Imágenes de páginas
PDF
EPUB

8. Article by William R. Kinney, Jr., “Predicting Earnings: Entity versus Subentity Data"

[From Journal of Accounting Research, Spring 1971]

Reprinted from JOURNAL OF ACCOUNTING RESEARCH

Vol. 9, No. 1, Spring, 1971

Printed in U.S.A.

Predicting Earnings: Entity versus
Subentity Data

WILLIAM R. KINNEY, JR.*

The reporting of subentity earnings by large diversified corporations has been the subject of much discussion and research. Attention has centered on the need for the information by investors and the theoretical and practical problems to be encountered in furnishing the information.1

Those who have supported the reporting of subentity earnings data have argued that rates of growth and profitability and degrees of risk differ among the segments of a company operating in substantially different industries. This makes the prediction of consolidated earnings of the diversified company unnecessarily difficult. Since little information on subentity earnings has been made public in the past, research in this area has been confined to calling attention to possible uses of the data by investors.

The purpose of this study is to test the relative predictive power of subentity earnings data for a sample of companies which have voluntarily reported sales and earnings data by subentity. Consolidated earnings for these firms will be predicted for 1968 and 1969 using subentity and entity sales and earnings data in conjunction with other investment and economic data available in early 1968 and early 1969, respectively. Specifically the question examined is: Will the disaggregation of consolidated earnings permit better predictions of next year's earnings using certain objective prediction models?

Relatively simple prediction models are employed as objectively as possible in order to isolate a measure of the value of certain investment in* Associate Professor, University of Iowa. The author acknowledges financial support by the Committee on Relations with Universities of the AICPA.

1 For example, see Robert K. Mautz, Financial Reporting by Diversified Companies (New York: Financial Executives Research Foundation, 1968).

128

JOURNAL OF ACCOUNTING RESEARCH, SPRING, 1971

formation in predicting operating results. Undoubtedly, better predictions could be made by skilled financial analysts using subjective judgments and more sophisticated models. The purpose here, however, is to assess whether, in a minimal sense, the reporting of subentity data adds to the investor's capability to predict earnings of the diversified company.

The Prediction Problem

Traditional investment decision models utilize expectations both for an industry and for a particular firm as a part of that industry. Using this approach, the investor would obtain predictions for overall economic activity, such as the gross national product, and try to identify a particular industry whose prospects seem to meet his objectives. To determine the prospects for the company the investor might use its accounting reports to determine trends in the company's profits, given trends in sales, current investments in plant and equipment and other data. Market share and trends in market share could be determined by reference to historial industry data, such as those reported in the Survey of Current Business.3 All of this information could be used to make a prediction of the firm's earnings.

However, for a diversified firm whose segments have different rates of growth, profitability, and risk, this type of prediction model would be inappropriate unless certain conditions hold. One condition is that the company be so completely diversified that it is in every industry and in each in the same proportion so that it becomes a model of the economy. Under this condition, future consolidated earnings could then be predicted by considering estimates of future performance for the economy as a whole such as gross national product. On the assumption that such a diversified company would experience earnings growth equal to the rate of growth of GNP, the rate of growth of the economy applied to the earnings of the company should give good predictions of the future earnings of the firm (given, of course, good predictions of GNP). Under this model:

[blocks in formation]

where E, is predicted consolidated earnings for year i, AGNP, is the predicted percentage change in GNP from year i 1 to year i and E-1 is consolidated earnings for year i 1.

Another condition would be that a less diversified firm is diversified in such a way that fluctuations in earnings among divisions offset and a

'However, an attempt was made to incorporate those measures of subentity performance which financial analysts consider important. See Robert K. Mautz, "Financial Reporting by Conglomerate Companies," Financial Executive (February, 1968), p. 59.

Published monthly by the Office of Business Economics of the U.S. Department of Commerce.

PREDICTING EARNINGS 129

constant overall rate of growth is experienced (implying that the covariances of earnings among divisions are large and negative). This rate of growth may, of course, be greater or less than the rate of growth for the economy as a whole. Consolidated earnings thus grow at a rate different from GNP but at a rate that can be predicted by an analysis of the trend in consolidated earnings. Thus:

[merged small][merged small][ocr errors][merged small][merged small]

where ƒ is some function relating changes in earnings over time.

(2)

However, if firms provide information on segments, it may not be necessary to rely completely on such assumptions. For example, the federal securities laws require that certain diversified companies "indicate, insofar as practicable, the relative importance of product or service or class of similar products or services which contributed 15% or more to the gross volume of business done during the last fiscal year." Using segment revenues the investor could predict future revenues by segment by using predictions of industry revenues and applying the rates of change in industry revenues (and possibly an expected change or trend in market share) to the past revenues of the segment. The subentity revenue estimates could then be summed to estimate consolidated sales and the sum multiplied by the consolidated profit rate to estimate consolidated earnings. Thus:

i,j

[merged small][ocr errors]

(3)

1

where AÍS..; is the predicted percentage change in sales from year i to year i for industry j, 8-1,; is the sales in year i 1 for the segment in industry j, Ē. is the average consolidated earnings and S. is the average consolidated sales.

