reverse causality problem in panel data

[Zinan Wang]

hello everyone,
i am writing my dissertation recently, but i got the reverse causality problem.
in fact, maybe not only x affect y, but y affect x as well. therefore we need trying to avoid such problems
the lagged x is the possible way to deal with it i think, as the present is less likely to affect past x.
But, is any other method to test or control reverse causality problem?

by the way, i use both x and lagged x to estimate the influences upon y by fixed effects model and random effects model.
is any bias of problems on my results? if so what kind of problems will be happened?

thanks for any help
zinan wang

[Aspen Chen]

For your first question, I would consider instrumental variables. The method has become increasingly common, and Stata has readily available packages to implement it.