Predictive Content of Earnings Components



PREDICTIVE CONTENT OF EARNINGS CLASSIFICATION

By

Sansaloni Butar-butar(

ABSTRACT

To make financial statement more informative and reliable, some contend that any gains and losses, directly or indirectly related to firm normal operation should be recognized in income statement report. At the opposite side, many believe the idea of recognizing irregular items in the body of income statement may mislead users of financial statement. At present, accounting profession require firms to report gains and losses from abnormal transaction apart from normal transaction. The study investigate whether the current practice to separate earnings components based on its irregularity can be justified. The result shows that disaggregating net income into income before special items and specials item do not increase the predictive content of earnings. But further disaggregation of income before special items into operating income and non-operating income and tax do increase the predictive content of earnings.

KEYWORD: irregular items, gains and losses, special items, normal operation

I. INTRODUCTION

Ball and Brown (1968) documented the first empirical result of the usefulness of earnings number in firm valuation. The result seems robust for large body of research using different data, different period of time and different stock exchange found that earnings has information content (( Beaver, 1968; Brown, 1970; McEnally, 1971; Beaver, Clark and Wright, 1979; Foster, 1975)

In addition to earnings information content studies , large body of research also seek how market participants react to earnings components. Analysis of earnings components may provide additional information about the persistence of earnings number. Does the earnings increase (decrease) come from normal operation or come from random events ? The question has drawn attention of many researchers. If earnings increase (decrease) is triggered by regular-normal activities of company, investor may expect that the increase will last to the next period. Quality of earnings is considered high when most of components come from normal operation.

Finger (1994) shows that current earnings can be used to predict future earnings and cash flow. Ou and Penman (1989) developed models to predict change in earning per share using balance sheet and income statement items. Givoly and Hayn (1992) had identified nonrecurring components of earnings and found that reported earnings excluding non-recurring items has better predictive ability than reported earnings including non-recurring items.

In order to gain insight on the effect of rare events classified as extraordinary items, discontinued operation and special items, Fairfield, Sweeney and Yohn (1996) collected sample consist of 33.334 firm/year observation for non-financial firms available 1991 Compustats tapes. They found that dividing net income into income from continuing operation, income before special items and operating income as a common standard classification in body of income statement report have increased predictive ability of earnings.

This study replicate Fairfield, Sweeney and Yohn (1996) work. Unlike Fairfield, Sweeney and Yohn (1996) work, the study uses three level of earnings classification which are net income, income before specials items and operating income. Three level of earnings classification are adopted simply because events classified as extraordinary items are very rare in firms listed in Jakarta Stock Exchange.

II. THEORY AND HYPOTHESES DEVELOPMENT

2.1 Usefulness of Earnings Information

Rational investors require adequate information to estimate return on investment and the associated risk. By analyzing this information, they attempt to discover new high return investment choices and quickly dispose high-risk investment of low return. Wide variety of information traverse over the market everyday. As one of information, financial statement reflects the ability of firm management in handling resources entrusted to them. Investors need this information to assess and to predict firm cash flow generating capacity (SAK, 2000). By using this information, uncertainty can be reduced and quality of decision-making is therefore increased.

At the time decision-making process takes place, investors are conditioned by many things they believed. Those subjective personal beliefs grow from investor cumulative experiences such as training, education, and previous investment experience (Beaver, 1998). Furthermore, past financial statements, analyst reports and articles in magazine and newspapers also influence those beliefs.

2.2 Earnings Components Classification

Accounting literatures provide detail explanation of principles and procedures used to record economics events on income statement. The principles and procedure to follow is formalized into general standard known as generally accepted accounting standard. The standard is not only intended to make company income statements comparable but also to enhance the usefulness of reported earnings. However, accounting profession acknowledges that earnings figure alone could not communicate all information concerning firm profitability.

Measurement and estimation errors, integrity of matching concept and the occurrence of unusual events are few factors that could effect the credibility of reported earnings. Therefore, accounting literature recommend firms to present income statement based on its components, eg. Operating income, income before tax, income before extraordinary items and discontinued operation.

