Jones 1991

Journal of Accounting ResearchVol. 29 No. 2 Autumn 1991

Printed in U.S.A.

Earnings Management DuringImport Relief Investigations

JENNIFER J. JONES*

1. Introduction

This study tests whether firms that would benefit from import relief(e.g., tariff increases and quota reductions) attempt to decrease earningsthrough earnings management during import relief investigations by theUnited States International Trade Commission (ITC). The import reliefdetermination made by the ITC is based on several factors that arespecified in the federal trade acts, including the profitability of theindustry. Explicit use of accounting numbers in import relief regulationprovides incentives for managers to manage earnings in order to increasethe likelihood of obtaining import relief and/or increase the amount ofrelief granted.

While studies of earnings management typically examine situations inwhich all contracting parties have incentives to "perfectly" monitor(adjust) accounting numbers for such manipulation, import relief inves-tigations provide a specific motive for earnings management that is not

* University of Chicago. I am especially indebted to my dissertation committee, JohnBound, Michael Bradley (cochairman), Michael Maher (cochairman), and Thomas Stoberfor their assistance and to others who have provided helpful comments, notably ananonymous referee, Victor Bernard, Dean Crawford, Linda DeAngelo, Bruce Ikawa, WilliamKinney, Richard Leftwich, ftobert Lipe, Peter Wilson, David Wright, and workshopparticipants at Carnegie-Mellon University, Northwestern University, Ohio State Univer-sity, University of Arizona, University of Chicago, University of Iowa, University of NorthCarolina, University of Rochester, Washington University, and Wharton School. I am alsoindebted to Ricbard Laulor from the United States International Trade Commission (ITC)for his help in obtaining information about the import relief investigation process at theITC. This paper has been funded by the Institute of Professional Accounting at theUniversity of Chicago and dissertation fellowships from the Arthur Andersen & Co.Foundation and the University of Michigan Dykstra Fund.

193

Copyright <P), Journal of Accounting Research 1991

194 JOURNAL OF ACCOUNTING RESEARCH, AUTUMN 1991

provided in other earnings management studies. Import relief is a wealthtransfer from a group of diffuse losers {consumers) to a group of concen-trated winners (all other contracting parties of domestic producers re-ceiving import relief).^ I argue that consumers do not monitor earningsmanagement as effectively as losers examined in other studies becausethe loss to each consumer is smaller, and their interests more diverse,than for the contracting parties examined in these studies.^ Regulatorshave less incentive to adjust for managers' earnings manipulations sincetheir ultimate payoff for such adjustment is less direct than in othersituations previously studied (e.g., union contract negotiations).^ Fur-thermore, interviews of ITC regulators indicate that the ITC does notadjust financial data for accounting procedures used or for accrualdecisions made by firms.

This study documents the use of accounting numbers in a federalgovernment program as a basis for wealth transfers (i.e., import relief).An estimate of the discretionary component of total acruals is used asthe measure of earnings management rather than the discretionarycomponent of a single accrual (as used in McNichols and Wilson[1988]). The discretionary component of total accruals is more appropri-ate in this context because the ITC is interested in earnings before taxes,which includes the effects of all accrual accounts, and, as such, managersare likely to use several accruals to reduce reported earnings. Firm-specific expectations models are developed to estimate normal (nondis-cretionary) accruals. The expectations models control for the effects ofeconomic conditions on the level of accruals. I conduct a cross-sectionalanalysis to test whether estimated discretionary accruals (i.e., residualsfrom the estimated expectations models) tend to be income-decreasingduring the import relief investigation period. The methodology developedin thi-s study extends the methodology used in other earnings manage-ment studies; specifically, time-series models are developed to estimatetotal nondiscretionary accruals and cross-sectional tests of the earningsmanagement hypothesis are applied. The results of these tests are con-sistent with the hypothesis that managers decrease earnings throughearnings management during import relief investigations. This evidence

' Examples of other contracting parties are management, sharebolders, debtholdere, andemployees. It sbould be noted that the long-run effects of import relief are not wellunderstood. It may be the case tbat the short-term costs to consumers (i.e., wealth transfersdue to import relief) are offset in the long-run by benefits from a stronger domestic industry,

^ For example, employees during union contract negotiations (Liberty and Zimmerman[1986]), shareholders in setting management compensation (Healy [1985)), shareholders inmanagement buyouts (DeAngelo [1986]), and shareholders in proxy contests (DeAngelo[1988]).

^ As I discuss later, the ITC bas been organized in such a way as to minimize the effectof voters on tbe ITCs actions; tberefore, the ITC may bave less incentive to adjust tbanregulators in situations in which voters have a more direct influence. In either case, tbepayoff to regulators for adjusting for managers' opportunistic accounting choices is iessdirect tban it is for other contracting parties such as debtbolders and stockholders.

EARNINGS MANAGEMENT DURING INVESTIGATIONS 195

is of particular importance in the light of the current interest in importprotection.

The next section provides institutional background for import reliefdeterminations. Section 3 develops the hypothesis to be tested. Section4 contains the sample selection procedures and descriptive statistics.Section 5 reports tbe results of the empirical tests. The last sectionprovides conclusions.

2. Role of Accounting Numbers in Foreign TradeRegulation

Foreign trade regulation provides avenues for granting import reliefthrough tariffs, quotas, marketing agreements, and/or federal adjustmentassistance. In most cases, an increase in import protection results in awealth transfer from domestic consumers, domestic importers, and for-eign suppliers to domestic producers of the protected good. Agents in thedomestic producers' nexus of contracts, such as employees, stockholders,debtholders, and suppliers, cannot be hurt directly by the import protec-tion and instead may benefit."* Managers of firms that would benefit fromincreased import protection have incentives to take actions to increasethe likelihood of obtaining such protection and/or increase the amountof protection granted. The ways in which managers can increase theexpected value of the import relief depend on the factors considered bythe regulators when making import relief decisions. In the remainder ofthis section, I review these factors.

