The time traveller’s CAPM

Abstract

This empirical study comprises six emerging market portfolios and five industry replicating portfolios from the USA, using data from 2005 to 2014. The purpose of this study is to test the ability of the ex-ante beta against the ex-post beta using six different generalised autoregressive conditional heteroscedasticity (GARCH) models and the machine learning artificial neural network (ANN) to construct the ex-ante models. Whereas most studies use GARCH models for ultra-short forecasts, little is known about long-run forecasts using these models. The ANN model is also trained to learn the capital asset pricing model (CAPM) to see if the computer can improve upon the traditional model. This study uses a one month (five week) forecast window in all its sample and subsample horseraces, 55 races in all. The ability to forecast the CAPM beta parameters, travel to the future, was found to be more accurate than the historical method across all 11 markets for the three three-year, two five-year and one full 10-year sample periods. Using out-of-sample testing (backtests), the SGARCH was the best predictor overall in developed markets while the IGARCH was in emerging markets.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 18 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.