A Comparative Study of Different Shrinkage Estimators for Panel Data Models

G. S. Maddala

The present paper uses small-sigma asymptotics to show that in general the shrinkage estimators have superior properties among the individual least squares estimators, the simple average estimators, the weighted average estimators, estimators obtained by shrinking towards the simple average, and estimators obtained by shrinking towards the weighted average. The shrinkage estimators are used to derive short-run and
long-run price and income elasticities for residential natural gas demand and electricity demand in the US based on panel data covering 49 states over 21 years (1970-90). They are also used for out of sample forecasting.