Real Activities Earnings Management to Avoid】Losses

日 本 管 理 会 計 学 会 誌

管理会 計 学 2009年 第 17巻 第1号

Japanese



Firms

^，

Real



Activities



Earnings



Management



to

Avoid

】

Losses

Chine^．min KevinPan Abstract

This　paperinvest萱gatesif　Japanese　firms　that　report 　small 　positiveprofitsparticipate^in　eamings

management through 　realactivities manipulation to　avoid　reporting　losses．　The　evidence 　indicates

that　these　firmsupwardly 　mamage 　earnings　bycutting　discretionary^{expenses 　and}overproducing ． Meanwhile _，　the　findings　also　suggest 血 t　they　simultaneously 　record 　income −decreasingaccruals， suggesting　that　firms　reporting smaU 　positiveprofitshave　a　stronger　incentive　to　avoid 　reporting

losses．

KeyWords ：　Accruals_，　Cash　Flow　ffom　Operations（CFO _），Discretionary　Expenses，Production　Cests_， Earnings　Management

目 本 企

業^の 実 質

^{活 動 に よ る}

報

^告

利

^益

管

^理

潘 　

^{健 民}

〈論文 要 旨^〉

　

^{本 研究で は、 日本企業の 実質活 動に よる報 告 利益 管理の検 証 を 行 っ た。2000 年}以 降、 会

計 ^ビッ グ ・^{バ ン の} 影響^{に よ り}、 日

本

^企業は 変 更 さ れ た 会 計 基準^{へ の}対^{応 を}強^い られ^た。 本

研 究^は 、先 行研 究 と同様^に 、営業キャ ッ シ ュ ・_{フ U}・・一、発 生 項 目、裁量的支出 及 び 生 産^コ

ス トに 焦点を 当て て分 析 を 行 ^っ た。分^{析 の結 果、 対 象 企}業^は裁 量 的 支出^の 削減を行 う^の と 同時^に 、大量生産^に よ り^{コ ス} トを低減 さ^せ る^こ とに よ り、

利

^益

数値

^を

作

^{り上 げるこ と が}確 認^で き^た。 同時^に 、 対 象^企業^は、 発生項目 に ^よ り利 益 数 値を 引 き 下げる^こ と も確認^で き、

対 象企業が報告 利益管理 を 行 う動 機が さ らに強 まっ た と考えられ る。 本研 究^にお け る 分 析 結 果を総 合 する と、対 象企業^は実 質活動^{に よ} る報 告利 益 管理 を 行^っ て^い た 可 能 性 が あ る^の で は な^い か と考^えら^れる。

〈 キー_ワー _ド〉

発 生 項 目、営業^{キャ ッ シ ュ} ・_{フ ロ} ー

、 裁 量 的 支 出、 生 産^コ ス ト_、 報 告利 益管^理

2008 年3月 11 日　 受 付

2DO8年 7月30日　 受理

早 稲田 大学大学 院 商学研 究 科 博 士 後 期 課 程

SubmittedllMarch 2008 Accepted 　30　July²⁰⁰⁸

DoctoralStudent，GraduateSchoolofCornrnerce，

WasedaUniversity

The Japanese Association of Management Accounting

NII-Electronic Library Service

The JapaneseAssociation of Management Accounting

ewpaker\ eg^]7tsng¹e

1.

Introduction

This_paper investigatesJapanesefirms'^eamings managernent through real activities manipulation

(i.e.,

operational activities that affect cash flows).lnpartieular,^thispaper investigatesifJapanesefums thatreport smal1 positiveprofits^engage inreal activities management toavoid reponing losses.

Startingin the 1ate1990s,theJapanesebusinessenvironment eniered an era of accounting ^change. During thistime, new standards were adopted in the Japanese GenerallyAcceptedAccounting^Principies

(GAAP)

^as part^{of the refbrm} known as the

Japanese Accounting ^"Big Bang."With so many changes inaccounting rules wnhin such a 1imitedtime span, itwould be a valuable exercise to examine how thesenew accounting standards haveinfluencedaccounting practiceinJapanese^firms.'

