Is the Assumption of Linearity in Factor Models too Strong in Practice?

The assumption of linearity of factor models is implicit in all empirical applications used in macroeconomic analysis. We test this assumption in a more general setting than previously considered using a well-studied macroeconomic dataset on the U.S. economy, and find strong evidence in support for regime-switching type non-linearity. Furthermore, we show non-linearity is strongly concentrated in certain groups (such as financial variables). Our results, which are robust to serial dependence, suggest the assumption of linearity underpinning factor models might be too strong and gives further support towards developing models which explicitly account for non-linearity