The Role of Disaggregated Accounting Data in Detecting and Suppressing Earnings Management

Transcription

1 The Role of Disaggregated Aounting Data in Deteting and Suppressing Earnings Management Eli Amir ondon Business Shool ondon NW 4SA United Kingdom Eti Einhorn Tel Ai Uniersity Tel Ai Israel Itay Kama Tel Ai Uniersity Tel Ai Israel Preliminary draft January 8 We appreiate the helpful omments and suggestions of Gilad ine and Amir Zi.

2 The Role of Disaggregated Aounting Data in Deteting and Suppressing Earnings Management Astrat We study the role of disaggregated aounting data in the ontet of earnings management. Our analysis is rooted in the notion that different earnings omponents tend to e noisily proportional to eah other y their fundamental eonomi nature. We argue that these fundamental proportions eteen the omponents of earnings are likely to e distorted y reporting manipulations. The distortion in proportions that inole an inome item and an epense item is due to the managerial ish to ias inomes and epenses in opposite diretions. In other ases diersity in the managerial aility to manipulate different earnings omponents generates a reporting ias that distorts their original proportions. Users of finanial statements an thus utilize the disaggregated earnings data in order to identify deiations of the reported earnings omponents from their epeted fundamental proportions and therey detet iases in reporting. Being aare of the informational role of disaggregated aounting data in indiating earnings management firms may adopt a more autious approah in managing their earnings reports. Our analysis therefore suggests that disaggregated aounting information ia its role in deteting earnings management might sere as a poerful mehanism for suppressing misreporting inenties. Keyords: Information asymmetry; Aounting; Finanial reporting; Earnings management; Disaggregated aounting data; Finanial ratios. JE lassifiation: D8; G4; M4.

3 . Introdution Aounting earnings are typially the aggregation of many omponents some of hih hae to e dislosed ithin the inome statement hile others may e dislosed in notes to the finanial statements. Though the informational ontent of suh disaggregated aounting data is the fous of numerous studies in aounting e.g. ipe 986; Rayurn 986; Wilson 987; Barth Beaer and Wolfson 99; Barth 99; Ohlson and Penman 99; Amir 996; Sloan 996; Nissim and Penman ; Barth Clinh and Shiano it has een largely ignored y the etensie literature on earnings management. Etant studies onsider iases in reporting earnings mostly at the aggregated leel of the reported earnings e.g. Dye 988; Stein 989; Fisher and Verrehia ; Kirshenheiter and Melumad ; Fisher and Stoken 4; Eert and Wagenhofer 5; Guttman Kadan and Kandel 6. By analyzing reporting manipulations at the disaggregated leel of the earnings this study eplores the role that disaggregated aounting data play in the ontet of earnings management. Our analysis is rooted in the notion that earnings omponents tend to e proportional to eah other y their fundamental eonomi nature. Eamples inlude the ratios that usually eist eteen reenues and ertain epense items suh as the ost of goods sold marketing epenses and administration epenses. Influened as they are y hanges in the usiness enironment suh ratios are noisy in most situations. Their degree of noisiness depends on the nature of the earnings omponents inoled as ell as on industry and firm-speifi harateristis. We argue that reporting manipulations are likely to distort the fundamental proportions eteen the earnings omponents. The distortion in proportions that inole an inomes item and an epenses item is due to the managerial ish to ias inomes and epenses in opposite diretions. In other ases diersity in the managerial aility to manipulate different earnings omponents generates a reporting ias that distorts their original proportions.

4 The leeay of managers in reporting earnings ithin aounting onentions indeed seems to ary in its degree aross different omponents of earnings. Due to the inherent differenes in their nature some of the earnings omponents are more easily managed than others. It is generally aepted for eample that arual-ased items suh as ad dets and loss reseres depreiations and amortizations asset impairments are more easily managed than ash-ased items. igh diersity in the managerial disretion in reporting different earnings omponents is also typial of segment reporting here the onsolidated earnings of a firm are deomposed into earnings from its operations in arious usiness or geographial segments. We demonstrate the distorting impat of reporting manipulations on the original proportions eteen the different omponents of earnings and eplore its impliations for the aility of users of finanial statements to detet earnings management and for the propensity of managers to engage in earnings management. To do so e model a single-period reporting game here a manager of a pulily traded firm has the aility and inenties to manage the mandatory periodial earnings reported to apital market inestors. In our model the firm s manager hose ompensation is linked to the firm s stok prie hooses a reporting strategy ased on her rational epetations aout the market priing rule. The inestors in turn inoke their rational epetations regarding the manager s reporting strategy hen priing the firm in an effort to detet earnings manipulations. Folloing to Fisher and Verrehia e inorporate an eogenous noise in the model hih does not allo inestors to perfetly resole the reporting ias in priing the firm. Our departure from the traditional setting of earnings management is made in onsidering a disaggregated earnings report rather than an aggregate report. An analysis of the equilirium in our reporting game highlights the poer of disaggregated aounting data in deteting and suppressing reporting manipulations. In

