Question

 

Even small RCTs are always preferable to quasi experimental approaches like RDD and Diff-in-Diff.

Answer 1

"Even small RCTs are always preferable to quasi experimental approaches like RDD and Diff-in-Diff"

The answer is uncertain...

There are other important considerations that you can think beyond statistical efficiency.

For example, lets take an example of Scholarship. In RCT (Randomised Controlled Trials) approach, you are going to give many scholarships to a randomly selected group of people, out of which many might be not strong performers. That is a wastage of resources.

In a Regression Discontinuity Design (RDD) approach, you normally don't provide any academic scholarships to the weak performers. So it is not a wastage of resources. But, the resulting inference might be statistically inefficient.

Actually. we don't want to give scholarships to the low performers, but we want to offer it to the top performers, and then randomise for a middle group.

So you gain some efficiency relative to pure Regression Discontinuity Design (but you're a little wasteful), and you're less wasteful than a pure Randomised Controlled Trials (but you lose some efficiency).

Similar is the case with Difference in Difference... So, you can't directly say that RCTs are always preferable to quasi experimental approaches like RDD and Difference in Difference as every process has its merits and demerits.

So, it is uncertain...