Inthispaper;Iexamine ifJapanese firmsmanage earnings toavoid reporting losses

through real activities earnings management. The existing literatureon eamings management shows that managers havea strong incentiveito arvoid negative earnings surprises otiand exceed specific eamings thresholds.2

Several_paststudies providesubstantial evidence thatJapanesemanagers engage in

eamings management. For example, Shuto

(2000)

^{and Suda and Shuto}

(2001)

provide

evidence on accruals management and Suda and Shuto

(2005)

present^{evidence of}

earnings managemeni combined withearnings distributionand accruals management.

SimilarIMHerrmann, Inoue,and Thomas

(2003)

^{linkthe eamings management}

activities of Japanese firms with the sales of fixed assets and marketable securities,

whileShuto

(2007)

establishes a connection with executive compensation schemes. In

other extensions, Yamashitaand Otogawa

(2e08)

investigateifJapanesefirmsmanaged earnings inresponse tothetaxreductions of thelate1990sand Onuma

(2004)

^analyzed

the^{use of valuation allowances} fordeferredtaxes for^{earnings management inJapanese}

comrnercia! banks.However,the evidence on earnings management in Japan!argely

ends inthe1999fisealyear3^{with only a} few studies, such as Onuma

(2004)

^{and Pan}

(2006),

^{continuing beyondthe 2000fiscal}year.

Thispaper^{conuibutes to} the literatureby linkingthe eamings management of

Japanesefirmsto real activities manipulation. Untilnovv) research on earnings management activities has focusedon accruals management with only a few studies extending theirscope to specific earnings management vehicles. Thispaperisalsoone of thefew studies thatexamine earnings management activities of Japanese firmsafter

the²⁰⁰⁰ fiscal_year.

2.Earnings^and RealActivitiesManagement

2.1Managing tolexceedEarnings Tlhresholds

One of themainstream research rnethods used to detectearnings management isto

depictthefrequencydistribution^{of earnings} byfirm-yearto indicateiffirmexecutives manage earnings to avoid reporting losses.More specifically, when plottingthe frequency^of eamings intoearnings intervals,researchers often finda discontinuityin thezero earnings interval.Thisdiscontinuitythen provides ^{evidence of the} firms'

attempts to manage earnings to exceed the zero eafr}ingsthreshold^to avoid ^reponing

losses.

^{The particularfindingof an earnings} distributiondiscontinuitywas firstdocumented byHayn

(1995)

^and Burgstahler^and Dichev

(1997).

^{Thismethod hasalsobeenapplied}

4

NII-Electronic

H"fi\o=pttaent:l6WkiUesrgN

by Degeorge,Pateland Zeckhauser

(1999),

^{Dechow) Richardsonand Tuna}

(2003),

Beave4McNicholsand Nelson

(2007),

^Xue

(2004),

^{and Durtschiand Easton}

(2005).

^In

addition, several studies on earnings management activitiesinJapanesefirmsemploy

similar methods, includingShutoand Suda

(2001),

^{Suda and Shuto}

(2005),

^{and Pan}

<2eo6).

2,2AccruaZs^{Management and Real}ActivitiesManagement

Othertharitheshape of thefrequencydistributionof earnings, theexisting literature

also presents^{models and methods tocapture earnings management activities} byfirms.A Jones-typeaccruals model

(Jones,

^{1991)or themodified Jonesaccruals model}

(Dechovvl

Sloanand Sweeney, 1995)has dominatedthe detectionof earnings management activities duringthe lastdecade.In addition to their^role as a detectorof earnings management, increases!decreasesinaccruals are also presented^{asevidence of earnings}

4 management.

More recently, real activities eamings management hasreceived increasingattention

from_younger researchers. InBalland Shivakumar

(2e06),

^{itisargtied thata lessnoisy}

earnings variable isobtained by adding working capital accmals tothecash flowsfrom

operations

(CFO).

^{In short, accruals models are used to detecteamings management}

activities withno directcczshflow conseguences

(Roychowdhury,

^2006).