5 equilirium reognizing the manager s inenties to ias inomes and epenses in opposite diretions and knoing that she faes different degrees of disretion in manipulating different omponents of the earnings report inestors rationally infer that earnings management is likely to distort the original proportions of the earnings omponents. Furthermore they an utilize the disaggregated earnings report in order to ompute the atual deiation of the reported earnings omponents from their fundamental eonomi proportions. Suh deiation may either stem from an eonomi noise that underlies the stohasti relationship eteen the earnings omponents or result from a reporting ias. It thus proides inestors ith a noisy indiator of earnings management hih enales them to more aurately ealuate the firm ased on the earnings report. The firm s manager ho is aare of inestors aility to utilize the reported disaggregated earnings data in order to imperfetly detet the ias in reporting eomes more relutant to manipulate the firm s earnings report in the first plae. Our analysis therefore demonstrates that disaggregated aounting information ia its role in indiating reporting manipulations might sere as an effetie mehanism that suppresses managerial misreporting inenties. Utilizing a omparatie statis analysis e further sho that disaggregated earnings information is more effetie in deteting and suppressing earnings management hen it onsists of items that are fundamentally more tightly proportional to eah other. The analysis also points to the merits of disaggregating earnings into pure-inomes and pureepenses omponents or alternatiely deomposing earnings into items that are ery dierse in the etent to hih their reporting an e managed. The paper proeeds as follos. The net setion models the reporting game. The equilirium in this game is deried and analyzed in Setion demonstrating the role of disaggregated aounting data in deteting and suppressing reporting manipulations. In an attempt to proide aounting standard setters ith useful guidane in seleting among alternatie disaggregating aounting proedures Setion 4 onsiders the sensitiity

6 of the equilirium outomes to the main modeling parameters. In Setion 5 e disuss the empirial impliations of our study. The final setion summarizes and offers onluding remarks. Proofs appear in the appendi.. Model In this setion e present a rational epetations model hih desries a singleperiod reporting game eteen a manager of a pulily traded firm and inestors in the apital market. The manager of the firm hose ompensation is linked to the prie at hih the firm is traded in the market eerises disretion oer the ostly ias in the earnings reported to the inestors. In designing the model e uilt on the earnings management setup of Fisher and Verrehia. We deiate from their model y onsidering a disaggregated earnings report rather than an aggregate earnings report. This allos us to flush out the important role of aounting disaggregation in the ontet of earnings management. The remainder of this setion details the parameters and assumptions underlying the model hih are all assumed to e ommon knoledge unless otherise indiated. We onsider a firm that is traded in a apital market for one period. The earnings that the firm yields during the gien period onstitute its equity alue. We model the firm s unertain earnings as a random ariale ~. To apture in the model the essene of earnings as an aounting aggregator e assume that the firm s earnings ~ is the sum of to omponents ~ and ~. That is ~ ~ ~. The earnings omponents ~ and ~ are assumed to e normally distriuted random ariales ith means μ and μ respetiely and arianes The assumption that the equity alue of the firm equals its earnings is a simplifying assumption that fits the single-period nature of the model. oeer the analysis an e generalized to the ase here the earnings measure is a noisy estimator of the firm s alue ithout qualitatiely affeting the results. 4

7 and respetiely here the oariane eteen them is. It follos therefore that the aggregate earnings ariale ~ is normally distriuted ith mean μ μ μ and ariane. The manager priately oseres the realization of the firm s earnings ~ as ell as the realizations and of the to earnings omponents ~ and ~ respetiely. Based on her priate information she issues an earnings report r r r hih inludes a report r of the aggregate earnings and to additional reports r and r of the to earnings omponents and respetiely. We assume that the manager eerises disretion in reporting hih allos her to ias the reported earnings. We further assume hoeer that the reporting ias is ostly for the manager. Biases and manipulations in reporting earnings an e assoiated ith a ariety of osts. When earnings management inoles the arrying out of ineffiient real transations it is assoiated ith the ost of distorting alue. In other ases it might e assoiated ith litigation osts reputation erosion osts osts that emerge from onflits ith auditors and audit ommittees and the osts of the antiipated mean-reersion in future reports. The normal distriution of the to ariales ~ and ~ implies that they an hae oth negatie and positie realizations. This allos us to apture earnings omponents that mi oth inomes and epenses and thus might e positie in some ases and negatie in other ases. Neertheless e an also apture in the model an inome-ased omponent y assuming that the orresponding ariale ~ ~ i N μi i has a positie mean μ i that eeeds y a large amount the standard deiation i so that the proaility of a negatie realization of ~ i is ery lose to zero. Similarly e an apture in the model an epense-ased omponent y assuming that the orresponding ariale ~ ~ i N μi i has a negatie mean μ i that its asolute alue eeeds y a large amount the standard deiation i so that the proaility of a positie realization of ~ i is ery lose to zero. This assumption is typial of earnings management models distinguishing them from heap talk models here misreporting is ostless e.g. Craford and Soel 98 and preluding aling equiliria in hih no information is oneyed. 5