2.1 STATUTORY PROVISIONS OF THE FOREIGN TRADE ACTS

The major statutory provisions of the foreign trade acts that relate toimport relief are summarized in Appendix A. The first three statutes,which pertain to general escape clause, countervailing duty, and anti-dumping investigations, are the primary focus of tbis study. Title VII oftbe Tariff Act of 1930 was designed to protect domestic industries fromimports that are sold at less than fair value (antidumping) or arebenefiting from foreign subsidies (countervailing duty). The generalescape clause investigations are based on section 201 of the Trade Act of1974, which was designed to aid domestic industries that are seriouslyinjured by increased imports. Section 201 is based on article XIX of theGeneral Agreement on Tariffs and Trade (GATT), which permits acountry to "escape" (hence the term "escape clause") temporarily fromits obligations under the GATT when increased imports of a specificproduct are causing or threatening to cause serious injury to domesticproducers of a like or directly competitive product. ITC investigationsconducted under section 201 provide a basis for the president to invokearticle XIX of the GATT.

* Except to the extent that the contracting parties are also consumers.

196 JENNIFER J. JONES

The antidumping, countervailing duty, and general escape clause stat-utes require the ITC to make a favorable injury decision before importrelief can be granted.'' In the case of countervailing duty and antidumpingcases, once the ITC has determined that an industry is being injured byimports, the Department of Commerce determines the increase in tariffsnecessary to offset the dumping margin or foreign subsidy. If the ITCrules favorably in general escape clause investigations, a recommendationis made by the ITC to the president to grant the industry some specifiedtype of import relief. The president has 60 days to make his import reliefdecision. If the president does not grant any import relief or grants reliefthat differs from that recommended by the ITC, Congress can overridethe president's decision and accept the ITCs recommendation by obtain-ing an affirmative vote in each House within 90 days after the president'sdecision. In each of these three types of investigations, the ITC mustfind that the industry has been injured before import relief can be granted.

The federal trade acts specify the factors to be considered when makingimport relief decisions. In the case of general escape clause investigations,the Trade Act of 1974 states that in determining injury:

. . . the Commission shall take into account all economic factors wbich it considersrelevant, including (but not limited to)—(A) witb respect to serious injury, the significant idling of productive facilities in the

industry, the inability of a significant number of ^rms to operate at areasonable level of profit, and significant unemployment or underemploymentwithin the industry;

(B) witb respect to threat of serious injury, a decline in sales, a higher and growinginventory, and a downward trend in production, promts, wages, or employment(or increasing underemployment) in the domestic industry concerned , . . (19 USC2251(b)(2)), [Emphasis added.)

The factors to be considered in antidumping and countervailing dutyinvestigations are as follows:

(1) actual and potential decline in output, sales, market share, profits, productivity,return on investments, and utilization oi capacity,

(2) factors affecting domestic prices, and(3) actual and potential negative effects on cash flow,'' inventories, employment,

wages, growth, ability to raise capital, and investment (Trade Agreements Act of1979, section 771 (19 USC 1677(7)). [Emphasis added.]

Since injury determinations specifically call for the use of accountingnumbers (i.e., profits, sales, and inventories) in foreign trade regulation,the remainder of this paper addresses the three types of investigationsthat require such determinations.

'' The ITC is responsible for making all injury decisions under the foreign trade statutes," The definition of cash flows used by the ITC is income before tax plus depreciation

expense.


2.2 USE OF ACCOUNTING NUMBERS BY THE ITC

The use of accounting numbers by the ITC is not only specified in thetrade acts but is also apparent in other ways. A review of 50 petitions^filed with the ITC for import relief investigations reveals that mostpetitioners cite the poor financial condition of the domestic producers asan indication of the industry's need for import protection. In an articleregarding the copper industry's petition for import relief, the Wall StreetJournal states that "the eleven copper producers that filed the complaintcited a $623 million loss last year" (Wall Street Journal [January 27,1984], p. 45). Another indication of the ITC's use of accounting numbersis reflected in the public version of its staff reports,^ which include asection devoted to industry financial performance. An analysis of theincome statement through net operating profit (or loss) before taxes forthe industry is always presented. The commissioners' injury opinions,which are included in the ITC staff reports, always include a discussionof the financial performance of the industry. All commissioners andcommissioners' aides that f interviewed agreed that the financial condi-tion of the industry is a key factor considered in injury determinations.^The use of accounting numbers by the ITC provides an incentive formanagers to manage earnings in order to increase the apparent injury tothe firm and, thereby, the industry. '

Staff members at the ITC indicated tbat the footwear industry was aprime candidate for inclusion in this study. Their reasoning was as

' In the case of general escape clause investigations, "a petition for eligibility for importrelief for the purpose of facilitating orderly adjustment to import competition may be filedwith the International Trade Commission . . . by an entity, including a trade association,firm, certified or recognized union, or group of workers, which is representative of anindustry" (Trade Act of 1974, 19 USC 225l(a)(l)). Investigations can also be instituted bythe president, the Special Representative for Trade Negotiations, Congress, or the ITC. Inthe case of antidumping and countervailing duty investigations, petitions can be filed by adomestic producer, or wholesaler, union or group of workers, or a trade association (see theTrade Agreements Act of 1979, 26 USC 3083(9)(C-E)).

