The Relationship Between Time-Series Models and the Security Market's Expectation of Quarterly Earnings.
用预测能力和市场关联测试两种标准，评估五种季度盈利时间序列模型，发现Brown-Rozeff模型最优，且意外盈利变化仅解释公告日市场信息的约15%。
Abstract ABSTRACT: The relationship between a set of five quarterly earnings time-series models and the security market's expectation of quarterly earnings was assessed using two evaluative criteria: 1) prediction of quarterly earnings numbers and 2) security market association testing. A relatively large sample of 240 NYSE firms was analyzed using an extensive time-series data base over 1962-1974 and a holdout test period from 1975-1977. Overall, the Brown-Rozeff model dominated both the predictive ability tests and the cumulative average residual analysis. These results are consistent with the security market adjusting for the presence of seasonality in the quarterly EPS data. Although a significant contemporaneous association between unexpected earnings changes and abnormal security returns was revealed, the informational content of the earnings numbers conveyed only approximately 15 percent of the total information available to the market as of the earnings announcement date.