1: \begin{abstract}

2: In this paper, we propose a varying coefficient panel data model with

3: unobservable multiple interactive fixed effects that are correlated with

4: the regressors. We approximate each coefficient function by

5: B-spline, and propose a robust nonlinear iteration scheme based on

6: the least squares method to estimate the coefficient functions of interest.  We also establish the asymptotic

7: theory  of the resulting estimators under certain regularity

8: assumptions, including the consistency, the convergence rate

9: and the asymptotic distribution. Furthermore, we develop a least squares dummy variable  

10: method to study an important special case  of the proposed model: the varying coefficient panel data model with additive fixed effects.

11: To construct the pointwise confidence intervals

12: for the coefficient functions, a residual-based  block bootstrap method is proposed to reduce the computational burden as well as to avoid the accumulative errors. Simulation studies and a real data analysis are also carried out to

13: assess the performance of our proposed methods.

14: 

15: \vskip10pt

16: 

17: \noindent\textit{Key words}: Varying coefficient model; Panel data; Interactive fixed effect;  Bootstrap; Least squares dummy variable method; B-spline

18: 

19: \end{abstract}

20: