Power properties of the Sargan test in the presence of measurement errors in dynamic panels

This paper investigates the power properties of the Sargan test in the presence of measurement errors in dynamic panel data models. The general conclusion from the Monte Carlo simulations is that the Sargan test, in many cases, leads the econometrician to accept misspecified models with sometimes severely biased parameter estimates as a result. This is especially true when the number of cross-sectional units is small and when there are measurement errors in the dependent variable. To investigate if the simulation results have any bearing in real applications, we used the data in Arellano and Bond (1991) and re-estimated their employment equations with the difference that we deliberately imposed additive and multiplicative measurement errors in the employment and wage variables. It turned out that the Sargan test always accepted the misspecified models while we at the same time ended up with biased parameter estimates. The conclusion from this paper is that in the very likely case of measurement errors in either the dependent or any of the independent variables, we will, if we rely on the Sargan test, quite likely accept a misspecified model and end up with biased results.