Comparative Study of Earnings Forecasts Techniques: Which one generates the most accurate earnings forecasts under Chinese environment?

Zhang, Huipu (2014) Comparative Study of Earnings Forecasts Techniques: Which one generates the most accurate earnings forecasts under Chinese environment? [Dissertation (University of Nottingham only)] (Unpublished)

[img] PDF - Registered users only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (709kB)

Abstract

This work has comparatively studied on the accuracy of earnings forecasts by random walk, exponential smoothing method and multiple-regression model, for different forecast horizons and different industries. The main difference of the current study from the prior literature is that this study firstly employs the data from Chinese listed firms, whereas the prior researchers generally use the US data. The primary objective of this study is to examine whether the prior findings of earnings forecasts are still held under Chinese environment and recommend which statistical forecasting method is most adaptive to the earnings forecasts for Chinese firms. The empirical result of this study shows that the exponential smoothing method is slightly superior to the random walk model while the multiple-regression model is outperformed by the other two methods for both one-year-ahead and two-year-ahead horizons. All forecast methods produce more accurate earnings forecasts for shorter horizon than longer horizon. Further, the combination method works only when there are large forecast errors and earnings forecasts generated by many techniques are combined. In addition, random walk model dominates the retail and electrical industries, while the exponential smoothing method dominates the real estate industry.

The current study recommends that, when making earning forecasts for firms under Chinese environment, neither statistical forecasting method could fit all firms. After inspection of the historical earnings patterns, random walk model is recommended for the firm with stable and persistent earnings while exponential smoothing method is recommended for the firm with strong seasonal or trend earnings pattern. Multiple-regression model and combination forecasts are generally not recommended unless the special cases appear to require the reduction of abnormally large forecast errors.

Item Type: Dissertation (University of Nottingham only)
Depositing User: EP, Services
Date Deposited: 23 Dec 2021 14:22
Last Modified: 23 Dec 2021 14:22
URI: http://eprints.nottingham.ac.uk/id/eprint/27181

Actions (Archive Staff Only)

Edit View Edit View