To pool or not to pool: Possible heterogeneous slopes in a functional coefficient model

This paper investigates the use of different estimators when dealing with possible heterogeneous slope coefficients among individuals in a panel data setting. We consider both parametric estimators and non-parametric time-variant estimators. The main estimator investigated is an averaging estimator proposed in Wang, Zhang, and Paap (2019). We review the performance of this estimator and extend it to a time-variant coefficient estimator. Results show the estimator performing particularly well with a group structure among individual slopes and performing quite robust when coefficients are completely homogeneous or heterogeneous.