Over the past decade of accounting and finance research, the Ohlson (1995) model has been widely adopted as a framework for stock price prediction. While using the accounting data of 391 companies from SP500 in this paper, Bayesian statistical techniques are adopted to enhance both the estimative and predictive qualities of the Ohlson model comparing to the classical approaches. Specifically, the classical methods are used for the exploratory data analysis and then the Bayesian strategies are applied using Markov chain Monte Carlo method in three stages: individual analysis for each company, grouping analysis for each group and adaptive analysis by pooling information across companies. The base data, which consist of 20 quarters' observations starting from the first quarter of 1998, are used to make inferences for the regression coefficients (or parameters), evaluate the model adequacy and predict the stock price for the first quarter of 2004, when the real observations are set as the test data to evaluate the predictive ability of the Ohlson model. The results are averaged within each specified group categorized via the general industrial classification (GIC). The empirical results show that classical models result in larger stock price prediction errors, more positively-biased predictions and have much smaller explanatory powers than Bayesian models. A few transformations of both classical and Bayesian models are also performed in this paper, however, transformations of the classical models do not outweigh the usefulness of applying Bayesian statistics.
| Identifer | oai:union.ndltd.org:wpi.edu/oai:digitalcommons.wpi.edu:etd-theses-1765 |
| Date | 06 May 2005 |
| Creators | Lu, Qunfang Flora |
| Contributors | Balgobin Nandram, Advisor, , , Huong Higgins |
| Publisher | Digital WPI |
| Source Sets | Worcester Polytechnic Institute |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Masters Theses (All Theses, All Years) |
Page generated in 0.002 seconds