Name four identification strategies for causal effects
Randomized Controlled Trails(RCT)
Controll variable approach (OLS and Matching)
Instrumental variables (IV)
Difference in Differences (DiD)
Identification assumption
are untestable
avoid overly strong once
clear stating the assumption
provid argumentation for why they hold
Based on what do we decide which identification strategie we are using
question we want to answer
data avilable
assigment rules
Causal language vs. Non-causal language
Causal language
indicates that one variable effects the other with direction
example: increase/reduce, promts, improves, impact, consequence, because/since/due
Non-causal
correlation and no direction
example: is associalted with, higher/lower than, get/have/linked
Disadvantages of RCT
non random participation
non random drop out
disapoitment if no placebo
interal v. external validity
often unfeasible due to ethical/financial/political issues
Advantages of Regression
easy to implement
intensively discussed
Advantages of PMS/IPW
no functional form assmution
focus on threatment effect
often clear statement of identification assumption
PSM does not solve the problem of ommitted variables
Zuletzt geändertvor 7 Monaten