Pooling data for surveys from different years

[Duman Rad]

Hi,

I really appreciate if someone can help me regarding the following.

To do a diff-in-diff analysis, I am running regressions with pooled data from different waves of a survey. My results are robust for including wave fixed effects. However, if I drop older waves of the surveys, the results are a little bit different.

Now, I wonder if there is anything wrong with pooling data from different waves (conducted in different years)? Note that I am including many fixed effects including year fixed effects in the regressions.

Thanks.

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

Your project appears predicated on using data from different surveys. If the data from the waves are really comparable, then it makes sense to treat them as a normal panel data set. That dropping some early years changes results a little change should not worry you - any change in the sample may change results. But if it makes a big difference, then I'd worry. I'm a little surprised this appears as a problem in a d-i-d analysis since the action is largely around the event.

[Duman Rad]

Thanks Phil for the reply.
Dropping early years doesn't change the results. I just wanted to make sure that I am not making an obvious mistake.