* The staff report is prepared by the investigative staff at the ITC for use by thecommissioners in making injury determinations. The report includes a wide range ofinformation including a discussion of the product, domestic producers, foreign producers,level of imports, and all the economic factors specified in the trade statutes to be consideredin injury determinations such as production, capacity utilization, fmancial performance,product prices, and other causes of injury. The public version of the staff report alsoincludes the final opinions of the commissioners.

® Interviews were conducted in December 1986 with Commissioners Seeley G. Lodwickand Alfred E. Eckes and staff aides Kenneth Novak (aide to Commissioner Lodwick) andRon Blum (aide to Commissioner David E. Rohr).

^° Management discussion and analysis sections in annual reports and 10-Ks may alsoaffect the expected value of the import relief. For example, the staff report for oneinvestigation lists the reasons for the firms' poor financial performance during the pastthree years as described in the firms' annual reports and 10-Ks (Investigation Number TA-201-32, Publication no. 905, p. A-44).


follows. The footwear industry originally petitioned for an import reliefinvestigation in 1984. Later that year, the ITC ruled that the footwearindustry was not being injured. One reason for this finding was the factthat the footwear industry was relatively profitable in the most recentperiods. The petitioners later took their case to their congressionalrepresentatives. The federal trade statutes specify that, in the case ofgeneral escape clause investigations, one year must elapse between theITC's recommendation to the president and commencement of anothergeneral escape clause investigation unless there is good cause to foregothe required waiting period {Trade Act of 1974, 19 USC 2251(e))." TheSenate Committee on Finance issued a resolution stating that theybelieved circumstances had changed since the ITC's recommendation tothe president in 1984; therefore, they requested that the ITC institute asecond investigation. ITC Chairwoman Stern found in the second inves-tigation (1985) that "while the data in the previous investigation showedthat producers of the majority of domestic production were experiencingstrong profits, our most recent data show otherwise" (United StatesInternational Trade Commission, Publication no. 1717, p. 19). Thesudden and dramatic drop in profitability of the footwear industry ledsome staff members at the ITC to wonder if managers of these firms hadtaken steps deliberately to decrease reported earnings during the secondinvestigation period. ITC staff members do not believe that the footwearindustry is the only case in which managers may have deliberatelyreduced profits during import relief investigations; instead, they believethat it depicts an obvious case in which deliberate reduction of profits isa possibility.

Managers have greater incentives to make income-decreasing account-ing choices if they believe that the regulators do not completely adjustfor these chokes. Neither the public nor the regulators are necessarilythought to be "fooled" by the accounting numbers reported by domesticproducers. Instead, the regulators may be "captured" or may simply notregard "undoing" the reported numbers to be cost effective. ^ On the

' The waiting period differs if the industry is granted import relief by the president asa result of the earlier investigation. A general escape clause investigation cannot beundertaken unless two years have elapsed since the last day on which import relief wasprovided (Trade Agreement of 1974,19 USC 2253(j)).

" I t should be noted that in 1916 the ITC (then called the United States TariffCommission) was established as a nonpartisan information-gathering agency. Many of theearly advocates of the tariff commission had hoped to 'take the tariff out of politics' byestablishing a nonpartisan agency to analyze the impact of various tariff structures (Dobson[1976]). But as Dobson also notes, this was an impossible task for an issue as emotional asforeign trade regulation. Over the past 70 years. Congress has attempted to insulate theITC from politics by limiting the terms of the six commissioners (who are piesidentialappointees) to one nine-year appointment, limiting the number of commissioners fromeach political party to three, and staggering the terms of the commissioners. The overalleffectiveness of these restrictions is not centra! to this study, but it appears as tboughpolitics have not been completely eliminated from the ITC.


Other hand, voters may realize that they could personally benefit fromopposing special interest groups (i.e., domestic producers seeking importrelief), but this personal gain might not exceed the costs of informationsearch, lobbying, and forming coalitions. As stated by Peltzman [1976,p. 212], "producer protection represents the dominance of a small groupwith a large per capita stake over the large group {consumers) with morediffused interests." In summary, while regulators might not adjust theaccounting numbers for various reasons, consumers would not be able toform effective coalitions to oppose this practice because the potentialbenefit to each consumer is too small, and their interests too diverse, tomake such opposition cost effective.

Peltzman [1976] also argues that regulators will not exclusively serveone economic interest (e.g., domestic producers) but will instead help thegroups that provide the greatest amount of personal gain to the regulator.Thus, not all domestic industries seeking import relief can be expectedto obtain the desired relief. If all domestic producers are granted relief,there may be less incentive for them to increase the apparent injury tothe industry.^^ In a subsequent section, a description of the sample usedin this paper is presented which is consistent with Peltzman's theorythat the regulators (i.e., the ITC) will not exclusively serve one economicinterest {that of domestic producers).

2.3 ITC SOURCES OF FINANCIAL INFORMATION

The ITC obtains its financial information from the domestic producers'audited financial statements, 10-Ks, and responses to an ITC question-naire called the Producer's Questionnaire. The financial informationcollected in the Producer's Questionnaire is similar to that reported infirms' annual reports, but it is disaggregated by product line and/orestablishment.^'' Appendix B contains a summary of the data requestedin the Producer's Questionnaire. A company official is required to affirmthat the questionnaire is complete and correct to the best of his/herknowledge and beliefs. Domestic producers are required under law toprovide the requested data and subpoenas can be used to obtain theinformation if the producers fail to comply. The ITC attempts to obtaininformation from all domestic producers unless there are a very largenumber of small producers, in which case the ITC requests informationfrom a subset of the small producers. The ITC aggregates the dataprovided by the producers into industry totals.

