Project title The Network Bias in Factor Models
Duration  6 weeks

Statistical factor models are a very popular statistical tool in macroeconomics, especially in the study of business cycles co-movements and macro-financial systemic risk. The popularity of factor models lies in the fact that they summarize high-dimensional data set in few principal components (aggregate shocks).

However, it has been emphasized that factor models can largely overestimate the importance of aggregate components, or alternatively largely underestimating the systemic importance of idiosyncratic shocks to industries (Foerster, Sarte, and Watson 2011).

This project aims to study under what conditions of the underlying network structure between assets/industries/banks; factor models deliver a consistent decomposition of aggregate versus idiosyncratic disturbances. The project will deliver a characterization of the network features that (in)validate the estimation as well as a test that identifies environments that are suitable for the estimation of factor models. Finally, the project aim to provide a series of quantitative applications (input-output models, banking networks, etc).

Expected outcomes & deliverables
  • Data processing skills
  • Quantitative and coding skills (Matlab)
  • If the student prefers, he/she can present the work at the end of the project.
Student qualities

Background in economics, preferably a honours or master student in applied econometrics or master in economics.

Primary supervisor

Jorge Miranda-Pinto, Christiern Rose

Further information Interested Applicants can contact Jorge Miranda-Pinto at