8 As the managerial leeay in reporting might ary in its degree aross the different omponents of the reported earnings e allo the manager to ias oth earnings omponents ut at potentially different osts. In ommon ith the literature e.g. Fisher and Verrehia ; Dye and Sridhar 4; Guttman Kadan and Kandel 6 e assume a quadrati ost funtion. Unlike prior models hoeer our ost funtion is quadrati not only in the oerall reporting ias ut also in the reporting ias in eah of the earnings omponents. That is hen the manager oseres earnings of and reports r r r she ears a ost of here. The terms r r r > r and r in the ost funtion apture osts that are related to the reporting ias in a ertain omponent of the earnings suh as the osts of distorting alue due to ineffiient real transations that are arried out just for reporting purposes or the osts that arise due to onflits ith auditors and audit ommittees. The term r in the ost funtion aptures osts that are related to the oerall reporting ias ut are not sensitie to the ias in partiular earnings omponents suh as litigation osts or reputation erosion osts. The ratio / measures the etent to hih the seond earnings omponent is manageale relatie to the first earnings omponent. The higher the ratio / the more manageale is the seond omponent relatie to the first omponent. 4 The manager s ompensation is linked to the firm s stok prie and thus the manager makes her reporting deision in onjuntion ith her epetations aout the impat of the 4 The etreme ase here onerges to infinity and so is the ratio / desries situations here only the seond omponent of the earnings report is manageale. The opposite etreme ase here onerges to infinity and so the ratio / onerges to zero desries situations here only the first omponent of the earnings report is manageale. In oth these etreme ases the reporting of one of the earnings omponents must e truthful and this omponent an thus e ieed as a puli signal. 6

9 earnings report on the market prie of the firm. For any gien earnings report r r r e define P r r as the market prie of the firm. After osering the earnings omponents and the manager designs her optimal earnings report r r r in order to influene the rationally antiipated market prie of the firm P r r ut sujet to the ost r r r assoiated ith the reporting ias. We denote y the marginal enefit to the manager from shifting upard the market prie of the firm y one additional unit. ene the utility of the manager takes the form P r r r r r gien that she reports r r r after osering. Similarly to Fisher and Verrehia and susequent papers it is assumed that is priately knon only to the manager. The inestors on the other hand do not osere and they onsider it as the realization of some normally distriuted random eent ~ ith mean μ and ariane. That is the inestors fae some degree of unertainty aout the manager s reporting ojetie. 5 This unertainty on the part of inestors preludes fully reealing equiliria here the market is apale of perfetly aking out the reporting ias. Inestors are assumed to e rational and risk-neutral. Aordingly they set the firm s market prie equal to the firm s epeted alue onditional on all the aailale information. In partiular they use their epetations aout the manager s reporting strategy in an effort to detet the reporting ias and therey most effetiely utilize the information oneyed in the 5 Folloing Fisher and Verrehia unertainty on the part of inestors aout the reporting ojetie of managers is idely assumed in the dislosure literature e.g. Fisher and Stoken 4; Dye and Sridhar 4; Eert and Wagenhofer 5; Einhorn 7. Suh unertainty is modeled y assuming imperfet information of inestors ith respet to either the enefit to the manager from shifting the firm s market prie or the osts that she ears hen iasing the report. 7

10 earnings report in priing the firm. Gien that and are the realizations of the earnings omponents ~ and ~ respetiely and is the realization of the random ariale ~ e define as the reporting ias in the i th i earnings omponent. Although B i inestors do not osere and they rationally antiipate that the manager reports r B and r B hen she oseres ~ ~ and ~. Therefore the inremental information that inestors eliit from the manager s disaggregated earnings report r r r eyond the already aailale puli information is ~ B ~ ~ ~ r ~ B ~ ~ ~ r. and Figure proides a timeline depiting the sequene of eents in the model. At the eginning inestors estalish their prior eliefs aout the firm s unertain earnings ~ ~ ~ and aout the random eent ~. Then the manager priately oseres the realizations and of the earning omponents ~ and ~ respetiely and the realization of the random eent ~. Based on her priate information the manager issues the firm s disaggregated earnings report r r r here r B and r B. Susequently using the earnings report and all other aailale puli information inestors set the firm s market prie P r r. At the end of the period the firm s earnings ~ ~ ~ are realized and eome ommon knoledge. [FIGURE ] The model is therefore a single-period game ith to interrelated deisions made y to players the reporting deision of the firm s manager and the priing deision of the inestors. Equilirium in the model onsists of three funtions: B B R R that represent : the manager s strategy in iasing the to omponents of the earnings report and P : R R 8

11 that represents the priing rule applied y the inestors. In equilirium the manager hooses the reporting strategy B B ased on her rational epetations aout the market priing rule P hih in turn is ased on the inestors rational epetations regarding the manager s reporting strategy B B. Equilirium is formally defined as a etor of three funtions B B : R R P R R that satisfies to onditions. The first equilirium ondition : pertains to the manager s reporting strategy B B in iasing the disaggregated earnings report requiring for any R that B B arg ma P. The seond equilirium ondition desries the market priing rule P imposing P r r E [ ~ ~ B ~ ~ ~ r ~ B ~ ~ ~ r ] for any disaggregated earnings report r r R. We restrit the analysis to equiliria ith a linear priing rule. That is e impose the priing rule P r r to e a linear funtion of the to reported earnings omponents r and r. inear equiliria are ommonly assumed in the earnings management literature. 6 When omined ith a quadrati iasing ost funtion and a normal distriution of the firm s equity alue and earnings a linear priing rule enales a tratale analysis and yields equilirium outomes that an e easily haraterized and intuitiely eplained. As linearity restritions are ommonly made in empirial researh the assumption of a linear priing rule also allos us to link our analytial results to empirial findings and to dra ne preditions that are empirially testale. 6 An eeption is the analysis of Guttman Kadan and Kandel 6 hih eplains kinks and disontinuities in the distriution of the reported earnings y fousing on non-linear equiliria. 9

