The Working Paper gives an overview about the topic of causal inference, covered in the Institute on Statistical Analysis for Education Policy organized by the American Educational Research Association in spring 2013. Because randomized experiments are often difficult to implement into large-scale studies in educational research, inference on the causality of different treatments (e.g. different teaching approaches) is limited. This paper discusses the possibilities to draw causal inferences in non-randomized experiments, and provides an introduction to different analytical approaches, namely propensity score analysis, sensitivity analysis and the regression discontinuity approach. The main idea behind each procedure is explained, advantages and limitations are discussed.