A Multivariate Analysis of Annual Earnings Forecasts Generated from Quarterly Forecasts of Financial Analysts and Univariate Time-Series Models
比较了财务分析师综合预测与单变量时间序列模型在年度盈利预测上的表现，探讨综合模型是否优于单变量模型，对投资决策和会计准则制定有参考价值。
The Financial Accounting Standards Board [1977] recently emphasized the importance of forecasted accounting earnings in the formulation of investment decisions. Empirical investigations into various aspects of the investment decision process, such as cost of capital, firm valuation, and the relationship between earnings and stock prices, have utilized forecasted accounting earnings extensively as a measure of earnings expectations. Thus, both policy boards and empirical research support the importance of forecasted accounting earnings. Current sources of these forecasts which are widely available are univariate time-series models and financial analysts. In the future, another may be management forecasts, if the SEC and FASB should desire to make those more widely available. Analysts' and managements' forecasts may be characterized as representing comprehensive models, in that input to these models can incorporate numerous variables both endogenous and exogenous to the firm. In contrast, the univariate timeseries models incorporate a single variable, past earnings. Both comprehensive and univariate models have advantages and disadvantages. Currently, a major question is the value of a comprehensive model relative to a univariate model. Another question is whether a univariate model should be identified individually for each firm or if a generally identified or premier model would provide forecasts that are