12 . Equilirium Analysis We start the analysis y inestigating the fundamental proportion eteen the to earnings omponents ~ and ~. For this purpose e define the oariane ratio ~ ~ / o ~ ~ o /. 7 The ratio an e either positie or negatie ut it must satisfy eause the ariane of the earnings ariale ~ is positie. The folloing oseration indiates that the normally distriuted random ariale ~ ~ denoted ~ is independent of the firm s earnings and alue ~. Oseration. et /. Then ~ ~ ~ is a normally distriuted ariale ith mean μ μ μ and ariane here o ~ ~. Sine the random ariale ~ ~ is alue-independent the measure an e ieed as the fundamental proportion eteen the earnings omponents ~ and ~. Using this interpretation the mean μ of the random ariale ~ ~ aptures the epeted deiation of the earnings omponents from their fundamental proportion hereas its ariane aptures the etent to hih the proportion is noisy. For eample in the speial ase here ~ represents the reenues and ~ represents the epenses measures the fundamental proportion eteen the ariale osts and the reenues μ measures the aerage leel of the fied osts and measures the etent of noise that underlies this stohasti relationship eteen the reenues and the epenses. 7 Sine ~ ar is stritly positie either the oariane o ~ ~ or the oariane o ~ ~ must differ from zero. So ithout loss of generality it is assumed that ~ ~ o ensuring that is alays ell-defined.

13 While the random ariales ~ ~ ~ N μ μ are the primities of the model it is more onenient to haraterize the equilirium properties in terms of the ommon distriution of the random ariales ~ ~ ~ ~ ~ ~ ~ N μ μ here / μ μ μ μ μ and μ. As μ μ and unequioally determine μ μ and the ariales ~ ~ ~ and ~ ~ ~ an sustitute for the ariales and in onstituting the primities of the model. 8 Aordingly hen applying the equilirium analysis e refer to μ μ and as primitie parameters so hanges in eah of them are made hile keeping the others intat. Though our results hold for any set of parameters μ μ and e espeially interested in the more desriptie situations here the epeted deiation μ of the earnings omponents from their fundamental proportion is loer than their epeted aggregate alue μ. In suh situations the sign of equals the sign of μ μ. 9 It follos thus that a negatie proportion eteen the to ariales ~ and ~ eists hen they are likely to hae opposite signs. This is typial of situations here one of these ariales represents inomes and the other represents epenses and thus equals the additie inerse of the epenses. A positie on the other hand aptures all other situations here the to earnings omponents are likely to share the same sign implying that oth of them represent inomes or that oth of them represent epenses. 8 Epliitly / μ / / / and /. μ μ μ μ μ 9 Assuming ithout loss of generality that > μ μ μ μ μ μ / share the same sign. it follos from the inequality μ < μ that and

14 Sine the manager ishes to ias inomes and epenses in opposite diretions the driing fore ehind our results for negatie alues of is the diersity in the manager s inenties ith respet to the diretion of the ias in the to earnings omponents. When is positie the manager seeks to ias the to earnings omponents in the same diretion so the diersity in the manager s aility to ias the to earnings omponents eomes the ruial fore that dries the results. Therefore e distinguish eteen three ases here the proportion is positie: i / ii < < / and iii > /. In the ase of / the fundamental proportion eteen the to earnings omponents oinides eatly ith the ratio / of their marginal iasing osts. That is the marginal ost of iasing the seond earnings omponent ~ y one additional unit equals the marginal ost of proportionally iasing the first earnings omponent ~ y the addition of units. We interpret this as implying that the to earnings omponents ~ and ~ are equally manageale. The ase < < / aptures situations here the first earnings omponent is less manageale than the seond earnings omponent. The opposite ours in the ase of > / here the first earnings omponent is more manageale than the seond earnings omponent. Throughout the analysis e fous on situations here / using the ase of / as a enhmark ase... The Benhmark ase of / We first onsider the edge ase of / here the fundamental proportion eteen the to earnings omponents is positie and oinides eatly ith the ratio / of It should e noted that the ase > / is symmetri to the ase < < / sujet to an ehange in the indees of the to earnings omponents.

15 their marginal iasing osts. The positie sign of implies that the to earnings omponents are likely to share the same sign and thus the manager ishes to ias oth of them in the same diretion. The fat that oinides eatly ith / implies that the manager also faes the same degree of leeay in manipulating the to earnings omponents. The ase of / is therefore the one and only ase here the reporting ias preseres the fundamental proportion eteen the to earnings omponents. It thus proides a natural point of referene sering in the analysis as a enhmark ase. Proposition estalishes the eistene and uniqueness of equilirium in the enhmark ase of / and haraterizes its form. Proposition. In the enhmark ase of / there eists a unique linear equilirium B B : R R P R R. The equilirium satisfies B / : β B / and P r r α β r r for any r r R here β and α β are salars suh that < β <. It follos from Proposition that the enhmark ase of / yields equilirium outomes that are onsistent ith traditional models of earnings management hih assume an aggregate earnings report. In partiular the equilirium outomes in the enhmark ase oinide eatly ith those of Fisher and Verrehia. The enhmark ase of / yields a unique linear equilirium ith an oerall reporting ias of B B / here. The sign of β the reporting ias is the same as the sign of. Its asolute alue is inreasing in the importane that the manager attahes to the market prie of the firm as aptured y the asolute alue of. The asolute alue of the reporting ias is also inreasing in the eight β assigned y the priing funtion P r r β r r α to the reported earnings r r r ut it is dereasing in the marginal iasing osts and.

