitly designed with two party systems in mind. In Paper 4, I test some of these techniques to determine their usefulness in complex, multiparty systems, and I develop some improvements that enable us to take advantage of more of the information in the data. In the final paper, I combine the two themes of polling and government stability by looking at how changes in government popularity affect the likelihood of premature dissolution. I find that governments, particularly single party governments, do, in fact, use terminations as a strategic response to changes in their popularity among the public. When support is high, governments tend to opportunistically call an early election, whereas they tend to abandon or reshuffle the government when support is low.