Earnings classification scheme assumes earnings components contain incremental information. Earnings components information is relevant to investors because they provide information that enable investor to asses the quality of earnings figure. Knowing that earnings increase come from normal operation enable investors to develop strong belief about firm future profitability. If the sources of increase comes from activity beyond firm normal course of action, the likelihood of increase persist to next period is small. Concept that describe this idea is known as earnings persistence concept.

2.3 Earnings Persistence

Earnings persistence measures the stability of earnings components and is first introduced by Kormendi and Lipe in 1987. According to Kormendi and Lipe, stock price movement is induced by the magnitude of expected earnings revision. Revision in expected earnings causes earnings surprise and if earnings surprise still persist in future period, stock price will move.

Based on earnings persistence concept, investors divide earnings figure into persistence and non-persistence components. Gains and losses from rare and random events such as disposal of fixed asset or loss from natural disaster must be recognized separately from components that come from normal operation. Rare and random events are usually uncontrollable and unpredictable.

Another way to perceive earnings quality is to classify earnings into transitory and permanent component. The concept is introduced by Beaver, Lambert and Morse (1980). They argue that earnings contain permanent component (ungarbled earnings) which has implication on stock price and transitory component which has no effect on stock price. If reported earnings contains many unusual events and is not expected to occur in near future, the ability of earnings to predict future earnings diminish. In this case, we can say that earnings figure contained many transitory components. Likewise, if reported earnings contained many permanent earnings, the ability of current earning to predict future earnings increases.

2.4 Reporting Irregular Items

Which components that should have been included in income statement have been long controversy (Kieso and Weygand, 1998). For example, would gain or loss from abnormal event be recognized in income statement or closed into retained earnings account ?

The proponents of current operating concept argue that earnings number is supposed to reflect only regular economic events or transaction that comes from normal operation. At opposite side, the proponents of all-inclusive concepts say hat income statement should include all economic transaction regardless of its irregularity. They argue that any gains and losses, related or unrelated to firm normal operation, contribute to firm long term profitability and therefore, any gains and losses from irregular events should be reported.

As of today, all-inclusive concept has been widely accepted. However, the lack of clear cut criteria that can be used as guidelines to identify regular and irregular events has prompted management to purposely choose reporting method that may alter their financial performance. This is done by reporting loss from abnormal events as irregular but report gains from abnormal events as a portion of normal earnings.

Accounting Principles Board (APB) No. 9 adopted all-inclusive concept and urges firms to use it in financial reporting practices. Irregular items are classified into five category: 1) Discontinued Operation. 2) Extraordinary items. 3) Unusual gains and losses. 4) Changes in Accounting principles. 5). Changes in estimates (Kieso and Weygand, 1998).

In income statement, separate category must be made to accommodate gains and losses from segment of a business which is still in operation and a business that has ceased to operate. Economic events classified as discontinued operations are increasing in practice. APB Opinion No. 30 provide guidance to help firms determine events that qualify discontinued operation.. To qualify as discontinued operation, assets, result of operations and activities of a segment of a business must be clearly distinguishable, physically and operationally from other assets, result of operations and activities of the entity.

Extraordinary items are defined as nonrecurring material items that differ significantly from entity’s typical business activities. According to APB Opinion No. 30, to qualify as extraordinary items, two category must be met. First, a transaction or event should possess high abnormality which are unrelated to firm normal activity. Second, a transaction or event should be of a type that would not be reasonably expected to recur in foreseeable future. Based on the category, the following gains and losses could not be classified as extraordinary items: (a). Write down or write off receivables, inventories, equipment leased to others and deferred research and development cost or other intangible assets. (b). Gains and losses from exchange or translation of foreign currencies, including those relating to major devaluations and revaluations. (c). Gains and losses on disposal of a segment of a business. (d). Other gains and losses from sale or abandonment of property, plant, or equipment used in the business. (e). Effects of a strike, including those against competitors and major suppliers. (f). Adjustments of accruals on long-term contracts.