16 Sine inestors do not osere the realization of the random eent ~ they annot preisely detet the reporting ias B B / and thus β annot deipher the manager s priate information from her report. Being unale to undo the reporting ias the market uses the earnings report r r r only as a noisy signal of the underlying earnings realization ~. Consequently the eight β that the enhmark equilirium priing funtion P r r α β r r assigns to the aggregate reported earnings r r r is loer than. More importantly in priing the firm the market plaes the same eight β on the to oserale omponents r and r of the reported earnings. That is hen / inestors only use the aggregate earnings datum r in priing the firm and find its deomposition into the omponents r and r redundant. Disaggregated data annot assist inestors in identifying the reporting ias hen / eause the equilirium reporting ias is proportionally emedded in the to omponents of the earnings report as refleted y B / B. Inestors rationally infer that the reporting ias / preseres the fundamental proportion eteen the to earnings omponents. They thus attriute any deiation of the reported earnings omponents from their epeted proportion to eonomi noise so suh deiation is not informatie to them in deteting the reporting ias. The enhmark equilirium presented in Proposition onstitutes an important referene point hen ealuating the poer of disaggregated earnings data in mitigating earnings management. This is eause the enhmark equilirium appears to e the unique linear equilirium that the model yields for any alue of under an aggregate reporting regime here the manager proides only the aggregate report r ithout detailing its omponents r and r. We formally state this result in the folloing orollary. 4

17 Corollary to Proposition. The enhmark equilirium B B : R R P R R : presented in Proposition is the unique linear equilirium for any alue of under an aggregate reporting regime here the earnings report inludes only the aggregate datum r and all other modeling assumptions are kept intat... The Case of / aing analyzed the enhmark ase of / e no turn to analyzing the more frequent situations here the fundamental proportion eteen the earnings omponents deiates from the ratio / of their marginal iasing osts. Sine the osts ratio / is positie y definition / implies that either the manager has different inenties regarding the diretion of the ias in the to earnings omponents hen is negatie or she has different ailities in iasing the to earnings omponents hen is positie. It appears that in oth situations the manager s optimal reporting ias distorts the fundamental proportion eteen the to earnings omponents hih makes the disaggregated earnings data useful in deteting and suppressing earnings management. While e annot rule out the eistene of multiple linear equiliria hen / e estalish in Proposition the eistene of a linear equilirium and haraterize the form of any suh equilirium. Proposition. In the ase of / there eists at least one linear equilirium B B : R R P R R. Any suh linear equilirium satisfies : B β β / B γ β / and γ γ P r α β r γ r for any r r R here and α r β γ are salars. The priing oeffiients β and γ satisfy β γ and β < β γ <. 5

18 Aording to Proposition hen / the equilirium priing funtion P r α β r γ r assigns different eights to the to omponents of the reported r earnings r and r. ene unlike the enhmark ase of / use the reported disaggregated earnings data and the priing funtion here inestors do not P r r α β r r assigns an idential eight of β to oth earnings omponents disaggregated earnings data eome informatie to inestors hen deiates from / enaling them to more aurately ealuate the firm. Proposition further suggests that the oerall reporting ias B B β / in situations here γ / is of loer magnitude relatie to the enhmark ias B B /. β The eplanation ehind the results of Proposition is the disproportional partition of the equilirium reporting ias aross the to earnings omponents hih distorts their fundamental proportion that is B / B. This enales inestors to utilize the reported disaggregated earnings data in reduing to some etent the eogenous noise ~ emedded in the reporting ias though they are still inapale of perfetly identifying and aking out the reporting ias B B β /. Based on the γ disaggregated earnings report they an ompute the etent r r to hih the reported earnings omponents r and r deiate from the fundamental proportion. The oserale measure r r is the sum of to unoserale alues: and r r. The former is the realization of the alue-independent ariale ~ ~ ~ hereas the When < < / the eight β of r in the priing funtion is higher than the eight γ of r. The opposite ours hen > /. In oth ases the market plaes more eight on the less manageale omponent of the earnings report. 6

19 latter is the realization of B ~ ~ ~ B ~ ~ ~ ~ γ β β /. γ Knoing this inestors reognize that any deiation of r r from μ ould e due to either the eonomi noise emedded in the stohasti relationship eteen the earnings omponents or the noise ~ emedded in the reporting ias r. By alloing them to diminish some of the noise ~ that underlies the reporting ias the oserale measure r r seres inestors as a noisy indiator of earnings management. ene in priing the firm inestors utilize the estimator r r in addition to the aggregate earnings measure r r r hih is the one and only releant signal in the enhmark ase. Aordingly the folloing orollary to Proposition offers an alternatie representation of the firm s market prie P r r r α β r γ as a funtion of r r and r r. Corollary to Proposition. In the ase of / the equilirium priing funtion P r α β r γ r presented in Proposition an e reritten as r P r r α δ r r τ r r for any r r R here δ β γ / and τ γ β /. The priing oeffiient δ satisfies < β < δ <. The priing oeffiient τ is negatie for < / and positie for > /. Sine inestors are more apale of deteting the reporting ias in situations here / relatie to the enhmark ase of / the aggregate earnings measure r r r eomes more informatie to them and thus the eight δ that they plae on it is higher than the enhmark eight β though it is still less than. To elaorate on the eight τ that inestors plae on the indiator r r of earnings management it is onenient to distinguish eteen negatie and positie alues of. When is negatie so that one of the earnings omponents represents inomes and 7

