Forecasting, Firm performance

Reviews the book "Business Decision Theory,' by Paul Jedamus and Robert Frame.

The purpose of this article is to describe a model of a forecaster's biasing behavior. The model should be of interest to accountants because the model provides a link between a forecaster's biasing behavior and the quality or acceptability of his forecast as an input to an accounting system. The article attempts to describe how a model which assumes that an individual bases his present action on his experience on the preceding trial can be applied to describe biasing behavior in a game theory-like situation. The individual forecaster (F) is asked to provide a forecast of some variable such as sales, cash flows or expenses for his area of responsibility. F works under a payoff structure which encourages him to intentionally bias his forecast within certain limits. F's desire to bias his forecast, and the desire of the recipient (R) to adjust F's forecast for the presence of bias places F and R in a situation not unlike a nonzero-sum game. However, most attempts by psychologists to use game theory as a descriptive theory of behavior have failed. The models discussed are abstractions of forecasting situations.

In recent years, several studies have appeared which provide some support for an inverse relationship between earnings variability and share price.' That is, increased variability in reported earnings, other things equal, appears to reduce the price of a firm's shares. A rationale for this behavioral phenomenon has been suggested by several accounting writers. Almost two decades ago Hepworth contended that :2

The article comments on the paper "Segmental Financial Disclosure by Diversified Firms and Security Prices," by R.F. Kochanek, published in the April 1974 issue of the journal "The Accounting Review." Kochanek studied the relationship between segmental disclosure by diversified companies and the forecastability of their earnings. Kochanek's test of the impact of segmental disclosure on earning forecasts employs an indirect measure of the forecastability of earnings. While Kochanek's sample companies are diversified, some would appear to fall primarily into industry groups which vary in earnings volatility. The possible impact of number of segments or industry factors on earnings forecastability bears on Kochanek's results only if quality of reporting varies systematically with number of segments or industry affiliation. To test for this, the authors computed the correlation between number of product lines and Kochanek's disclosure scores. Kochanek executed a creative but inadequately controlled piece of research in attempting to study this relationship.

This article presents information on the smoothing hypothesis in accounting. This hypothesis assumes that managers perceive their performance measure to be a decreasing function of earnings variability. Based upon this assumption, managers could be expected to make accounting policy decisions which tend to smooth reported earnings. The focus of this paper is on the differential impact of the cost versus the equity method of accounting for unconsolidated subsidiaries. Assuming that dividends typically fluctuate less than earnings, one might reason that the cost method would generally result in smoother reported earnings for the parent. However, closer examination shows that even if subsidiary dividends fluctuate less than subsidiary earnings, the equity basis can still result in smoother reported earnings. The least squares criterion was used to develop a linear relationship between earnings and time on each basis. The slope co-efficient of the line was used as an estimate of the rate of growth in before-tax earnings. Thus, each firm had a growth rate estimate for each valuation basis. The variability of earnings about the linear trend line described above provided the basis for a measurement of earnings variability. The actual measure of earnings variability used was the mean square error (MSE) of earnings about the linear trend line standardized by dividing by average earnings over the time period considered.

The article presents a comment on the interpretation of API. In "Interpreting the API," Ronald M. Marshall concludes that API does not always provide a proper measure of either the private value of accounting data or the association between unexpected accounting signals and unexpected market returns. In addition, Marshall concludes that an alternative formulation of API always produces measures of these attributes, which are at least as good as those obtained via API. On the basis of these conclusions, Marshall argues that API constitutes the more appropriate tool for use in accounting research. Authors do not disagree with Marshall's conclusions from a conceptual viewpoint, but they do question the desirability of using API in accounting research because of its inherent subjectivity and costliness in terms of time. Since API does not possess these defects, authors believe that it constitutes the better research technique when, conceptually speaking, it can be expected to yield results, which are equivalent to those that would be produced using API. One objective of the paper is to identify an important sufficient condition, under which the two API will produce equivalent results.