Gains and losses classified as unusual gain and losses came from transactions which are unusual or infrequent but not both. If a transaction meet both criteria, it is classified as extraordinary items. Write downs of inventories, gains and losses on disposal of asset, gains and losses from foreign exchange fluctuation are few examples of unusual gain and losses or commonly known as specials items.

Changes in accounting method occur frequently in practice and recognized by including cumulative effects net of tax in currents year’s income statement. The effect of new accounting method adopted should be disclosed as separate items following extraordinary items. Changes in estimates should be disclosed in period where the changes occur or future period if the changes effect both period.

Many empirical studies have been conducted to seek the relation between earnings number and firm valuation. Earnings-return relation studies that exist today stem from Ball and Brown (1968) work. Large number of studies which involved different period, different methodology, and different stock market have found that earnings has information content (Lev and Ohlson, 1982).

In addition, there exists large body of research on information contained in earnings components. Gonedes (1975) reports that market returns are associated with unusual earnings components. Bowen (1981) found that investors put different values on operating and non-operating components of electric companies. Strong and Walker (1993) report that the strength of earnings-return relation can be enhanced through separating earnings into ordinary and unusual components. Beside investigating correlation between earnings components and stock return, the researchers also test earnings predictive ability. Finger (1994) found that current earnings can predict future earnings and cash flow.

Some studies divide financial statement in various ways and investigate the incremental information content of special items on financial statement. Ou and Penman (1989) developed model to predict the signs of earnings per share change. Their model used balance sheet and income statement items. Givoly and Hayn (1992) identified non-recurring items from annual financial statement and found that exclusion of these items from earnings figure increase ability of earnings to predict future earnings.

Fairfield, Sweeney and Yohn (1996) address the question of whether specific income statement classification yield incremental information for predicting future profitability. Future profitability is proxied by return on equity (ROE). They found that disaggregation of earnings into extraordinary, discontinued operation and specials items increase forecast accuracy.

Based on theoretical analysis and previous empirical result, the following hypotheses can be stated.

Hypothesis 1:ROEt-1 can predict ROE t.

Hypothesis 2: Separating ROEt-1 into income before special items and specials items improve forecast of ROE one year ahead .

Hypothesis 3: Separating ROEt-1 into operating income, special items and non operating income and tax improves forecast of ROE one year ahead

III. RESEARCH METHOD

3.1 Sample and Data

The empirical tests are conducted using firms listed in Jakarta Stock Exchange that meet following criteria:

1. Firms are listed in Jakarta Stock Exchange from 2000 through 2002 excluding insurance company, banks and financial institutions.

2. Release financial statement for four consecutive years starting 1999 through 2002 and accessible via website www,jsx.co.id.

3. Financial ratio are available in Indonesian Capital Market Directory.

4. Experience unusual gains and losses.

The first criteria is imposed because insurance companies, banks and financial institutions do not report cost of good sold and thus never reported operating income. Second criteria is intended to provide adequate period to form estimation and out of sample prediction. Third criteria is intended to collect sample firms as many as possible. Fourth criteria is imposed to highlight the partitioning of earnings into regular and irregular components. After screening process, there are 65 samples firms left for further analysis.

3.2 Research model

To make analysis easier to follow, decomposing earnings into its smaller components will be given special names. Hypotheses testing are conducted by using three regression model based on its aggregation level. The highest aggregation level is ROE (net income divided by equity at the beginning of the year). ROE is disaggregated into income before special items (IBSI) and Special items (SPECIAL). After first disaggregation, IBSI is further disaggregated into two variable; operating income (OPINC) and non-operating income and income tax (NOPTAX). The three regression model used to test the hypotheses are stated as follows:

Model 1. ROE[pic] = [pic] + [pic]ROE[pic] + [pic]

Model 2. ROE[pic] = [pic] + [pic]IBSI[pic] + [pic] SPECIALt-1 + [pic]

Model 3. ROE[pic]= [pic] + [pic]OPINC[pic] + [pic]NOPTAXt-1 + [pic]SPECIALt-1 +[pic]

Where :

ROE t : Return On equity for period t

IBSI[pic] : Income before special items for period t-1

OPINC t-1 : Operating Income for period t-1

SPECIAL t-i : Special items for period t-1

NOPTAX t-1 : Non operating income and income taxes period t-1

In this study, income is not disaggregated into extraordinary items and discontinued operation because the number of firms experiencing both events are very small. It is so small that makes it impossible to use regression analysis. Out of 65 sample firms, there are only 8 firms recorded extraordinary item and discontinued operation.