20 the other represents epenses the ish of the manager to ias inomes and epenses in opposite diretions generates a reporting ias that distorts the fundamental proportion eteen the to omponents. In this ase the indiator r r inreases dereases hen is positie negatie and the manager iases earnings upard donard. Therefore the priing funtion assigns a positie eight δ to the aggregate reported earnings r r r and a negatie eight τ to the indiator r r of earnings management. That is for any fied leel of aggregate reported earnings r r r the firm s market prie is negatiely related to the estimator r r that noisily indiates the etent of the reporting ias. When is positie the manager ishes to ias the to earnings omponents in the same diretion ut she an more easily ias one of the omponents relatie to the other. In the ase < < / the first earnings omponent is less manageale than the seond so inestors an infer that the reporting ias orks to inrease derease the alue of r r hen is positie negatie and the manager iases earnings upard donard. As a result the priing funtion assigns a positie eight δ to the aggregate reported earnings r r r and a negatie eight τ to the indiator r r of earnings management. The opposite ours in the ase of > /. ere the first earnings omponent is more manageale than the seond so inestors infer that the reporting ias orks to derease inrease the alue of r r hen is positie negatie and the manager iases earnings upard donard. In this ase therefore the priing funtion still assigns a positie eight δ to the aggregate reported earnings r r r ut no it plaes a positie eight τ on the indiator r r of earnings management. In all situations here / the manager ho reognizes the poer of disaggregated aounting data in enaling the market to imperfetly detet the reporting 8

21 ias eomes more relutant to manipulate the earnings report and thus her optimal hoie of the reporting ias is loer ompared to the enhmark ase. This is refleted in Proposition y the inequality β < β γ < hih implies that the equilirium oerall reporting ias B B β / is loer in magnitude than the γ enhmark reporting ias B B /. We thus onlude β that disaggregated aounting information ia its role in imperfetly deteting the reporting ias orks to suppress managerial misreporting inenties. We emphasize hoeer that despite their restriting impat on managerial aility to engage in earnings management disaggregated aounting data are not neessarily undesirale from the iepoint of managers. It has een ell estalished in the literature that managers an e orse off ith the option to ias their earnings reports see Stein 989. Managers might end up taking ostly ations to ias their reporting een hen they kno that they are unale to fool the market. Managers are trapped into suh ineffiient ehaior eause they take the market s onjetures as fied knoing that inestors ill suspet their report in any ase. Disaggregated aounting data might mitigate this prolem as they enale inestors to etter assess the reporting ias and therey allo managers to derease their ompelled engagement in ineffiient ations of earnings management. Additional interesting insights arise hen e onsider the reporting ias in eah of the earnings omponents separately. Aording to Proposition the reporting ias in the first earnings omponent is B β β / hereas the reporting ias in the γ seond earnings omponent is B γ β γ /. The term β / in B implies that the manager is more inlined to ias the first earnings omponent hen its oeffiient β in the priing funtion P r r α β r γ r is higher. Similarly 9

22 the term γ / in B implies that the manager is more inlined to ias the seond earnings omponent hen its oeffiient γ in the priing funtion is higher. The terms β / and β / in B and B respetiely reflet the γ γ propensity of the manager to shift the ias from one of the earnings omponents to the other relatie to the optimal iases under an aggregate earnings report. The manager inreases the ias in one of the earnings omponents y an amount of β γ / hile dereasing the ias in the other omponent y the same amount. This shift of the reporting ias from one omponent to the other is enefiial to the manager eause it fogs the deiation of the reported earnings omponents from their fundamental proportion as aptured y the indiator r and therey dereases the market s estimate of the reporting ias ithout atually r hanging the oerall reporting ias. 4. Comparatie Statis Analysis The analysis proided in the preious setion highlights the importane of disaggregated aounting data in mitigating reporting manipulations shedding light on the merits of epanded disaggregated dislosure in ases here the aounting system allos a relatiely high degree of disretion in reporting. In this setion e attempt to proide aounting poliy-makers ith useful guidane in seleting among alternatie disaggregating aounting proedures. We do so y utilizing a omparatie statis analysis to inestigate ho the eat struture of disaggregated aounting information affets its effetieness in deteting and suppressing earnings management. When multiple equiliria eist all the omparatie statis results presented in this setion remain the same ithin eah equilirium.

