fixed effects question I am analyzing survey respondent data. Respondents are nested within regions and are surveyed 4 times. Over the 4 time periods, different regions undergo a policy change (go from 0 to 1) randomly (or let's hope). I'm interested in the change in attitudes when R's got the new policy.
This is probably a econ 1 question (but I'm not an economist). Is the proper way to analyze the effect of the policy change just regression with unit-level fixed effects? Is there any reason to include wave fixed effects as well? Then do I need to cluster for the region?
thanks!
 A: assuming that all individuals in the survey get exposed to the policy change, it seems what you can do is more like an event study although a fixed effects regression helps you to get rid of the unobserved individual heterogeneity and remove year specific factors. So the simplest regression you can run is:
$Y_{it} = \alpha_{i} + \beta_{t} + \beta_{0} X_{it} + \beta{1} Z_{it} + \varepsilon_{it}$
where $\alpha_{i}$  and $\beta_{t}$ are your unit and year level fixed effects. $X_{it}$ here is your policy variable whihc say is 1 in years when the policy came into place and 0 before that. $\beta_{0}$ is interpretable as the change in an individuals views when she is exposed to the policy change. Note that interretation may not be causal but you need to clarify what exactly is the nature of policy change, and if individuals get to select into the policy or if everyone is exposed to it regardless. $Z_{it}$ are other unit level time varying factors that may affect attitudes like schooling, or marriage status, and so on.
Hope this helps!