Conversely, ^{real activities earnings management} focuseson detectingearnings

management ^{activities with} directcash

flow

conseguences. The literatureon real activities earnings management detailsthefbllowingtransactions.First,cutting research and development

(R&D)

^{expenses, or selling,} general^and administrative

(SG&A)

expenditures to increaseincome.Second,overproducing to reduce the cost of goods

sold to increaseincome.[[hird,cutting prices!offeringpricediscountsto boestsales in

thecurrent period.FinallMselling fixedassets withumealized heldinggains^or losses.5 In a seminal studM Roychowdhury

(2006)

proposes ^and findsevidence thatfirms

manage eamings through real activitiesmanipulation. Followingthe procedures developedinRoychowdhury

(2006),

^Gumy

(2005)

^{analyzes the} consequenQes of ^real activities ^{eanrings management} and findsthatmanaging eamings through ,real^activities significantly hinder the futureperformance^of firms.Zang

(2007)

^{also studies the}

trade-off betweenaccruals management ^{and real} activities management and findsthat

managers opportunistically switch fromaccruals toreal activities and vice versa. In a recent paper,Cohen, Dey and Lys

(2008)

investigateand findthat accrual-based earnings management declinessignificantly after tihepassage Sarbanes-OxleyAct

(SOX)

^{in2002, while real activities earnings management increases}sigriificantly after

thepassage^of SOX.

3.RealActivitiesHypotheses 3.JRealActivities^{Earnings Management}

A ^{number of studies} harvediscussedthepossibility^{thatmanagerial} interventioninthe reponing process^{can occur via not only accounting estimates and methods,} butalso operational decisions.6Shortlyafter Healyand Wahlen

(1999),

^{Dechow and Skinner}

(2000)

propose thattheacceleration of sales, alterations inshipment ^{schedules and} the

delayof R&D and maintenance expenditures are some of the earnings management methods ayailable to ^managers. Reychowdhury

(2006)

investigatesseveral other real

The Japanese Association of Management Accounting

NII-Electronic Library Service

The JapaneseAssociation of Management Accounting

Pmpkst4 th¹⁷igac1g

activities management methods, including_price discouritsand the reduction of

discretionaryexpenses, as possibly^{optimal actions giventhe economic} circumstances ofthe firm.

InRoychowdnury

(2006),

^{two features}characterize ^real activitiesmanipulation:

(al

departures^{from normal} operational practices-these.departuresare, by themselves, potentiallydetrimentalto firm value,7 and

(b)

^{a desireto mislead at leastsorne}

stakeholders intobelievingcertain financialreponing goalshavebeenmet inthenorrnal course ofoperations.8 Realactivities management affbcts cash fiowsand, insome cases, accruals. Zang

(2006)

^{also argues thatreal activities and accruals management are}

substitutes. Hence,thepresent^analysis alsoinvestigatesaccruals management.

This _paper fbcuseson the fbllowingreal activities earnings management manipulation methods documentedinthe literatureand theireffects on excess CFO:

1.offering pricediscountstoincreasesales,

2.reducing discretionaryexpenses, and

3.overproducing todecreasethe costs ofgoods sold.

3.2OLt7ZiringPriceDiscountstoincreaseSales

Salesmanipulation is_generallydefined^as a managerial attempt to temporarily

increasesales beforethe fiscalyear-end.^Itisargued that by offering price^discounts,

managers can rush inmore sales volume befbrethe_year-end.However,managers likely

undertake such actions even inthenormal course of business.Whethertheseaetivities are more extensive thannormal among firrnstryingtomeet earnings targetsisthenan empirical question.^{Also, itisexpected that the additional sales volumes created} by pricediscounts^{will reverse} inthe next fiscalyear^{once the} firmstops offering such

discounts.

3.3ReducingDiscretionar:),Ebepenses

One effective ^way toincrease^earnings is^redueing discretionary^expenses. Following Roychowdhury

(2006),

^this paper investigatesif Japanesefirms make use･of

discretionaryexpenses to manage eamings by examining advertising, research and

development

(R&D)

^{and selling,} general^and administrative

(SG&A)

^expenses.

[Ibgether,advenising and R&D typically comprise a Iargeshare of discretionary

expenses and managers can temporarily increaseearnings through theirreduction. For

example, Mande,File,and Kwak (2000)provide^evidence thatJapanese firmsinthe 1990ssmoothed incomethroughtheadjustmerrt ofR&D expenses.

