Balanced randomization within clusters

Question

I was hoping someone could point me in the right direction on a randomization method. We want to randomize teachers to receive an educational intervention. The catch is that we have a number of schools and these schools are very different in terms of various socioeconomic factors so we don’t want to end up in a situation where all (or nearly all) of the teachers in a given school are randomized to the same intervention (the number of teachers per school is fairly small so this is a possibility). Can someone point me in the direction of a randomization procedure that will result in a near 50/50 split within each school?

Thank you

Answer 1

Assuming you have one treatment, a simple way to do this would be:

1. Pick a school
2. Randomly pick 50% of teachers in that school
3. Assign those 50% to treatment, and the rest to no-treatment
4. Repeat for each school

This will assign 50% of all teachers and in each school to treament and 50% to no-treatment