Regression model 1, model 2 and model 3 are used to test hypothesis 1, hypothesis 2 and hypothesis 3 respectively. To satisfy assumption of data normality and to reduce problem of heteroscedasticity, each variable in regression model are deflated by owner’s equity at the beginning of the year.

Cross-sectional time-series regression models are employed to investigate the contribution of different level of earnings disaggregation to forecast next year ROE. This can be accomplished by two ways. First, forming equation from coefficients regression and use the equation to predict next year ROE. The 1999 and 2000 data are used for estimating parameters. Second, compare out of sample forecast errors from each of three model to test the incremental predictive ability of income statement components. Out of sample test uses 2001 data to forecast ROE for 2002. To test whether forecast improvement are statistically significant across models, Wilcoxon signed rank test on paired differences is used.

IV. EMPIRICAL ANALYSIS

1. Descriptive Statistic

Table 4.1 summarize statistics on ROE and its components. ROE and IBSI have positive mean and SPECIAL has negative mean. The positive sign of ROE suggest that negative impact of transactions represented by SPECIAL variable on net income is not so strong. This also suggest that special items are relatively infrequent and most of earnings contain permanent component.

Table 4.1

Descriptive Statistic

[pic]

The negative sign of SPECIAL mean indicate that on average firms suffered more losses on irregular transaction. Observing financial statement of firms sample shows that most firms experienced gain or loss from exchange rate fluctuation and disposal of fixed asset. Almost all firms recorded gains on disposal of asset. This is not the case with exchange rate fluctuation. Most firms suffered great loss. But on average loss from exchange rate fluctuation exceed gain on disposal of asset. This is why SPECIAL has negative sign. Negative sign of NOPTAX is caused by the amount of tax paid is greater than non-operating income.

2. Diagnostic Test

Original least square requires data to be normally distributed, constant in variance and absence of autocorrelation among observations. This is of necessity in order to draw valid conclusion of the regression result and to make sound inferences. Table 4.2 present the result of diagnostic test. Kolmogorov-smirnov statistic is employed to

Table 4.2

Result of diagnostic test

|Model |Kolmg.-smirnov |Durbin-Watson |Glejser test |

|I |0.162 |1.779 |0.512 |

|II |0.185 |1.736 |0.582 |

|III |0.081 |1.859 |0.452 |

test data normality. Several observations from each model are deleted to satisfy data normality. Autocorrelation and heteroscedasticity are diagnosed through durbin-watson and glejser test.

3. Test of Hypotheses

To test predictive content of earnings components, three set forecast models are employed. Model 1 uses ROE[pic] as a single explanatory variable to predict ROE[pic]. Model 2 disaggregate ROE[pic]into two explanatory variables: income before special items (IBSI) and special items (SPECIAL). Model 3 retain special items as an explanatory variable and further disaggregate IBSI into two explanatory variable: operating income (OPINC) and non-operating income and tax (NOPTAX). Table 4.2 report regression result of model 1, model 2 and model 3.

Table 4.2

Regression of ROE[pic] on Disaggregation of ROE[pic]

Using 1999-2000 data

|MODEL |EXPLANATORY |[pic] |COEFF. |P-VALUE |AdjR2 |

| |VARIABLE | | | | |

|1 |ROEt-1 |11.02 |0.308 |0.000 |0.34 |

|2 |IBSI |11.27 |0.209 |0.000 |0.54 |

| |SPECIAL | |0.200 |0.000 | |

|3 |OPINC |5.67 |0.417 |0.000 |0.63 |

| |NOPTAX | |0.231 |0.000 | |

| |SPECIAL | |0.227 |0.000 | |

It can be clearly seen that correlation between explanatory variable and explained variable are very strong. P-value for each model is less than 0.01. Hypothesis 1 states that ROE t-1 can predict ROEt . The findings on table 4.2 support hypothesis 1.