Otherthanadvertising and R&D expenses, SG&A usually includes,amongst other things, employee training, maintenance and travelingexpenses. These items are

generallyinthe fbrm of cash. Reducing discretionaryexpenditures haspositive^eflects

on excess operational cash fiowsinthecurrent period,butonly atthe expense of lower

cash inflowsinthe future.Atthe same time, when reducing discretionaryexpenditures such advertising and R&D, thelong-termcompetitiveness and profitability^{of the}firrnis

adversely affected.

Roychowdhury

(2006)

^expresses discretionary^{expenses as} advenising expense +

R&D expense + SG&A expense. SG&A expense inRoychowdhury

(2006)

^istaken

from COMPUSTAT data item#189. The datadefinitionin COMPUSTAT #189 indicatesthatthecalculation of SG&A alreadyincludesadvertising and R&D. Inother

6

E2pAmea)2Riswta:*4wsijf]esgpt

words, itis^very _possiblethatRoychowdhury

(2006)

^doublecounts advertising and

R&D expenses indiscretionaryexpenses. ^'Ib avoid this,thispaper ^employs SG&A

expense ^as the discretionary^expense.

3,4 Overproducing toDecnease theCost_qf'GoocisSbld

Recalltheprinciple^of profit-maximizationfromany introductoryeconomics text:

firrnsmaximize profits^at thepointwhere marginal costs equal marginal revenues.

Roychowdhury

(2006)

^{argues that}"...manufacturing firmscan overproduce

(produce

more goods ^{than necessary tomeet expected} demand) to manage eamings upwards

(p.

340."

As the_production^volume of goods increases,^variable costs also increase.Howeveg

when production^volumes increase,fixedcosts do not change and the per-unitfixed

overhead costs decline.As longas theincreaseinmarginal costs per ^unit doesnot exceed the decreaseinfixedcosts per^{unit, totaleost} per^unit declines.Withlowertotal

costs per^{unig the reported} CostsofGoods Sold

(COGS)

^{are lower,and firmsare able to}

report betteroperating margins.

Thispaperdefinesproduction^{costs as} thesum of Costsof Goods Soidand the

ChangeinInventoryduringa period.Withthisdefinition,ifthefrrmdecidestomanage earnings through overproduction, thefirmwili need tobeartheproduction^and holding

costs of the'overproduced goods thatthefirmisunable to sell beforethe end of the fiscal_yeai

Nonmanufacturingfirmsalso have _production costs within the scope of this definition.Itispossible,fbrexample, thatnonmanufacturing firmsreceive price discountsfrom bulkpurchases.Ifthe firmreceives pricediscounts,itreports lower COGS and higheroperating margins. At the sarne time, thefirmalso incursthe

inventoryholdingcosts ofthe overpurchased goods thatthefirmisunal)le tosell befbre the fiscal_year-end.Therefore,fora given ^sales level,cash flowsfrom operations a!}e expected tobelowerthan^normal.

3,5^.Eormal

thpotheses

Similarto Roychowdhury

(2006),

^{the discussioninthis section suggests that the}

relation betweenreal activities and sales isas fbllows:offering pricediscountsand overproducing leadstohigher_productioncosts relative tosales volume, while reducing

discretionaryexpenditures leadsto lowerdiscretionaryexpenses relative to sales.

Conversely,the^relation between^{real activities and sales} isas fbllows:pricediscounts,

channel stuMng, and overproducing may simultaneously decreaseCFO, while reducing

discretionaryexpenditures may increaseCFO.

FollowingRoychowchury

(2006),

^{theformalhypothesesare as fbllows:}

^,Fbrmal

Hlpothesis for

OveT:pn)duction:Aftercontrolling forthelevel^of

^sales, targetfirrnsshould haveunusually high_production^costs.

^jFbrmal

H)tpothesis for

Discretionan}Jlixpenses:Aftercontrolling fbrthe

levelof sales, targetfirmsshould have either or bothunusually low cash

fiowfromoperations

(CFO)

^{and low}discretionaryexpenses.