If subentity earnings data are also reported, the subentity sales method suggested above should be improved by applying subentity profit rates to the subentity sales estimates and summing the subentity earnings estimates to predict consolidated earnings. That is:

i

P1 = Σα

(1 + ATS¡ ̧;)8;–1,j

8. .j

(4)

4 Form S-1 under the Securities Act of 1933 and Form 10 under the Securities Exchange Act of 1934. Product line revenue reporting requirements have recently been extended to product line earnings by the Securities and Exchange Commission through Releases 4988 and 8650. The change will apply to registration statements filed on or after August 14, 1969.

As one source of industry predictions, the Business and Defense Service Administration of the U.S. Department of Commerce makes available annually predictions of the "value of shipments" by industry for the coming year. For example, see U.S. Industrial Outlook 1968, Business and Defense Services Administration, U.S. Department of Commerce, December, 1967.

[blocks in formation]

j

where ē, is the average earnings for the segment in industry j and §.; is the average sales for the segment in industry j.

The Current Study

Pacter lists 32 companies which voluntarily reported subentity earnings data in 1967, with some of them providing such reports for a decade. As Pacter notes, diversity in types of data and the nature and degree of earnings disclosure is common, but it appears that at this time these reports contain the best information available to the general public. The 32 companies formed an initial sample for testing the prediction models described above. However, eight of these were omitted for the following reasons: three were omitted because they were relatively small, two because of reporting inconsistencies among divisions, one firm liquidated during 1968, one sold its second division and hence became a single industry firm, and one firm did not respond to requests for financial state

ments.

Two prediction bases were used. The first served as a standard of comparison and consisted of the consolidated sales and earnings and segment revenues as required by security laws. The second prediction basis used the segment earnings data. Results using four particular prediction models were compared. These were (1) 1967 (and 1968) consolidated earnings multiplied by the predicted increase in gross national product, (2) an extension of the linear trend of consolidated earnings using double exponential smoothing, (3) the sum of predicted subentity sales multiplied by consolidated profit rates, and (4) the sum of predicted subentity earnings.

The measures of consolidated sales and earnings used in the tests paralleled the ones reported by the subentities. Thus, the earnings measure predicted ranged from consolidated net income after taxes to consolidated net income before common cost allocations.

For model one, 1968 consolidated earnings were predicted by multiplying the 1967 consolidated earnings by 1.076 and 1968 consolidated earnings by 1.096. These were the predicted increases in gross national product from 1967 to 1968 and from 1968 to 1969 and were taken from the industry predictions of the Business and Defense Services Administration."

For the second prediction model, consolidated earnings were predicted by using double exponential smoothing to estimate the linear trend in consolidated earnings (as originally reported) and extending the trend to

• Paul A. Pacter, "Some Recent Examples of Earnings Reports by Division," Journal of Accountancy (December, 1968), pp. 40-51.

'U.S. Industrial Outlook 1968, p. vii. The 1969 prediction was obtained directly from the Business and Defense Services Administration.

Pooling of interest accounting plays an important part in the uncertainty surrounding the prediction of earnings of some diversified companies because pooling gives rise to an ever changing base for prediction. Since acquisitions accounted for

PREDICTING EARNINGS 131

1968 and 1969. The base period was from eight to ten years and a smoothing constant of .4 was used to make the trend fairly responsive to change. Models 3 and 4 both utilized subentity sales classified by industry. Industry memberships were based on subentity activity descriptions. As much specificity as the descriptions permitted was used. For example, when a "chemical division" was indicated, the entire 2800 Standard Industrial Classification Code was used unless the text of the report indicated a concentration in a particular type of chemical or chemical product such as plastics (2821) or soaps and detergents (2841).

The segment earnings section of the annual reports provided only a very broad guide for determining the exact output of an operating unit, and text descriptions were necessary to determine the composition in more specific terms. "Industrial division" outputs ranged from construction equipment to industrial sewing machines and industrial solvents. Better predictions may have been obtained if the reporting units had provided further breakdown of their products to the extent different products reflected varying rates of growth, profitability, etc. For example, in one reporting segment titled "consumer" products, outputs ranged from fountain pens to zippers, snowmobiles, and cologne. No further breakdown was provided.

With the specific industries identified in as much detail as possible, the expected change in performance for a particular industry as a whole was determined or predicted and multiplied by the 1967 (1968) sales of the segment of the diversified company for that industry. Various predictors of industry sales were used. The primary source (accounting for about 85 percent of the predictions) was the predicted "value of shipments" by industry as made and reported by the Business and Defense Services Administration in the U.S. Industrial Outlook 1968 and U.S. Industrial Outlook 1969. Exponentially smoothed extensions of the linear trend in historical industry revenue data from the Survey of Current Business was made for certain services not reported in the Industrial Outlook. Finally, an extension of trend in earnings performance of the segments was made for those segments operating in industries for which no industry-wide data were available.10

as pooling of interest could be expected to continue, the trend extension was based on consolidated earnings as originally reported. The alternative (basing the predictions on the artificial construct of the "1967" company as if it had existed as such in prior years) would likely underpredict the rate of growth of expanding companies. 9 Larger smoothing constants were also tested. See footnote 14.

10 Kaiser Industries Corporation has not reported segment revenues in the annual report and thus earnings estimates using method 3 could not be made. In applying method 4 for Kaiser, extensions of trends in historical data were used. For three of Kaiser's four segments an extension of the linear trend in the profits of particular industries were used (corporate profits for 15 broadly defined industry classifications are reported in the Survey of Current Business). For the other segment, an extension of segment earnings was used.

« AnteriorContinuar »