Beta forecasting at long horizons

Download files
Access & Terms of Use
open access
Copyright: De Oliveira Ferrazoli Ribeiro, Fabio
Altmetric
Abstract
Since the inception of beta as the main risk measure for equity assets, practitioners and academics are required to estimate its future value for uses such as the calculation of hurdle rates of projects, cost of capital and valuation of equities. With the goal of providing guidance, this thesis follows the recently developed literature on realised betas and the literature on beta forecasting and extends it to evaluate forecasts of betas on long horizons such as 6 months and one year. Our sample is of 15 US stocks from the Dow Jones index with 60 years of data. We implement constant, autoregressive and heterogeneous autoregressive beta forecasting models and compare their accuracy with that of the standard Fama-MacBeth beta (Fama and MacBeth, 1973). We conclude that the autoregressive model with one lag, based on in-sample sizes of 30 years is the most suitable alternative. This model presents the smallest forecasting error (and variance) and is also statistically superior in terms of bias. By implementing the model suggested in this paper, practitioners are able to reduce forecast error by as much as 30.2% when forecasting 6 months ahead and by as much as 29.8% for forecasts one year ahead when compared to the Fama-MacBeth beta, the industry benchmark over the last 40 years. The economic application of using the autoregressive model is also significant with an estimated increase of 0.54% (0.5%) in cost of equity of firms. This could translate into a 3-6% decrease in the net present value of a five year illustrative investment project.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
De Oliveira Ferrazoli Ribeiro, Fabio
Supervisor(s)
Reeves, Jonathan
Cenesizoglu, Tolga
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2013
Resource Type
Thesis
Degree Type
Masters Thesis
UNSW Faculty
Files
download whole.pdf 1.28 MB Adobe Portable Document Format
Related dataset(s)