Which (Nonlinear) Factor Models?
Abstract:
We extend traditional tests of factor models to incorporate nonlinearities. The metric for model evaluation becomes the Sharpe ratio of the mimicking portfolio of a nonlinear stochastic discount factor (SDF) pricing the model factors. Empirically, we investigate the implications of an economically meaningful family of nonlinear SDFs for evaluating popular factor models. We find that, relative to the linear case, introducing nonlinearities substantially improves pricing performance and changes rankings among competing models. The preferred model depends on the test assets: unlike the linear approach, test assets are relevant for model comparison as they are needed to mimic nonlinearities in the factors.
The seminar will be held in person.
Language: English
Supported by the Luxembourg National Research Fund (FNR) (2022/17573036)