We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A COMPARISON OF CROSS-SECTIONAL AND TIME-SERIES BETA ADJUSTMENT TECHNIQUES.
- Authors
Ushman, Neal L.
- Abstract
The Stable-beta mode! is by far the least accurate of the six models examined. It generates biased forecasts and consistently has the largest Mean Square Forecast Error. The naive model is the most accurate of the models in terms of generating unbiased forecasts. However, in terms of the bias percentage of the MSE, these forecasts are not significantly better than those generated by any of the four beta adjustment models. Both time-series models have significantly smaller MSE's than the naive model and are significantly more efficient in all cases. While the MSE of either cross-sectional model is not significantly lower than that of the naive model, the cross-sectional autoregressive model is significantly more efficient in all cases and the Bayesian revision model is significantly more efficient in two of the three cases. In no case is the MSE of the naive model significantly lower than that generated by the stable beta model. In conclusion, there are models available to forecast beta which improve upon estimates which either ignore potential shifts in beta entirely or naively assume beta in period t + 1 equals beta in period t. Most previous research has approached the selection of such a beta-adjustment model using a cross-sectional approach, and researchers have conducted their tests in an environment where no shifts in beta were expected. This study has offered for consideration two time-series models which have sound analytical and/or statistical underpinnings, and has tested them in an environment where beta did shift. The results presented indicate that either approach improves the forecast of beta.
- Publication
Journal of Business Finance & Accounting, 1987, Vol 14, Issue 3, p355
- ISSN
0306-686X
- Publication type
Academic Journal
- DOI
10.1111/j.1468-5957.1987.tb00100.x