Actually, I think it's more sound to approach this with clustered standard errors. Basic intuition is similar, but the sample size is what it is per person, and your observations aren't independent across draws but are across people.

https://en.m.wikipedia.org/wiki/Clustered_standard_errors