Discussion

Discussion of “Forecasting macroeconomic variables using collapsed dynamic factor analysis” by Falk Bräuning and Siem Jan Koopman

The use of factor models

Factor models have become an important tool in the modern forecaster’s toolkit. They provide a means of producing forecasts when the number of indicator series, $N$, exceeds the number of time series observations, $T$. A further perceived benefit is that they relieve the pressure on the forecaster to select the preferred indicator(s), and in turn indicator-based forecasting model, from a potentially large set.¹

Forecasting at times of structural change

BK invoke the 2007–8 global financial crisis, and the collapse of Lehman Brothers, as a reason for a renewed interest in forecasting. While this crisis has indeed prompted new research themes, BK’s applications end in 2003, precluding an assessment of the performance of their factor model over this particular period of great change.

This discussion therefore seeks to provide empirical out-of-sample evidence on the performance of the BK factor method, relative to benchmarks, at times of

The future is uncertain

While forecasters increasingly pool both information and forecasts in the face of both data and model uncertainty, increasingly they also look beyond the conditional mean forecast. This is because economic loss functions are not generally quadratic, as is assumed when MSE evaluation criteria are used. Under more general loss functions, requiring the production of density forecasts, AR forecasts prove far easier to beat than in many MSE competitions. This is because they fail to capture the

Acknowledgments

Many thanks to Falk Bräuning and Siem Jan Koopman for making the data/code available to enable us to replicate their results.