23 The omparatie statis analysis fouses on to equilirium outomes. The first is the ariane of the firm s true earnings ~ onditional on the disaggregated earnings report hih is used to ompare the effetieness of alternatie disaggregating proedures in deteting earnings management. The seond is the asolute alue of the oerall reporting ias hih is used to ompare the effetieness of alternatie disaggregating proedures in suppressing earnings management. Speifially a loer leel of the onditional ariane of ~ implies that the disaggregating proedure is more effetie in deteting earnings management. Similarly a loer asolute alue of the oerall reporting ias implies that the disaggregating proedure is more effetie in suppressing earnings management. We analyze the sensitiity of these to equilirium outomes to the modeling parameters and that depit important properties of the disaggregated data. We first analyze the sensitiity of the equilirium outomes to the fundamental proportion eteen the earnings omponents. We are ale to proide onlusie omparatie statis results only hen is positie. Fousing on positie alues of here the to earnings omponents are likely to share the same sign e espeially interested in analyzing ho the equilirium relates to the etent to hih the proportion differs from the ratio / of the iasing osts. It follos from Propositions and that disaggregated aounting information is effetie in mitigating earnings management only hen / and the managerial leeay in manipulating reporting is different for the arious earnings omponents. Proposition omplements this argument y shoing that the effetieness of disaggregated aounting information in this role monotonially inreases hen the departure of from / gros larger and the aility to manage the different earnings omponents eomes more dierse.

24 Proposition. Assume. Both the ariane of ~ onditional on the disaggregated earnings report and the asolute alue of the oerall reporting ias initially inrease in reahing a maimum at / and then derease in. The more departs from / the greater the diersity in the aility to manage the to earnings omponents and the larger the disproportion in their reporting ias. As a result the deiation r r of the reported earnings omponents from their fundamental proportion eomes more indiatie of the reporting ias. Consequently the market s aility to detet the reporting ias improes enhaning the informational quality of the earnings report. The manager eing aare that the market is etter ale to disern earnings management eomes more autious in manipulating the earnings report so the magnitude of the oerall reporting ias dereases. Proposition therefore points to the adantages of disaggregating aounting proedures that deompose earnings into omponents that are ery different in the etent to hih their reporting an e manipulated suh as pure ash-ased items and pure arualased items. While Proposition pertains only to positie alues of partiular attention should e also paid to situations here is negatie so that one of the earnings omponents represents inomes and the other represents epenses. Sine the osts ratio / is positie y definition any negatie must satisfies /. It follos thus from Proposition that disaggregated earnings report that deomposes earnings into inomes and epenses is alays effetie in mitigating earnings management regardless of the eat degree of leeay in managing the different earnings omponents. When is negatie een if the manager eerises the same degree of disretion in managing the to earnings omponents her ish to ias the inomes and epenses in opposite diretions generates a reporting ias that distorts the fundamental proportion eteen the to omponents. Under suh irumstanes the

25 oserale measure r r is alays indiatie aout the reporting ias regardless of the iasing osts and improing the informational quality of the disaggregated earnings report in priing the firm and reduing the managerial misreporting inenties. This sheds light on the merits of disaggregating aounting proedures that deompose earnings into pureinome and pure-epense omponents. oeer the dependene of the equilirium outomes on takes a rather ompliated shape hen is negatie hih does not enale us to proide lear-ut omparatie statis results for negatie alues of. This is due to the eistene of to fores that are at ork hen is negatie. The first is the diersity in the manager s inenties regarding to the diretion of the ias in the to omponents and the seond is the diersity in the manager s ailities to ias the to omponents. This is unlike the ase of a positie here the manager ishes to ias the to earnings omponents in the same diretion so the only fore at ork is the diersity in her aility to ias the to omponents. We no turn to analyzing the sensitiity of the equilirium outomes to the etent to hih the fundamental proportion eteen the earnings omponents ~ and ~ is noisy as aptured y. Proposition 4 demonstrates that hen dereases and the fundamental proportion eteen ~ and ~ eomes less noisy disaggregated aounting data eome more effetie in deteting and suppressing earnings management. Proposition 4. Assume /. Both the ariane of ~ onditional on the disaggregated earnings report and the asolute alue of the oerall reporting ias inrease in. When dereases and the fundamental proportion eteen the earnings omponents ~ and ~ eomes less noisy the deiation r r of the reported earnings omponents from their fundamental proportion proides a more aurate indiator of the reporting ias.

26 Consequently the aility of inestors to detet the reporting ias improes as dereases and the earnings report eomes more informatie to them in ealuating the firm. This redues the manager s inenties to manipulate the earnings report so the reporting ias dereases as dereases. It thus follos from Proposition 4 that the effetieness of disaggregated aounting information in deteting and suppressing earnings management is monotonially inreasing in the strength of the fundamental proportion eteen its omponents. This suggests that aounting standards setters should seek to deompose earnings into omponents that fundamentally ehiit the most stale proportions. 5. Empirial Impliations Assessing the usefulness of aounting earnings to inestors has een the fous of numerous empirial studies. Our study ontriutes to this line of researh y proiding a rational eplanation for the effet of earnings management on the assoiation eteen stok pries and aounting earnings. In partiular our analysis points out to a potential misspeifiation in empirial models ommonly used in the literature. More speifially it follos from Proposition that inluding the aggregate earnings measure r r r as a single eplanatory ariale of share pries in a regression model may e insuffiient. The orollary to Proposition points to the deiation r r of the earnings omponents from their epeted eonomi proportion as an additional eplanatory ariale that aptures the etent of the reporting ias. The equilirium prie funtion P r r α δ r r τ r therefore r suggests a methodologial refinement that might improe the speifiation of the relation eteen share pries and earnings. Proposition 4 further predits that the eplanatory poer of the ariale r r is espeially large hen the regression model is applied to firms that ehiit less noisy proportions eteen their earnings omponents. 4