^ Domestic producers may still bave incentives to increase the apparent injury to tbeindustry in order to provide an "excuse" for the ITC to recommend import relief. Also, iftbe industry granted import relief appears to be injured, it may reduce the risk of retaliationby foreign governments.

" An establishment is defined by the ITC as "each facility in the United States in whichproduct A is produced, including auxiliary facilities operated in conjunction with (whetheror not physically separate from) such production facilities." This definition was taken froma sample Producer's Questionnaire provided by the ITC.


The Trade Act of 1979 provides for the verification of informationsubmitted to the ITC by domestic producers. The Act provides that " . . .the administering authority shall verify all information relied upon inmaking a final determination in an investigation. . . . If the administeringauthority is unable to verify the accuracy of the information submitted,it shall use the best information available to it as the basis for itsdetermination . . ." (Trade Act of 1979, section 776 (19 USC 1677e)).

The following details of the verification process used by the ITC wereobtained from interviews with members of the investigative staff of thejrpQ 15 rpi ^ YYC does not verify any information in the audited financialstatements or 10-Ks, nor do they make any adjustments to these data.Verification of the submitted data is restricted to the product line and/or establishment data provided in the Producer's Questionnaire. Most ofthe verification process is aimed at determining whether the cost aliO'cation methods used were appropriate and consistently applied. The ITCdoes not attempt to adjust the financial data for accounting proceduresused or for accrual decisions made by the firms' managers.

3. Hypothesis Development

3.1 EARNINGS MANAGEMENT HYPOTHESIS

The ITC's use of reported earnings in injury determinations providesan incentive for managers to make accounting choices that increase theapparent injury of the firm. By doing so, managers may increase theprobability of obtaining the desired import relief and/or increase theamount of relief granted; therefore, the link between accounting numbersand injury determinations may result in managers' accounting choiceshaving economic consequences (i.e., wealth transfers from consumers todomestic producers due to import relief). This incentive leads to thefollowing hypothesis.

EARNINGS MANAGEMENT HYPOTHESIS: Managers of domesticproducers that would benefit from import protection make accountingchoices that reduce reported earnings during ITC investigation periodsas compared to noninvestigation periods.

3.1.1. Conflicting Incentives. An assumption underlying this hypothesisis that the import relief incentive to decrease reported earnings is greaterthan other incentives the managers have to increase reported earnings.Prior research in other contexts (see n. 2) indicates that managers faceother economic consequences of their accounting choices that motivatethem to make income-increasing rather than income-decreasing account-ing choices—for example, debt covenants and management compensation

' Interviews were conducted in December 1986 with Richard Laulor, supervisor of theFinancial Analysis and Accounting Division of the Office of Investigations, Chand Mehta.accountant, and Dan Dwyer, investigator.


plans. ^ By increasing reported earnings, managers can reduce the restric-tiveness of the debt covenants and increase their own compensationthrough higher bonuses.

Debtholders would benefit by tolerating managers' income-decreasingaccounting choices during import relief investigations, even if it requiresthem to waive or amend covenants that are violated,'^ since the financialperformance of the firm can be expected to improve if import relief isgranted. Managers would also benefit from obtaining import relief if thefuture earnings of the firm are higher than without the relief and higherearnings result in higher bonus payments. Thus, during import reliefinvestigations managers have less incentive to increase reported earningsthan they would at other times because it is in the best interests of allcontracting parties (except consumers) for the firm to obtain the desiredimport protection.

3.1.2. Free-Rider Problem. The fact that all domestic producers withinan industry stand to benefit if import relief is granted results in a free-rider problem. As a result, managers of firms within an industry may nothave equal incentives to manage earnings during import relief investi-gations. The ITC evaluates the overall results of the industry but doesnot require that all firms be injured in order to recommend import relief;therefore, the managers of some domestic producers may decide that theresults of their operations will not alter the ITCs ultimate decision and,consequently, they may not manage earnings during import relief inves-tigations. Of course, if a large number of managers adopted this attitude,then, as a group, they could potentially affect the ITCs ultimate decision.Also, some firms may not decrease reported earnings during import reliefinvestigations because their management compensation and/or debt cov-enant incentives to manage earnings upward override the import reliefincentive.

In order to address the free-rider problem, a supplemental test of theearnings management hypothesis restricts the sample to firms thatpetitioned for import relief. The petitioners bear the costs, thought to bequite substantial in many cases, of supporting their claims of injurybefore the ITC Thus, petitioners may have greater incentives than otherdomestic producers in the industry to maximize the probability of ob-taining import relief and, as a result, petitioners may have greaterincentives to manage earnings during import relief investigations.

3.2.5. Investigation Types. The type of import relief investigation may

'Êvidence presented in Healy [1985] indicates that management compensation plansdo not motivate managers to make strictly income-increasing accounting choices but insteadtbe accounting choices depend on the relation of the earnings number (before these choices)to any upper or lower limits specified in the plans.

