A1388
Title: Training economic tracking portfolios using trees of assets
Authors: Onno Kleen - Erasmus University Rotterdam (Netherlands) [presenting]
Anastasija Tetereva - Erasmus University Rotterdam (Netherlands)
Rasmus Lonn - Erasmus University Rotterdam - Econometric Institute (Netherlands)
Abstract: Stock returns can help in nowcasting and forecasting economic variables such as consumption, inflation, or unemployment. Typically, economic tracking portfolios are based on aggregated portfolio returns. A modified version of regression trees is introduced that allows us to employ individual stocks to track economic variables. We build cross-sections of managed portfolios that serve as tracking portfolios, and that can be used in asset pricing models. Empirical results show that our portfolios display superior tracking properties across various market conditions.
