The Overlapping Data Problem

This paper provides a guide to handling the overlapping data problem.Overlapping data are often used in both economics and finance,but applied work often uses inefficient estimators. Thus there appears to be a need for a better understanding of the overlapping data problem. Under strict exogeneity,generalized least squares (GLS)is asymptotically efficient.Yet,the main reason to ever use GLS that holds up to scrutiny is some form of missing observations. Frequently used procedures such as the Newey-West procedure have large small-sample bias unless accompanied by a bootstrap procedure.Using nonoverlapping data when overlapping data are available can be grossly inefficient.Monte Carlo results are presented to support the arguments.In the case of errors in variables or without strict exogeneity,GLS performs poorly. Thus,no method is always preferred. Key words:autocorrelation,Monte Carlo,Newey-West,overlapping data