The Japanese Association of Management Accounting

NII-Electronic Library Service

The JapaneseAssociation of Management Accounting

eNkat\ ag¹⁷iges¹e

4.Data and VlairiableDefinitions 4,1Dataand Sbmple

Financialand stock pricedataare obtained fromthe NikkeiEconomic Electronic Datal)ankSystem-FinancialQUEST

(hereafteg

^{NEEDS-FQ) on-line databaseservice.}

Allfirmsⁱⁿthesample ^are listedinthefirstand ^second section ofthe [fbkyoSecurities Exchange.^Allfirmsincludedinthe sample are also March fiscalyear-endfirms.The dataperiod^eovers theperiodfromfiscalyear2eOOto fiscalyear2004.Tb ensure data

consistency, firms that change theirfiscalyear-end duringthe sample period ^are

removed. Inaddition,al1firmsin

the

^{sample arerequired tohave}consecutive firm-year datathroughout the entire sample period.^{'Ibcalculate market} capitalization, firm-years

are excluded ifthestock pricedataismissing.

The Niki(eiindustryclassification codes are used to identifyif a firm is a manufacturing or nonrnanufacturing firm.Inaddition, thetwo-digitNikkeimedium

industryclassification codes are used to identifya firm'sindustry affTiliation.

Furthermore,firmsinregulated industries

(e.g.,

utilities),banks,and other financiai institirtionsare excluded fromthesample.

The regressions inthepaper^{are estimated} fbrevery industryfbrevery year.Any industrywithlessthan 15 observations ina yearisexcluded fromthesample. As a result, the fu11sample consists of 8,205observations from 24 industriesand l,641

individualfirms^over theperiod2000 to2004.

4.² Choice

ofSbaling

^jFbctor

The ^{choice of scaling} factoris intendedto incorporatethe uniqueness of the

accounting environment inJapan.Inthis paper,^{all variables are} deflatedby market capitalization-unless specified otherwise.

The choice betweentotalassets or market capitalization as the scaling factorisa controversial issuein_theaccounting literature.Inthecase ofJapanese firms,themarket value of equity isa betterdeflatorthan thebook ^value of totalassets. Thisisbecause duete a deep-rootedhistoricalcost accounting convention inJapan,thebookvalue of

landheldby long-establishedfirrnsmay be undervalued.9 These unrealized holding gains hiddenwithin landprices^{are very comrnon arnong} long-establishedJapanese firms.By using market capitalization, the expectations of market participants^{to some}

extent

justify

^{theunrealized holding}gainshiddeninthefirm'stotalassets.

43.Expectation Modelsfor ]lixcess lhlues

Following^the proceduresfbundinthe literaturetodetect^real activities management

(Roychowdhury,

^{2006; Gunny, 2006;Zang, 2006),Ifirstestimate the `normal level}

value of thevariable' and thendefinethedeviationbetweenthis^and the^realized value ofvariable as the^`ex ¢ess value ofthe variable'.

[foprovide^specific detaiion the^empirical procedure,^we give^a detaileddescription

of how we derivethe^{excess cash} flowfromoperations

(CFO).

^{Ifirstuse themodel}

presentedinDechowl Kothariand Watts

(1998)

^{to estimate thenormal CFO fbreach}

firm-year.Then ^using the sales in thisfirm-yearand the estimated mean relatien

betweenCFO and sales forthecorresponding industry-year,Idefinethenormal CFO.

Finally,Idefinethedeviationsbetweenthe estimated normal CFO and the realized CFO

as ^`excess CFO.'

Four ^{models are selected} from Dechow} Kothariand Watts

(1998)

^{to definethe}

8

HJShiSOscpttaste:X4sue*ilasewge

excess cash flowsfrom operations, excess production ^{costs, excess} discretionary

expenses, and excess accmals. All fourof thesemodels are represented ds a 1inear functionof salesor thechange insales.

^The firstrnodel definesthe^{excess vaiue ofcash} flowfromoperations. As discussed,

theTChr2S;:'oSedCe?2.sneakSirneesVedMaOsdfebiilgeXs{?gressedoneveryindustryeveryyear.

CFq ⁼ qo^{+ai *S, +a2 *AEL +6,}

(1)

where

CFq ^:= cash flowsfromoperations in_period t,

S

⁼⁼ salesgeneratedinperiod t,and

AS,⁼ changes insales inperiod^t.

The ^{regression rnodel assumes a} linearrelation between the cash fiow from

operations and sales and change insales inthe current period.^{The excess} CFO or excess cash .17ows.17om operations^are definedaS thedeviationsbetweenthe^realized CFO and theestimated CFO.