''' Asarco, Incorporated is an example of a domestic producer that obtained debt covenantamendments wben the covenants were violated during an ITC import relief investigationof the copper industry (Asarco, Incorporated. 1984 Annual Report).


also affect managers' incentives to manage earnings during import reliefinvestigations. Antidumping and countervailing duty cases are institutedwhen there is evidence that unfair trade practices exist, whereas generalescape clause cases are instituted when there is no such evidence. Man-agers may have greater Incentives to manage earnings in general escapeclause investigations (section 201) than in antidumping or countervailingduty cases because "to be found eligible for relief under section 201,industries need not prove that an unfair trade practice exists, as isnecessary under the antidumping and countervailing duty laws andsection 337 of tbe Tariff Act of 1930. However, under section 201, agreater degree of injury, 'serious' injury, must be found to exist."^^ Sincea greater degree of injury is necessary to obtain relief, managers mayhave greater incentives to decrease reported earnings in general escapeclause investigations than in other types of investigations; therefore, inan alternative and, perhaps, more powerful test of the earnings manage-ment hypothesis, I restrict the sample to general escape clause investi-gations.

3.2 INVESTIGATION PERIOD

The ITC normally requests information for five years prior to the datethe petition was filed for general escape clause cases, and three years forantidumping and countervailing duty cases.'^ In some cases, data for themost recent quarter are also requested. The actual information requestedby the ITC in each investigation is indicated in the staff report.

The length of time taken to complete an ITC investigation depends onthe type of investigation.^" The ITC must complete general escape clauseinvestigations within six months after the filing of the petitions. Inantidumping and countervailing duty investigations, the ITC must com-plete their investigations within 120 days after an affirmative preliminarydetermination (dumping or subsidy) or 45 days after an affirmative finaldetermination by the Secretary of Commerce.^'

Tests of the hypothesis are restricted to the two most recent yearsreviewed by the ITC; the year the investigation is completed (year 0) andthe prior year (year -1).^^ It is not clear when managers first anticipate

' United States International Trade Commission, 1985 Annual Report, p. 1.^^The period for which information is requested can be altered under special circum-

stances, for example, in the case of a cyclical industry.^ Interviews with members of the investigation staff (see n. 15) indicate that the I'l'C

normally uses the maximum amount of time allowable to investigate cases.^ Each final antidumping and countervailing duty investigation conducted by the ITC

is preceded by a preliminary investigation in which the ITC has 45 days after the petitionis filed to determine whether there is a reasonable indication that the industry is beinginjured by imports. If the ITC's preliminary investigation results in a negative determina-tion, the case is terminated.

^ In most cases, the ITC completes its investigation in the same year the petition isfiled, but in some cases the petition is filed late in the year and the investigation is thencompleted in the following year. None of the firms in this study was part of an investigation


a future import relief investigation but is unlikely that they wouldanticipate it prior to year —1; thus, they would have no import-relief-related incentive to manage earnings during periods prior to year —1.The ITC does not formally request information for year 0 (except, insome cases, for the most recent quarter), but there is evidence that thisinformation enters the ITC's decision process either through the publichearings or by voluntary submission of the data.^^ Interviews with theITC commissioners and commissioners' aides (see n. 9) and statementsmade by commissioners in their injury opinions provide evidence thatthey place greater reliance on results for years —1 and 0 than on prioryears. The commissioners are looking for a downward trend in financialperformance or a drastic decline in the most recent periods, providinggreater incentives for managers to decrease reported earnings in the mostrecent

3.3 LIMITATIONS

The empirical tests might not support the earnings managementhypothesis for several reasons. First, managers may believe the ITCadjusts for their discretionary accounting choices, reducing their incen-tives to use accounting choices to manage earnings. Interviews conductedat the ITC indicate tbat the ITC does not adjust for accounting choices,and most of the information that the ITC uses in its injury determinationis publicly available; therefore, managers should be aware of the ITC'spractices. Second, financial performance of the affected firms may be sobad that managers do not need to use accounting choices to manageearnings. If tbe amount of injury found by the ITC impacts the amountof relief granted, then firms will still have an incentive to manageearnings. Third, managers may rely on cost allocations rather thanaccruals to manage earnings for the product line investigated by the ITC.Cost allocations can be used by man&gers to shift revenues and expensesbetween the product line being investigated by the ITC and other productlines. Finally, tbe power of the tests may not be sufficient to detect

in which the ITC did not formally request information for the year preceding the completionof the investigation.

^ For example, data through June 30, 1980 were formally collected by the ITC in theautomobile investigation but several of the commissioners' opinions refer to results fromoperations subsequent to June 30, 1980. Commissioner Stern's opinion states that "thecurrent year (ending December 31, 1980) will be the first in recent history in which theindustry shows aggregate losses" (United States International Trade Commission, Publi-cation no. 1110, p. 119).

^ For example, in tbe general escape clause investigation of automobiles which coveredthe period 1975 through 1979 and the first six months of 1980, Cbairman Alberger statedthe following in his injury opinion: "In the aggregate most of the indices of the U.S.automobile producers' performance during the period of investigation reveal a healtbypicture from 1976 through 1978 and rapidly declining trends thereafter" (United StatesInternational Trade Commission, Publication no. 1110, p. 17). Commissioner Albergerfound that the industry was injured.


managers' income-decreasing accounting choices. The sample selectionprocedures and empirical tests described in subsequent sections aredesigned to mitigate as many of these limitations as possible.

4. Data and Descriptive Statistics

4.1 SAMPLE SELECTION

The sample used in this study is restricted to import relief investiga-tions that require the ITC to make an injury determination: antidumping,countervailing duty, and general escape clause investigations. Further,only investigations pertaining to a broad product line, such as automo-biles and footwear, are examined because earnings management fornarrow product lines may not be material relative to the consolidatedfinancial statement data used in the empirical tests. Table 1 describesthe five industries included in the sample: automobiles, carbon steel,stainless steel, copper, and footwear. These industries represent sixinvestigations {as discussed earlier, there were two footwear investiga-tions). Five of the investigations were considered under the general escapeclause provision, and the other was considered under both the antidump-ing and countervailing duty provisions. Favorable rulings were issued bythe ITC in three of the investigations {stainless steel, copper, and the1985 footwear cases), and the president granted relief for two of the three(the stainless steel and 1985 footwear cases).

