Week 8 : Natural Experiments

Question 1

What is the key difference between natural experiments and real experiments?

Answer 1

In Natural experiments, the experimenter has no control over assignment.

Question 2

What is the idea of “as-if random”?

Answer 2

It all hinges on the assumption that the treatment happens as if randomly assigned.

Question 3

What are the advantages of as-if random experiments?

Answer 3

• High external validity
• No Hawthorne effect

Question 4

What is standard natural experiment?

Answer 4

Treatment is assigned as-if randomly through
some process.

Question 5

List 3 types of natural experiments.

Answer 5

• Standard NE
• Regression Discontinuity designs
• Instrumental Variables

Question 6

What is regression discontinuity design?

Answer 6

A non-randomised study design that permits strong causal inference with relatively weak assumptions.

Question 7

How does regression discontinuity design work?

Answer 7

To assess the effect of a programme / policy / event, we compare those above (treatment) and below (control) the cutoff within the specified bandwidth, and assume that they are nearly identical. Then calculate the difference between treatment and control.

Question 8

Give 3 examples of a running variable in regression discontinuity design.

Answer 8

• Temporal distance to event / policy
• Spatial distance to border
• Distance to threshold, (e.g. election, competition, eligibility)