财务危机预警文献,仅供学习研究
firms (2,843 firm-year observations) (See Panel A of Table 2). Panel B of Table 2 provides sample distribution among types of bankruptcy and Table 2-3 presents sample distribution by industry. Eleven types of bankruptcies are studied in this paper, with “the firm has checks bounced” (32%) and “The firm receives financial supports from the government” as the most frequent types (31%). The bankrupt firms are from various industries, heavily concentrating in electronics industry.
Information used to predict bankruptcy (including financial ratios and auditors’ opinions, macroeconomic factors) is for one year prior to the event year. Company bankruptcy event and predictor information is obtained from various sources including Taiwan economic journal, [including basic company data for public listed companies, financial data for public listed companies, auditors opinions database, and macroeconomic database. 3. Methodology
3.1. Discrete-time hazard model
We use discrete hazard model to analyze the prediction ability of auditors’ opinions, macroeconomic variables, and industry variables. Model parameters are estimated using maximum likelihood functions. The significance of individual variable is examined using Wald statistics. The overall goodness-of-fit for models are evaluated based upon likelihood ratio. Following prior literature (Sun 2007), Vuong test (1989) is employed to compare the overall fits of different models.