THE ECONOMIC IMPLICATIONS OF A DECOMPOSITION ANALYSIS OF THE RISK CHARACTERISTICS OF SELECTED PROPERTIES OF EARNINGS FORECASTS AND FORECAST ERRORS

ANDREW GRANT SNYIR, Purdue University

Abstract

Investors must make decisions regarding which particular assets in which to invest their limited resources. Information of all types is used by the investors in the market to set security prices through their trading actions. While most accounting data is historical in nature, forecasts of future accounting data such as earnings per share and resulting dividends may be very important in the investment decision. Also, since systematic security risk is developed by decomposing total return variability into systematic and nonsystematic components, the variability of EPS over time, the difficulty or uncertainty of predictions of EPS, and decompositions of these quantities may be related to decomposed components of security risk. If this is so, a more complete understanding of the connection of accounting data and investor behavior as it affects security returns may benefit from the study of this relationship. There are a number of theoretical models which suggest testable relationships between forecast error and related measures, and various properties of security return. These include systematic security risk, variance of total security return, and individualistic security risk as measured by the variance of the error term of the Capital Asset Pricing Model. The purpose of this study is examine various properties of analysts' EPS forecasts with a view toward determining underlying relationships of EPS forecasts with other endogenous and exogenous variables, such as security risk and macro economic projections. This study examines these relationships for a large sample over a relatively long time period. Additionally, it decomposes the EPS forecast error into various components to determine whether these components of forecast error are more closely related to other characteristics of the firm than the total EPS forecast error. The forecast consensus of a number of analysts' forecasts is also decomposed and related to other characteristics of the firm. Five hypotheses are tested. The first hypothesis tests for a relationship between an individual firm's EPS forecast error and the macro forecast error of Gross National Product estimates. A significant relationship is found, indicating that individual forecast error is related to the difficulty in making contemporaneous macro forecasts. The second hypothesis test results in the finding of a significant relationship between systematic EPS forecast error and systematic security risk. The third test supports the hypothesis that Theil decomposed forecast error components are not significantly related to systematic security risk. The test results of hypothesis four indicates that there is a significant relationship between individual forecast consensus and the macro forecast consensus of GNP. The fifth hypothesis test results in the finding of a significant relationship between individual forecast consensus and systematic security risk. The conclusion of the forecast error regression test is that any standard of forecast error performance should take the relationship of a firm's EPS forecast error and GNP forecast error into account. The conclusion of the Theil decomposition correlation analysis is that it does not reveal any new sign of relationships with forecast error and security risk. The conclusion of the forecast consensus test is that the individual EPS-macro GNP forecast consensus relationship does exist. Further study of the forecast consensus-security risk relationship may be even more successful in helping to set forecast evaluation standards and procedures.

Degree

Ph.D.

Subject Area

Accounting

COinS