This paper provides an impressive, yet compact and easily accessible review of the econometric literature on panel data analysis. Professor Cheng Hsiao has succeeded in surveying, in a coherent manner, classic results as well as more challenging recent developments on nonlinear models, crosssectional dependence, and long time-series panels. The coverage of topics in the article reflects the breadth of Professor Hsiao's important contributions to panel data econometrics over many years.In my comments I will focus on two aspects of recent work on nonlinear models. Firstly, I will discuss state dependence and dynamics from a treatment effect perspective. Secondly, I will consider the issue of choice of population framework and its implications for identifiability. Both comments are closely interconnected.