Predicting Credit Rating Using the Dynamic Ordered Probit Model with Autocorrelation Structure
| Autor: | Cheng-Chun Liu, 劉政君 |
|---|---|
| Rok vydání: | 2008 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 96 Ordered probit model for using a single cross-section of companies to predict bond ratings have been proposed by Kaplan and Urwitz, and extend by Blume et al. for using the panel data of companies. This is a dynamic ordered probit model (DOPM). The methodology of Blume et al. assume that the time series of each individual company is zero correlation structure. In this paper, we will use generalized estimating equations (Lipsitz et al., 1994) approach to estimate the parameters of DOPM with autocorrelation structure. Through the literature, the explanatory variables was selected from accounting variables in Poon (2003) and Pettit et al. (2004), market-driven variables in Shumway (2001), and the KMV-Merton default probabilities (Merton, 1974; Bharath and Shumway, 2007). Incorporate industry effects in Chava and Jarrow (2004) and Pettit et al. (2004). We will compare the difference of estimation and prediction between DOPM with autocorrelation and zero correlation structure. The empirical result of the holdout sample demonstrates that, the prediction accuracy rate of the model with autocorrelation structure is more than the model with zero correlation structure. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|