Panel Data Econometrics
I was impressed to discover today that a decade-old paper by Richard Blundell and Steve Bond, “Initial conditions and moment restrictions in dynamic panel data models,” has 1503 cites on Google Scholar. The paper discusses the validity of panel data estimation approaches using particular selections of GMM identification restrictions. Techniques validated by the paper have become ubiquitous for their generality.
Saltwater empiricists in the US, however, tend to expect clean experimental variation. And freshwater empiricists in the US also generally pursue a different approach, relying on identifying restrictions that come from models with more structure. The Blundell-Bond paper and the related literature suggest a generically valid approach conditional on certain lags and lagged changes of error terms and dependent variables being uncorrelated. I am led to the question: which real-world settings reliably satisfy the required conditions?