The R2 for Model 1 that uses single explanatory variable is 0.34. The aggregation of ROE t-1 into IBSI and SPECIAL increases R2 to 0.54. There is an increase 0.20 compared to model 1. The same thing also occurred for model 3. Model 3 increase R2 to 0.63. There is 0.9 increase over model 2. If R2 for model 3 is compared to model 1, the increase is quiet large, that is 0.29. The findings suggest that earnings components have incremental information.

To further investigate earnings components predictive ability, equation gathered from model 1, model 2 and model 3 are used to predict ROE 2002 with 2001 data. For example, from model 1 we have equation ROEt = 11.02 + 0.308 ROEt-1. Using 2001 data, we substitute ROEt-1. The actual ROE is then compared to estimated ROE from 2002 data. The different between actual ROE and estimated ROE is forecast error. Because negative and positive forecast errors will cancel out, the forecast accuracy is examined by absolute forecast error. Table 4.3 reports descriptive statistics on absolute forecast error from each model and forecast improvement from disaggregating earnings.

Table 4.3

Descriptive Statistics of Forecast Errors

and Forecast Improvement

Panel A

| |Model 1 |Model 2 |Wilcoxon Signed Rank Test | t- test |

|Mean |22.59 |20.97 | |0.062 |

|Median |11.50 |11.00 |0.552 | |

|Standard Deviation |33.18 |28.78 | | |

Panel B

| |Model 2 |Model 3 |Wilcoxon Signed Rank Test | t- test |

|Mean |20.97 |19.53 | |0.302 |

|Median |11.00 |10.00 |0.003 | |

|Standard Deviation |28.78 |32.34 | | |

Panel C

| |Model 1 |Model 3 |Wilcoxon Signed Rank | Paired t- test |

| | | |Test | |

|Mean |22.59 |19.53 | |0.009 |

|Median |11.50 |10.00 |0.001 | |

|Standard Deviation |33.18 |32.34 | | |

To test whether forecast improvement are statistically significant across model, nonparametric Wilcoxon signed rank test on paired differences is used. Significance levels (asyimp.sig) are reported for medians. In addition to Wilcoxon signed rank test, paired t test is also reported for comparison. Panel A compares forecast errors of model 1 and model 2. Median and mean absolute forecast error for model 1 are 11.50 and 22.59 respectively and for model 2 are 11.00 and 20.97 respectively. Comparison of these two statistics suggest that model 2 is superior to model 1. However, Wilcoxon signed rank test asyimp.Sig of 0.552 indicate that there is no different between forecast error model 1 and model 2. Thus, the result reject hypothesis 2.

Panel B report statistics on forecast error of model 2 and model 3. Median and mean for Model 3 are smaller than model 2. Once again, this suggest that disaggregated model outperform the more aggregate model. Wilcoxon signed rank test Asyimp.sig of 0.003 suggest that the absolute forecast error for model 2 and model 3 are statistically different. Consequently, hypothesis 3 is accepted.

To gain more insight , model 1 and model 3 are compared. Panel C report statistics on forecast error of model 1 and model 3. As expected, median and mean of model 3 are much smaller than model 1. Wilcoxon signed rank test Asyimp.sig of 0.001 suggest that forecast error of the two models differ significantly. Taking as a whole, we can conclude that disaggregating ROE into two explanatory variable does not significantly increase predictive accuracy. This finding is rather confusing. It suggest that the current practice separating items from regular and irregular operation could not be supported. To make things complicated, further disaggregating income before specials items into smaller components do increase predictive accuracy. It seems that investors put considerable concern on operating income, amount of tax paid and nonoperating income greater than income before special items.

V. SUMMARY AND CONCLUSION

Accounting literatures provide detail explanation haw to measure and report wide variety of economic events into income statement. The explanation is intended to increase the usefulness of reported earning.