Ihe ^treatment inregression

(1)

^{isalsoapplied to} regressions

(2)

^and

(3)

^{to define}

exces,s productipnQosts^{and excess} discretionaryexpenses. Again, the cross-sectional regressions are estimated fbrevery industryand year:

PRODL ⁼

1%

⁺ 13i^"S)+

13i

^{fAS,+ ,CL3" ASr-i+e}

Disex,⁼ 16⁺ 7i^*

a-i

^{+ et}

(2)(3)

where

PR04 ⁼⁼ _productioncosts, definedas CostofGoods Sold⁺ ChangeinInventory, Disex,⁼ discretionaryexpenses, definedas SG&A,

S,-,⁼ sales generatedinthepriorperiod t-1, and

AS,-,⁼⁼ changes in^sales inthepriorperiod t-1.

The excess .PROD, or excess production^{costs, and excess} Disex,or excess

discretionatyexpenses, are aisodefinedas thedeviationsbetweentherealized PROD

and the estimated PROD, and ^realized Disex and theestimated Disexrespectively.

Regression

(2)

^{assumes a linearrelation between} production ^costs between ^sales,

change insales and change insales inpreviousperiod.^Regression

(3)

^{assumes a linear}

relation betweendiscretionaryexpenses and sales. Productioncosts hereare definedas

the^{sum of the costs} of goods ^{sold and} thechange ininventory.Discretionaryexpenses

hereare definedas selling, general^and admmistrative expenses

(SG&A).

^{Finally,accruals are} introducedfbrthepuipose^{of the comparison} betweenacc;uals management and real management. ^`Normal' accruals are estimated byusing themodel

inJones

(1991):

The Japanese Association of Management Accounting

NII-Electronic Library Service

The 　Japanese 　Assooiation 　^of 　Management 　Aooounting

管 理 会 計 学 第 17巻 第 1号

Tab 且e　l−_{Excess 瓶} i滋）le　Expectation　Models

PanelA：Exectation ModelsEstirnated^withCross−sectional Re essions

CFO ， PROD _， DiseXt AccrualSt

Intercept StSt

＿1

△LS_，

△St．_1PPE

，

＊ ＊ 串0．084817

　（1Ll8291 _）

＊ ＊ ＊O，038102 　（25．45533_）

一_〇．007335

（−₀．36414_）

＊ ＊ ＊騨α24238

　 （−₂₁．4619_）

＊ ＊ 噌O．834371 　^（247，7071_）

　　■0．06204 　 （■₁．47825_）

＊ ＊ ＊0．098254 　^（3．_{054039 ）}

＊ ￥＊O．240766 　（2850354 _）

＊＊ ＊O．140022 　（75．96929_）

＊0．⁰¹³¹³⁸

（1．658283_）

　0，020035

（1．_{022601 ）}

＊ ＊ ＊凾O．12137 　（−415501）

＃Obv ． Ad ’

ustedR2

　82050

．2965

　82050

．9624

　82050

．5914

　82050

．3526

＊，^{串 寧}，^{and ＊串 寧indioate　variabl 已}signiflcantat10％，^5％，^{and　1％ level}respectively ．　The　sample includes　24　industries　and　120 industry−_years　over 　2000　to　2004 五sc径1−_years．　Numbers　presentedin　the　table　are　mじan　coe 伍 cients　and　mean 　a｛賑^usted R−_{squared 　estimated 　by　each 　industTy　and　by　each 　ycaτ}，　The　t−_{statistics　reported 　in　the　paren山 esis　ar巳calculated} 　from　standard errors　of　the　mean 　acress 　al1　industry−_years．

PamelB^：DescritiveStatisticsofVariables toRun theExectation Models RealNumbers _（inMillionYen）