27 In addition to the empirial preditions regarding the ay inestors utilize disaggregated earnings data in deteting earnings management and in priing firms our study proides testale empirial preditions ith respet to the managerial tendeny to engage in earnings management. Proposition 4 predits a loer leel of earnings management for firms that are haraterized y less noisy proportions eteen their earnings omponents. A loer leel of earnings management is also epeted in firms that proide more detailed dislosures on the omponents of earnings. Similarly our study predits a redution in earnings management folloing the issuane of an aounting standard that mandates additional dislosure on the omponents of earnings for eample mandating segment dislosures of sales and operating inome. Empirial eamination of our preditions requires the areful design of proies for the eonomi relationships eteen different earnings omponents. Finanial ratios seem natural andidates for this role. Partiular attention should e gien hoeer to the fat that finanial ratios are measured using reported earnings data rather than the underlying unoserale true earnings data. This prolem ould e mitigated y aeraging the finanial ratios of firms oer seeral reporting periods proided that deiations of the reported earnings items from the orresponding true earnings items mean-reert oer time. The theory deeloped in our study points to finanial ratios that inole only inome statement omponents suh as net profit margin operating profit margin gross profit margin and effetie ta rate. Neertheless finanial ratios ased on alane sheet line items suh as asset turnoer might e employed in empirial analysis as ell. This is so eause of the linkage eteen the inome statement and the alane sheet. In partiular any item that appears in the inome statement is assoiated ith a hange in alane sheet items. A similar argument an e applied to finanial ratios that inole items from the ash flo statement as these ratios apture relationships eteen ashased earnings omponents and arual-ased earnings omponents. 5

28 While most of our analytial results hae not een diretly tested there is empirial eidene onsistent ith our preditions. For eample in their reent study irst Koone and Venkataraman 7 find that disaggregation in management earnings foreasts enhanes their rediility. Consistent ith our predited priing rule e and Thiagarajan 99 argue that the deiation of the effetie ta rate measured as inome ta epense diided y pre-ta inome from the statutory ta rate is a sign of loer earnings quality. Also Ohlson and Penman 99 find that different omponents of earnings hae different aluation oeffiients ut these oeffiients onerge in magnitude as earnings are aumulated oer a suffiiently long period of time. They attriute this result to the diersity in the measurement errors emedded in different earnings omponents hih diminishes as the time indo eomes longer. Our study proides a slightly different interpretation of their result suggesting that diersity in the aility to manage the different earnings omponents might e the reason for their different aluation oeffiients. Oer a long period of time if earnings management is mean-reerting the differene in aluation oeffiients shrinks. In the ontet of our model diersity in the aility to manage the different earnings omponents might also e the reason for the empirial eidene doumenting that inestors plae a higher alue on the ash omponents of earnings than on arual-ased earnings omponents see Sloan Conluding Remarks This paper highlights aounting disaggregation as a poerful tool in deteting and suppressing earnings management. It suggests that epanded disaggregated aounting dislosure is espeially adantageous in ases here aounting standards allo management to selet among seeral reporting alternaties. In ases here the reporting disretion is sustantial restritions on the use of materiality onsiderations might also e reommended in order to aoid the ithholding of information that an assist in mitigating earnings 6

29 management. The analysis also offers guidane to aounting poliy-makers in seleting among alternatie aounting disaggregating proedures. In partiular it demonstrates that disaggregated earnings information is most effetie in mitigating earnings management hen its omponents are most tightly proportional to eah other due to their fundamental eonomi nature. The analysis further points to the merits of deomposing the earnings report into pureinome and pure-epense omponents or alternatiely deomposing it into omponents that are ery different in the etent to hih their reporting an e manipulated. While our study sheds light on onsiderations that ould e important to aounting standards setters in seleting the proper aounting disaggregation proedure these onsiderations oiously should e regarded in the ontet of other onsiderations that are not analyzed in this study. They should e ieed for eample in light of the differenes in the proprietary osts implied y different disaggregating aounting proedures. They should e also ieed in light of the differenes in the likelihood of different kinds of disaggregated data to e oluntarily dislosed y managers een in the asene of mandatory requirements see Einhorn 5. We also emphasize that the onlusions dran from our analysis only pertain to disaggregated information that details additie line items of an aggregate aounting measure suh as inomes and epenses that are aumulated into the net earnings measure. These onlusions do not neessarily hold hoeer ith respet to other types of detailed aounting data. Dye and Sridhar for eample onsider aounting information that details the sustitutale non-additie ojetie and sujetie measures on hih a single summary aounting datum is ased suh as the historial ost and the fair alue measures that underlie the familiar oer Cost or Market aounting measurement. They demonstrate irumstanes here this kind of detailed aounting data ould enhane managerial 7

30 misreporting inenties rather than suppress them eause more eight is attahed to sujetie measures. Our results are onsistent ith etant empirial eidene. In addition the results offer hitherto unidentified empirial preditions ith regard to the ay that inestors utilize disaggregated earnings data and ratio analysis in deteting earnings management hen priing firms. They also proide interesting testale preditions ith regard to the managerial tendeny to engage in earnings management. In a ompanion paper e proide empirial eidene that support our preditions. 8