To make financial statement more informative and reliable, some contend that any gains and losses, directly or indirectly related to firm normal operation should be recognized in income statement report because they contribute to long term profitability. But some do not support the idea of recognizing irregular items in the body of income statement. At present, accounting profession require firms to report gains and losses from abnormal transaction apart from normal transaction.

The study investigate whether the current practice to separate earnings components based on its irregularity can be justified. Three regression model are employed: model 1 test earnings predictive ability without desegregation, model 2 test earnings predictive ability with two explanatory variable from disaggregation of model 1, model 3 further disaggregate model 2 explanatory variable.

The regression result of three models show that disaggregate ROE (net income devided by owner’s equity) into income before special items and specials items do not increase predictive accuracy. It may suggest that investor in jakarta stock exchange do not consider gains and losses from irregular transaction as important factors in estimating future earnings. Another possibility is the insignificant effect of irregular transaction on reported earnings that can adjust investors belief of future earnings.

Model 3 further disaggregate income before specials item into operating income and nonoperating income and tax. Comparison of absolute forecast errors of the two models indicate that predictive accuracy of model 3 outperform predictive accuracy of model 2. It suggest that investors consider operating income and amount of tax paid and nonoperating income are more important than income before specials items.

REFERENCE

Ball, R and Philip Brown, 1968, “An Empirical Evaluation of Accounting Income Numbers, “Journal of Accounting Research (autumn)

Barth, M., W. Beaver, and M. Wolfson, 1990, “Components of earnings and structure of bank share prices,” Financial analysts journal 46: 53-60

Beaver, W.H, 1998,”Financial Reporting : An Accounting Revolution, 3 rd Edition, Prentice hall.

Beaver, W.H., R. Lambert, and D. Morse, 1980,”The Information Content of Security

Price ,”Journal of Accounting and Economics 2: 3-28

Beaver, W.H., R. Lambert, and S. Ryan, 1987,”The Information Content of Security Price : A Second Look ,”Journal of Accounting and Economics : 438-458

Bowen, R, 1981, “The evaluation of earnings components in the electric utility industry, “ The Accounting Review 56: 1-22

Fairfield, P.M., R.J Sweeney and T.L Yohn, 1996, “Accounting classification and the predictive content of earnings, “ The Accounting Review: 337-353

Givoly, D., and C. Hayn, 1992, Transitory accounting items: Information content and earnings management, working paper, Northwestern University, Evanston, IL

Gonedes, N, 1975, “Risk, information and the effects of special items oc capital market equilibrium, “ Journal of Accounting Research 13: 220-256

Finger, C, 1994, “The ability of earnings to predict future earnings and cash flow, “ Journal of Accounting Research 32: 210-223

Ikatan Akuntan Indonesia, 2000, Standar Akuntansi Keuangan, Salemba Empat, Jakarta.

Kieso, D.and Jerry .W, 1998, Intermediate Accounting, 9-th Edition, John Wiley & Sons, Inc.

Kormendi, R and R. Lipe, 1987,” Earnings Innovation, Earnings Persistence and Stock Return, “ Journal Of Business : 323-345.

Lev, B and J A Ohlson, 1982, “Market based Empirical research in accounting: a review, interpretation and extension, “ Journal of Accounting Research.

Lipe, R. C, 1986, “The information contained in the components of earnings,” Journal of Accounting Research.

Ohlson, J.A., and S.H Penman, 1992, “Disaggregated accounting data as explanatory variables for returns, “ Journal of Accounting, Auditing and Finance 7: 553-573

Ou, J, and S. Penman, 1989, “Accounting measurement, price-earnings ratio, and the information content of security prices, “Journal of Accounting Research 27: 111-144

Syafriadi, H, 2000, “Kemampuan earnings dan arus kas dalam memprediksi earnings dan arus kas masa depan, “ Jurnal Bisnis dan Akuntansi.

Strong, N., and M. Walker, 1993, “The explanatory power of earnings for stock return, “ The Accounting Review 68: 385-399

( Accounting lecturer of Universitas Katolik Soegijapranata, Semarang, 0248441555

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download