Mean 　 　 　 　Median

ScaledNumbers

Mean Median

CFO

ProductionCestS DiscretionaryExpenses

AccruaIs

13，555 176，918 43_，351．85

一₅，³⁶⁰

2，209 34_，176 7_，688

一487

0，198718 4，144555 0．7424 一_〇．16483

0．ユ38653 2．375578 0．529861

，0．07547

Sales 229，140 49．217 5．040252 3．097302

Change　in　Sales ・8，004 一₁₃ 一_〇．05138 0，020469

Property，Plantand　Equipment MarketvalueefEquiry

53，586122

，437

8，⁷²⁰¹⁴

，⁸⁴⁹

0．9653 0．563489

＃Obvs8，205

Definition　ofVariables^：（］FO^二Cash且ow 丘om 　operations ；Accruals：【Eamings　befbre　extraordinary 　items−CFO］；Praduction Costs_（PRα ））^：【Costof　GoodsSold÷ Changeininventery_】；DiscretionatJyErpenses（Disex）：［Selling，^{General　and} Administrative　Expenses_］；∫：Sales_；S，一，：Sa工es　in　PriorYear_；A^{∫；Change  Sales； △＆ノ：Change　tn　Sales　in　Prior　Year}；PPE ：

Propcrり〜Plant跏d　Eqロipment；ScalingFactor^：Market　Capitalization．

Accruals_、 ^＝

_％

^＋

4

^{＊△s，＋}

ろ

^{＊PPE ，＋et} （4_）

where

Accruals_， ^＝ accruals _，　definedas　earnings 　befbreextraordinaryitemsm 血 usCFO ，　and

PPE _，^＝ propert）^r，plant^{and 　equipment ．} 10

N工 工一_{Eleotronio 　Library}

Hptdi\`D$qtson:l6maftIUasgve

lflable^{2 -Descritive}StatisticsofVat-iahles

O<PRETAX<e,03 (650>

Mean Median

(t-stats) (z-stats

RestofSampte (7555)

Mean Median

Matket Valueof Equity

(millien^JPY)

MarkettoBookRatio TbtalAsset

(million^JPY)

Sales

(million^JPY)

EarningsbeforeExtraordinary '_Items(million^JPY])

Earningsafter Extraordinary Items(mi]Iien^JPY)

NetExtraordinaryItems

(mi11ion^JPY)

Cash Flow from Operations

(mi11ion^JPY)

Accruals

(million^JPY)

Sales/

MarketCapitalization

EBEIt '

MarketCapitalization PrctaxInceme /

Matket Capitalization CFO1

MarketCapitalization Accruals1

MarketCapitalization ProductionCostsf

MarketCapitalizatien Disex/

MarketCapitalization IriventoryTurnoverRatio Receival)leTurnoverRatio

***295,268

(7.96)

***4.275569

(6.82)

^*30,189

(-1.84)

^.292,808 (1.85)

7,122 (O.65)

4,481

(-O.56)

-2,641

(-O.12)

^**19,384 (2.69)

***-12,263

(-4.82)

**i2,851591

(-8.32)

***O.e50634

(-8.72)

O,Ol6g82 (-O.30)

***O.122431

(-3.83)

-O.0718

(-O.95)

***2.286623

(-7.67)

O,485651 (-8.31)

*9.085802

(1.83)

9.293S5S (-O,82)

***31,414

(7.5239)

***1247726

(10.7938)

59,267 (1.2294)

^**47,993

(-O.2420)

1,380 (-O.9506)

^***442 (-6,4830)

*-418

(.1.7792)

2,551 (1.0659)

^"*-761 (-3.5936)

***1.8e7746

(-12.0135)

***O,035068

(-20.6786)

***O.O18e84

(-26.5644)

***O.084057

(-10.2968)

-O.04223 (-O.5690)

***123074

(-11.5230)

O.344778 (-9.48)

**3,747806

(-2.6126)

5,058046 (-1,2230)

107,567 1.20S363 260,689 223,662 8,288 5,565-2,723

13,OS3-4,766

S228555 O.114281 O.035351 O.205282

-O,091 4.304404 O.76449 6.676297 12,92603

14,186 O.771787 52,726 49,269 1,649 1,102-319

2,193-467

3.258519 O,le8696 O.080284 O.144477 -O.03908 2.499414 O.5475S8 4,O15724 5,250318

i. *', and *** indicatevariable differencesfrorntherestof thesample significant at10%,5%,and 1%levelr¢spectively. The t-statisticsintheparentheses^areproducedbyt-testforthcdifferencesinmeans; andz-statisticsbynon-parametrietestforthe differencesinmedians.

