Bjørn Høyland

Bjørn Høyland

Professor of Political Science

Teaching

STV4027 - Causal inference and prediction

Graduate course, University of Oslo, Department of Political Science, 2017

This course introduces students to recent developments in the scholarly effort to derive causal explanations using quantitative methods. The bulk of the course will be concerned with how to identify and estimate causal effects in observational studies. It can be argued that this effort amounts to a paradigm shift within quantitative social science away from regression models and explained variance to identification and measurement of causal effects. Taking the randomized experiment as the ideal, we clarify the challenges faced by social scientists seeking to draw causal inferences from observational data. Units in observational studies usually select into their causal status (their ”treatment” status) through processes outside of the control of the researcher rather than being assigned to these causal states by the researcher, such as in controlled experiments. The characteristics of this selection process are central throughout the course. We present a range of approaches for identifying its’ core features and for drawing valid causal inferences given those features. In the process, we highlight the limitations and difficulties associated with causal estimates obtained via the different techniques. More than anything, the course aims to develop a critical, yet constructive, mindset towards claims of causal effects in observational studies The course will also give a brief introduction to basic techniques and concepts used for prediction purposes, and discuss how prediction differs from, and relates to, causal explanation. Together, causal inference and prediction constitute the two main activities of contemporary quantitative social science, and students of this course will be familiarized with the challenges and promises of both of them. Former title (before spring 2017): STV4027 - Causal Inference.

STV4358 - Comparative Political Institutions

Graduate course, University of Oslo, Department of Political Science, 2016

In this course, we compare how political institutions shape politics and policy-outcomes. In the first part of the course, we review spatial models before investigating approaches for identifying spatial positions of political actors and policies. In second part of the course, we apply these models and measurements to analyze government formation and duration; executive -legislative relations; and legislative coalition formation and organization. Empirically, the course will draw on insights from a range of political systems, including, but not limited to, the EU, US and national European nation-states. Students are expected to familiarize themselves with existing datasets and be able to rely on these for the assignments.

STV4025 - Quantitative political science

Graduate course, University of Oslo, Department of Political Science, 2013

This is a course in quantitative political science. The aim is to provide a rigorous foundation for students interested in applying quantitative methods in their own research. The rst part of the course provides a thorough treatment of generalized linear models. We will cover both standard models and their hierarchical extentions. The main text for this part is Long, S. J. (1997). Regression Models for Categorical and Limited Dependent Variables. SAGE, London. The second section of the course covers event-history models. Here, the main text will be Box-Ste ensmeier, J. M. and Jones, B. S. (2004). Event History Modeling: A Guide to Social Scientists. Cambridge University Press. We will use R, a free statistical programming language. It is available at www.r-project.org. Here, you will also nd a lot of useful information about R. I also recommend you to use Rstudio, www.rstudio.org for running R on your computer. I use LATEX to typeset my papers and presentations. It works great in combination with R and Rstudio. If you are serious about learning R, study Matlo , N. (2011). The Art of R Programming: A Tour of Statistical Design. No Starch Press: San Fransisco. An introduction to LATEX can be found here: http://ctan.uib.no/info/lshort/english/lshort.pdf. The course aims to teach students how to do reproducable research. In order to integrate LATEX and R, we will use knitr, http://yihui.name/knitr/. Each week will have two lectures and one computer class. One lecture will be devoted to the statistical properties of the model, assumptions and generalization. The second lecture will focus on estimation and interpretation of the estimates obtained from the model. The purpose of the computer classes is to develop programming tools to become a productive quantiative political scientist.