We invite you to the Department of Political Economy meeting, which will take place on 7.11.2023 at 17:00 in room A409.

Przemysław Kurek will present his research on "How to apply Causal Inference in observational studies? Conditional Instrumental Variable analysis based on the Causal Diagram approach: The impact of Foreign Language Skills on Wages" (abstract below). This part of the meeting will be held in English.

In the second part of the meeting we will discuss current organizational issues of the Department (in Polish).

Abstract

While the impact of Foreign Language Skills on wages is the phenomenon investigated by many researchers, strict causal evidence in this matter can not be present, as this setting is purely observational. Existing studies base mostly on cross-section correlational modelling and are subject to many sources of bias: Attenuation Bias, Bicausality Bias, Confounding Bias and Selection Bias. In this study we present how to improve the quality of the current approach using Causal Diagram methodology. First, we point out that simple correlational models with control variables are addressing only one of those sources of endogeneity leaving in most cases other sources intact and possibly result in the systematically biased estimations. Second, by the inclusion of control variables, the possibility of including Bias-Inducing variables also arises. In order to better address these problems we suggest using a Conditional Instrumental Variable (IV) approach based on Causal Diagram methodology taking mother and father education as conditional instruments. These potential instruments were not used in this setting before, as the classical IV assumptions were not satisfied. However, in the Causal Diagram methodology the IV assumptions are formulated differently, and therefore allows for much more convenient variable selection in the Conditional IV models. This opens the possibility to explore new instruments, which still have the downside of being control-dependent, but have the potential to bring additional information as they address all sources of endogeneity.