Relief granted to the footwear industry was in the form of adjustmentassistance paid to displaced workers and, as such, did not directly benefitthe domestic producers. Stainless steel was the only industry that ob-tained a substantial amount of relief as a direct result of the ITCinvestigations. Voluntary import limit agreements were reached withforeign governments in the automobile and carbon steel industries.

Data from firms' annual financial statements are used to construct aproxy for firms' earnings management. Data for overall operations areused rather than segment data because the segment data do not provideenough information to compute an estimate of firms' earnings manage-ment (i.e., accruals). Table 2 summarizes the sample selection criteria.Financial data for 49 firms in the five industries are available on Com-pustat. One firm with foreign ownership is excluded from the samplebecause this firm may not have benefited from import relief. Also, twofirms that expressed opposition to import relief are omitted from thesample because they have incentives to increase rather than decreasereported earnings during the investigation.^^ Five firms are excluded from

^ Firms' positions on import relief were obtained or inferred from ITC staff reports,transcripts of ITC hearings, industry data identifying importers, and Wall Street Journalarticles. Both of the opposing firms were substantial importers of tbe goods in question;therefore, import relief may hurt tbese firms. Thus, these firms may have incentives todecrease the apparent injury to the firm so as to decrease the likelihood of obtaining reliefand/or amount of relief granted.



TABLE 2Summary of the Sample Selection Criteria

Description

Total number of domestic producers—onCompustat^

Firms omitted from the sample:Foreign ownershipOpposed relief (importers)Division/subsidiary of a firm in another

line of business .Too few time-series observations . .Highly diversified firms . . . .

Total number of firms omitted from thesample

Total included in the sample

Autos Copper

4 11

(1)(1)

(5)

0 (7)

4 4

CarinonSteel

12

(4)(•i)

(7)

5

„ , Stainlessr ootwear ,-•, ,.Steel

15 7

U)

(4)(2) (5)

(7) (5)

8 2

Tota

49

(1)(2)

(;•.)

(S)(1(1)

(261

23

"The domestic producers were obtained Irom the ITC staff reports, petitions filed with the VVC andcorrespondence files maintained on microfiche at the IT(",

the sample because they are divisions or subsidiaries of a firm in anotherline of business. Firms with too few time-series observations (less than14 years) are also excluded from the sample as estimation of the expec-tations models would be hindered (eight firms). Domestic producers withoperations in several different industries (highly diversified) are excludedsince earnings management for the segment being investigated by theITC may not be detectable in the aggregated data (ten firms). Also, theITC may rely less heavily on the overall results of these diversified firmswhen making injury decisions. The empirical tests are based on theresulting sample of 23 firms from five industries.

4.2 MEASURE OF EARNINGS MANAGEMENT

Earnings management can be achieved by various means such as theuse of accruals, changes in accounting methods, and changes in capitalstructure (e.g., debt defeasance, debt-equity swaps). This study focuseson total accruals as the source of earnings management. More specifically,discretionary accruals are used as a measure of managers' earningsmanipulations during import relief investigations. Previous studies suchas DeAngelo [1986], Healy [1985], and McNichols and Wilson [1988],which also use some type of discretionary accruals measure, discuss thepartitioning of total acruals into discretionary and nondiscretionarycomponents.

The discretionary portion of total accruals is used in this study tocapture earnings management rather than the discretionary portion of asingle accrual account (as used in McNichols and Wilson [1988]) becausetotal accruals should capture a larger portion of managers' manipulations.


Total accruals are calculated as the change in noncash working capitalbefore income taxes payable less total depreciation expense. The changein noncasb working capital before taxes is defined as the change incurrent assets other tban cash and short-term investments less currentliabilities other than current maturities of long-term liabilities and in-come taxes payable.^^ This accrual measure excludes accruals related toincome taxes because the ITC bases its evaluation on income beforetaxes.

4.3 DESCRIPTIVE STATISTICS

The descriptive statistics presented here are based on the expectationsmodel used by DeAngelo 11986]. DeAngelo used total accruals from aprior period (t — k) as a measure of the "normal" total accrual. Shedefines the "abnormal" total accrual iATA) as the difference betweencurrent total accruals and normal total accruals, which, in turn, can beseparated into discretionary and nondiscretionary accruals;

- TA,-,) = iDA, - DA,-k) - iNA, - NA,-,). (1)

DeAngelo tested whether the average value of the abnormal accrualwas significantly negative during the event period. This test relies on theassumption that tbe average change in nondiscretionary accruals,iNA, - NAt-k), is approximately zero, so that a change in total accruals,

— TAt-k), primarily refiects a change in discretionary accruals,— DAf-k). Table 3 summarizes scaled changes in accruals, earnings,

cash flow, and revenue before taxes for the sample for years —5 through-hi. The scaled changes are computed as first differences of the variables(Xt — Xt-i), divided by total assets at time t — l . Table 3 presents themean and median change for each of the variables, as well as the numberof negative and positive changes, ^-statistics (null hypothesis that theaverage change is zero), and significance levels for the nonparametricWilcoxon signed-ranks test.

