On CCE estimation of factor-augmented models when regressors are not linear in the factors Ignace De Vos Lund University Joakim Westerlund ∗ Lund University and Centre for Financial Econometrics Deakin University January 8, 2019 Abstract In empirical research it is often of interest to include non-linear functions of the ex- planatory variables, such as squares or interactions, in the specification. A popular tech- nique to estimate such models in the presence of common factors is the Common Cor- related Effects (CCE) methodology. However, this approach assumes that the regressors are linear in the factors, which is not the case if variables enter non-linearly. In this note we show how CCE should be implemented when some regressors violate the linear factor model assumption. JEL Classification: C12; C13; C33. Keywords: CCE; factor-augmented regression models; non-linear regressors. 1 Introduction One of the most popular estimation techniques for panel data models with common factors is the Common Correlated Effects (CCE) approach of Pesaran (2006). The methodology is based on taking the cross-sectional averages (CA) of the observables as estimated factors, and applying Ordinary Least Squares (OLS) conditional on these estimates. A major reason for the popularity of this approach is its simplicity and generality. The CCE methodology as originally presented is, however, restrictive in the sense that all the regressors are assumed to satisfy a linear static factor model, which is not always the * Department of Economics, Lund University, Box 7082, 220 07 Lund, Sweden. Telephone: +46 46 222 8997. Fax: +46 46 222 4613. E-mail address: joakim.westerlund@nek.lu.se. 1