‘Analysis of Big Data in Economics and Empirical Applications’

Geert Mesters, Majid Al-Sadoon and Christian Brownlees, Pompeu Fabra University.

Over the last years, the analysis of large datasets has become an active topic of research in econometrics, with numerous applications in macroeconomics and finance. Large panels of economic and financial time series are commonly encountered in many research areas.

Ever increasing efforts are made to collect new panels and to extend existing ones. One of the main objectives in the analysis of large panels of time series is forecasting. However, forecasting in a high dimensional setting comes with a number of challenges.

The first hurdle is that standard estimation methods cannot always be implemented. Further, when they can, classic inferential theory often breaks down and optimality is no longer guaranteed. Second, determining the dependence structure among the time series in large panels can be a daunting task. For instance, a given series of interest has typically many possible candidate predictors in a large panel and brute force model selection is notoriously challenging when the number of possible predictors is large.

In addition, another compelling reason why large dataset of predictors might be useful in forecasting in Economics is the fact the predictors performance changes over time; therefore, it is reasonable to expect that models? forecasting performance would improve by extracting predictors from a large dataset, which includes all possible predictors. Instability is empirically well-documented for several key economic variables.

The objective of this project is to develop new methodologies for the analysis of large panels of time series geared towards forecasting. The project contributes to the emerging literature on forecasting in high dimensional environments that has been an extremely active area of research over the last years. The current research proposal is divided in a number of projects. Each project aims at producing a paper introducing a new methodology, providing rigorous theoretical analysis and documenting its empirical performance.