Cbanges in accruals scaled by total assets are reported in panel A oftable 3. Prior to year 0, all of the accrual changes are relatively small.The change in accruals in year 0 is negative with a (-statistic of —1.824.If the change in accruals is viewed in isolation, year 0 suggests thatmanagers are making income-decreasing accrual decisions. These results,bowever, must be interpreted cautiously because panels B through Dindicate that changes in earnings, cash flows, and revenues are alsosignificantly less than zero in year 0. ^

'^ The composition of total accruals (TA,) is as follows: TA, = [iHiCurrent AssetSi (4) -ACash, (1)] - [ACurrent Liabilitiesi (5) - ACurrent Maturities of Long-Term Debt, (44) -Aincome Taxes Payable, (71)] — Depreciation and Amortization Expensei (14), where thechange (A) is computed between time t and time t — l ; Compustat data item numbers areindicated parenthetically.

^' Cash flows are defined to be earnings (or income) less accruals throughout this paper.


f-H

o

p 00 pd T-H d[ I I

^ OO CO O I-* ClP CO O ^ - • *O O O .-I OI I

in t~ Tfcq Ol OO -I O

O r- Ol O -^l O 1-H O T-H WP f-: o ^ -HO O O 1-H O

o

O >o r-CO T-H Tt<O ri; "H

Ol 00 loO OJ OlT-; CO pd i/i r^

I I I

f l

A1

t ood1

CO

IO1-H

od

l O<MOd1

l OCO

F-H

1

O

d1

CTJ CO

d

(O CO 00.-H T-H .-Hp (i^ oOd d

I I

0 0COod1

t -aCOld1

OJ a iT-H "^

p ;ô

1—(

o( M

d

G*J " ÔJ Ô IOd i>i

,-HCO

od

o oT-H O

d

O C J " Ô T-H 00O JJ CO

fl

ed

O

Med

ian

oa

>

«CCO

- H O

1a SUl c

ce

m

Pan

el 1

Mea

n.t-

stat

isl

Med

ian

oQ.

>Ca:

=*; CO


^ •"! ' ^^ <6 dI I I

I 1

O O OI I I

d dI I I

O CD 00ro tD (oT-t CO o

-5 ?f c

C-- CD O O lOT-I .—1 - ^ .—I inq 's; q ^ iq<:i '^ di r-i dI I I

I I I

C-l CD

id d

as

s raFS; qfc m

a, 016 £

S ^> i

- O

-ali

ivid

•ajc

s

a

aj

oot'

^4.H

•IOJ

X !

o

x:

!.>u^hjjC

CJ 3

the

c

s a

"•2 "O ^

o £

C T3

2 o

.S IOJ ^

So S

^ C * ™

^ +


As Kaplan [1985] notes, changes in several working capital accountsand, thereby, accruals, depend upon the economic circumstances of thefirm. For example, if nondiscretionary accruals are a function of revenues,then the negative change in accruals may simply be due to changes innondiscretionary rather than discretionary accruals. The effect of declin-ing revenues on accruals is of particular importance in this study becausefirms in industries facing rising imports can be expected to have decliningrevenues. If revenues affect the level of nondiscretionary accruals, thenan expectations model used to measure nondiscretionary accruals musttake this relation into account. A test of the earnings managementhypothesis that attempts to control for the effect of changing economiccircumstances on accounting accruals is presented in section 5.

Since the sum of a firm's income over all years must equal the sum ofits cash flows, managers must at some point in time reverse any "exces-sive" earnings-decreasing (or increasing) accruals made in the past. Also,after the ITC investigations are completed, managers' incentives toincrease reported earnings for reasons related to compensation and/ordebt covenants will return. If managers do not expect to petition forimport relief again in the near future (the earliest consideration of a newinvestigation is limited by the trade statutes) and there is no ex postsettling up by the foreign trade regulators, managers will bear no costsby increasing reported earnings after the final import relief decision hasbeen made. The results for year +1 indicate that changes in accruals arenot significantly greater than zero (^statistic of 1.488) nor are changesin earnings ((-statistic of 1.295). Changes in cash fiows and revenues inyear -I-l also are not significantly different from zero."** These resultsmay be due to the fact that managers tend to reverse "excessive" earnings-decreasing accruals over a period of more than one year or that they faceother incentives that conflict with the reversal such as the intention topetition for import relief again in the near future or to avoid any ex postsettling up by the regulators.

5. Tests of the Hypothesis

5.1 ACCRUALS MODEL

The descriptive statistics presented in section 4 can be interpreted assupport for the earnings management hypothesis only if one assumesthat the difference between current- and prior-year accruals is due solelyto changes in discretionary accruals because nondiscretionary accrualsare assumed to be constant from period to period. To relax this assump-tion, I use the following expectations model for total accruals to control

^ The independence assumption for both the ;-test and Wilcoxon signed-rank.s lewt ismost likely violated because the sample is clustered by industry and time period. In ihonext section, I present a model that attempts to control for the effect of economircircumstances on accruals and a test of the earnings management hypothesis thai addressesthe prohiem of contemporaneous correlation in accruals.


for changes in the economic circumstances of the firm:

where:

(2)+ ut

it = total accruals in year t for firm i;it = revenues in year t less revenues in year t — 1 for firm i;

PPEu = gross property, plant, and equipment in year t for firm i;Ait-\ = total assets in year t - 1 for firm i;

tit = error term in year t for firm i;i=l,...,N firm index (N = 23);t=l,. . . ,Ti, year index for the years included in the estimation

period for firm i iT, ranges between 14 and 32 years).

