Hello dear Stata Community,

I have a general question about estimators and the number of observations.

For my paper, I have a pooled data set from three waves of data (2008-2012, 2010-2014 and 2012-2016).

I am estimating my research question of interest using a difference-in-differences design using pooled data. I have three treatment period. My treatment variable is binary and takes the value of "1" between 2009-2011, 2013 and 2011, and 2015 and 2013. The pre and post-treatment periods are thus the year before and after the treatment measurement.

I am confused about the point whether I should restrict each individual to only one wave or pool them. My estimators turn out to better when I apply this restriction. The question is about a tradeoff between observation number and estimators.

My current observation number is about 17,000. When applying the restriction, I almost loose half of these, however, my estimators are better.

I would really appreciate your input!

Greetings,
Miena