Definitionof Nlariables:Accruals:[Earnings^beforeextraordinary items-_CFO].PrvdltctienCosts(PROD):ICost^{of Goods}

Sold⁺ ChangeinInventoryl,Diseretionae,Eig,ensesCDisex):[Selling,^Generaland AdrninistrativeExpenses],EBEI:Earnings beforeExtraerdinaryItems,MZEi:The MarketValueof Equity(MarketCapitalization),Adhrhet-to-bookuaew:^{The Ratioof} MVE totheBookWhlueefEquity,inventorJ,71tmoverilatie:[Cost^{ofGoodsSold]1}[eeginning^{inventory+Endinginyentory)} f21,iaeceivable7i`rneverllatio:SalesX[(Beginning^GTossReceivables⁺ EndingReceivables)12].

4.4DescriptiveSiratistics

of,lixcess

^lhlues

Tal)le1 providesthe^{estimated results} from thefbur^models. Panel A detailsthe

regression estimates foreach industryand year.^{Panel B} providesdescriptivestatistics ofthe variables used teestimate themodels.

PanelA in^'fable 1 shows thecoefficients and t-statisticscalculated fromthe^mean

standard ^{errors across} al1industry-yearsfrom2000 to2004. Allregressions reported in PanelA are estimated fbr evgry industryineach year.Hence,theregressions are estimated ^{with a very strict assumption} thatall firrnsinthe same industry-yearhave

The Japanese Association of Management Accounting

NII-Electronic Library Service

The JapaneseAssociation of Management Accounting

ereekstgeac^17g M ¹e

Figure1-FrequencyDistributienofEarnings beforeExtraordinaryIterns(OTdinary^Incorne)

250

200

m 150.ELbts2i

100

50

opt8SSSSsstSSSSSSSSSSSS"geSSSSSSSSSSSS

' -

tt....ttttt-.t' t--';,!.ttt ./t'tt'atttt/tt,'.tt.'d

./.,tl'tt'.t,/t.',/t-!'' t/.t''g.tt/'tt

''tttt.t -tt-t/,.t/'/s .t't.''t;'''''tt'tL;tt.///;tt.tt't..

.t/t' ''

''' 'tt.t'''//tt-

'.t''t./･/ttt.t''tt'tt'

k''tt'' 't/t'tt'.ttttt/,}.tt'tt

tttt'tttt.t..tt''''

t/,t..ttt.ttt'ttttt/.t.t/ttt/'tl.tt/t1.

' ' '''''l'tt//tt''.ttt/tttttt.'

tt tt

''

.tt.'tttt.tt''t.'t/ttt-tt/tt''.,:'

tt.t 'tt'.t tt,ttt.tt./t.tt

'

exactly the sarne parameters.Inessence, ai1of theindependentvarial)les ineach ^model

(except

^{fbr AS,}

)

^are significantly correlated with theircorresponding dependent

variables.

^{Panel A also reports the adjusted R2 fbreach} regression. Since the cross-sectional regressions are estimated fbrevery industryand'year, the adjusted R2 in^'Ilable 1are

the mean adjustedR2. The mean adjusted R2 is O.2965 fbr CFO, O.9624 fbr production^costs, O.5914fordiscretionaryexpenses, and O.3526fbraccruals,

5.Selectionof [[largetFirms and DescriptiveStatistics 5.JStimplesIlxamined

This_paper examines firmswhere Isuspect managed earnings byconsidering iffirrns

manage eamings through^real activities manipulation to avoid reporting losses.The

existing literatureen earnings management

(Burgstahler

^{and Dichev,1997;Degeorge,}

Patel,and Zeckhauser, 1999; Matsumoto, 2002) shows that managers have a strong

incentivetoavoid negative earnings surprises or exeeed specific eamings thresholds.

The thresholdsthatmanagers have a strong incentiveto exceed include:

(aj

^zero

earnings

(to

^{not report a loss),}

(b)

^lastperiod's^earnings

(to

^{not report decreased}

eamings), and.(c) analysts' forecasts

(to

^{not miss analysts'} expectations).

Followingprior^wotk

(BurgstahIer

^{and Dichey l997;Beaver,McNichols, and Nelson,}

2007),firmsthatIsuspect of managing eamings toexceed thezero earnings threshold

are definedas firmsthat^report small positiveprofits.Smal1positive^profitsare defined

as earnings after extraordinary itemsdeflatedby rnarket capitalization thatare greater thanorequal to^zero, butlessthan 3%.

12