Proposal on Analysis of Causal Relationship between Fertility and Female Employment using Propensity Score Matching

Abstract

What is the first thing that comes to your mind when you read the headline: ’Increasing red meat intake linked with heightened risk of early death?’ It probably depends on what you were thinking on meat intake patterns and how you were feeling about the word ’death’ before you started reading this sentence. Some people may find it sounds causal, if they believe that eating red meat results in high health risks and the word ’death’ provokes a strong negative emotion. Some may be rather sceptical, as the meaning of the word ’linked’ does not necessarily imply causation. So what should the reader conclude based on this headline in ScienceDaily? A closer look suggests that the outlet describes a mere association between processed red meat intake and heightened risk of death. It signals that the study does not infer any causal inference. So it clearly demarcates a boundary between correlation and causation. ’Causality’ remains a source of confusion even for the academic world. To address the controversy of the topic scholars create a formal field of causal inference, which gains momentum since about 1970s. Causal modelling as its own area of statistical research introduces the formal definition of causal effects and assumptions necessary to identify causal effects from the data (’causal assumptions’). It quantifies the impact violation of assumptions may have on conclusions (so called ’sensitivity analysis’) and highlights how to find the right variables to control for (’causal diagrams approach’). This summer term I took the course ’Econometrics of Common Factor and Causality’ introduced by Dr. Jingjing Lyu. This essay summarizes what I have learned during the course. The first part describes main takeaways from the lecture, which mainly focused on causal inference from observational studies. The second part introduces the research proposal on analyzing a causal relationship between fertility and female employment using Propensity Score estimates.