Week 1 - RCTs

Question 1

4 limitations of RCTs

Answer 1

1. Failures to follow the experimental protocol
   - Good randomisation ensures that the treatment and control groups are COMPARABLE, ie. Participants are EQUALLY represented in both groups
2. Attrition = ppl DROPPING OUT of expt.
   - typically a serious problem in expt.s (but not concerning if happens in both treatment & control groups)
   - likely to be a SYSTEMATIC issue that REDUCES the COMPARABILITY of the treatment & control group, complicating the estimation of the ATE
   - will cause DIFFERENCES other than the treatment assignment between the 2 groups
   - Only scenario in which attrition is not a problem is ENTIRELY RANDOM ATTRITION.
   > If an entirely random subset of participants dropped out, then attrition doesn’t change the average composition of the treatment and control group
3. Experimental effects
   - some ppl change behaviour when they’re taking part in an expt, whether intentionally (out of malice) or not
4. External validity
   - Ppl react to the CONTEXT they live in
   - outcome may not hold in diff. country, ppl or point in time

+ considerations about ETHICS

Question 2

Russell’s chicken problem

Answer 2

Things evolve over time. The same action doesn’t mean it will give the same outcome in the future.