Assumptions of experiments

Question 1

Excludability

Answer 1

The assumption that z i can be omitted from the schedule of potential outcomes for Y_i (1) and Y_i (0).

Potential outcomes respond only to d and z is irrelevant.

Potential outcomes respond solely to receipt of the treatment, not to the random assignment of the treatment or any indirect by-product of random assignment.

Jeopardized when different procedures are used to measure outcomes in treatment and control groups.

Question 2

Non-interference

Answer 2

Potential outcomes for observation i reflect only the treatment or control status of observation i and not the treatment or control status of other observations.

No matter which subjects the random assignment allocates to treatment or control, a given subject’s potential outcome remains the same.

Question 3

Random assignment

Answer 3

…

Question 4

Solutions to excludability

Answer 4

Double-blindness and parallelism.

Double-blindness: Neither subjets nor researchers are aware of which treatments the subjects receive

Parallelism: Symmetri, i.e. same surveys, same times and so forth