In order to provide a longer time series of observations, the definitionof total accruals used in tests of the earnings management hypothesisreported in this section has been modified from that used in section 4.The total accruals measure used in this section is not adjusted for currentmaturities of long-term debt and income taxes payable because severalearly observations are missing from the Compustat tapes.^^ Excluding theadjustment for current maturities of long-term debt and income taxespayable increases the average number of observations for each firm from12.9 to 25.2.-'"

In equation (2), gross property, plant, and equipment and change inrevenues are included in the expectations model to control for changesin nondiscretionary accruals caused by changing conditions. Total ac-cruals [TA) includes changes in working capital accounts, such as ac-counts receivable, inventory and accounts payable, that depend to someextent on changes in revenues. Revenues are used to control for theeconomic environment of the firm because they are an objective measure

® The composition of total accruals (TA,) used in section 5 is as follows: TA, = [^CurrentAssets, (4) - ACash, (1)] - [ACurrent Liabilities, (5)] - Depreciation and AmortizationExpense, (IA), where the change (A) is computed between time t and time t - 1; Compustatdata item numbers are indicated parenthetically. The descriptive statistics for changes inaccruals, earnings, and cash flows using this modified definition of total accruals for years—1,0, and-Hi are very similar to those presented in table 3, although some of the significancelevels are somewhat lower.

•'" Since regression equation (2) is used to estimate "normal" accruals, levels of totalaccruals rather than the changes in total accruals are used in this equation. In addition tobeing conceptually superior, tbe use of total accruals makes estimation of the expectationsmodel more amenable. Unlike other time-series processes which need to be differenced inorder to obtain stationarity, total accruals tends to be a white noise process, Tbe averagefirst-order autocorrelation for total accruals (TA) is 0.075, wbereas it is -0,498 for cbangesin total accruals (ATA). Autocorrelations for tbe accnials measure used in section 4 (i.e.,total accruals adjusted for current maturities of long-term debt and income taxes payable)are similar, with an average of —0.076 for total accruals and —0.505 for changes in totalaccruals.


of the firms' operations before managers' manipulations, but they arenot completely exogenous." Gross property, plant, and equipment isincluded to control for the portion of total accruals related to nondiscre-tionary depreciation expense. Gross property, plant, and equipment isincluded in the expectations model rather than changes in this accountbecause total depreciation expense (versus the change in depreciationexpense) is included in the total accruals measure.^^ All variables in theaccruals expectations model are scaled by lagged assets to reduce heter-oscedasticity. As described in Kmenta [1986], a weighted least squaresapproach to estimating a regression equation with a heteroscedasticdisturbance term (i.e., the unsealed regression equation) can be obtainedby dividing both sides of the regression equation by an estimate of thevariance of the disturbance term (i.e., resulting in a scaled regressionequation). In this case, lagged assets (Au-i) are assumed to be positivelyassociated with the variance of the disturbance term.'''

Ordinary least squares is used to obtain estimates Oj, bi,, and b ; of a,,û, and 02i, respectively. This model assumes the relation between non-discretionary accruals and the explanatory variables is stationary. Theprediction error is defined as:

Uip = I Aip/Aip.-](3)

-H b[REyjA] bAPPEJA])

where p = year index for years included in the prediction period. Theprediction error, a p, represents the level of discretionary accruals at timep. The model is estimated using the longest time series of observationsavailable prior to year —1 for each firm. The use of a long time series ofobservations improves estimation efficiency but also increases the like-lihood of structural change occurring during the estimation period.

Table 4 provides descriptive statistics for the multiple regressionsestimated over all available observations through year —2. The averageresidual first-order autocorrelation is —0.171. The Durbin-Watson two-tailed test statistics indicate that the first-order autocorrelation is notsignificant at the .05 level for 17 of 23 firms and is inconclusive for the

• ' Reported revenues may be affected to some extent by managers' attempts to decreasereported earnings. For example, managers may postpone the shipment of merchandiseduring import relief investigation years in order to postpone recognition of revenue untilthe following year.

^ Some support for the choice of these variables can be found in a study by Kaplan[1979].

^^ The need for scaling was assessed by correlating the squared residuals obtained fromthe unsealed expectations mode! (i.e., equation (2) without scaling) with squared laggedassets. The resulting correlation of 0.643 indicates that the error term from the unsealedexpectations model is highly correlated with lagged assets. Other scaling factors have beenused to reduce heteroscedasticity, for example, Lipe [1986] deflated earnings componentsby the Consumers Price Index and Rayburn [1986] scaled earnings components by themarket value of equity.

EARNINGS MANAGEMENT DURING INVESTIGATIONS

TABLE 4Descriptive Statistics for the Multiple Regression Equations for Total Accruals'

(Estimated over years prior to year —!)

213

a.

(-statisticft.t-statistic/32i

t-statisticR-squaredAutocorrelationDurbin-WatsonNumber of years

Mean

11.0880.0840.0350.220

-0.033-1.269

0.232-0.171

2.24425.261

Median

0.2080.034

-0.008-0.172-0.030-1.385

0.249-0.151

2.22828.000

Std. Dev.

49.7950.8280.1442.3720.0471.3940.1520.1670.3955.902

Minimum

-13.771-1.954-0.196-3.315-0.141-4.030

0.000-0.476

1.40414.000

Ql"

-1.350-0.377-0.068-1.850-0.050-2.238

0.132-0.294

1.98421.000

Q3

4.4470.5720.1631.835

-0.017-0.186

0.310-0.048

2.54731.000

Maximum

238.5401.7050.3754.4400.0801.0860.5810.2102.818

32.000

" The descriptive statistics presented are for the estimated multiple regression equation:

TAJA,,., - .,,,| -t-

where:

i, — total accruals in year ( for firm i;;, = revenues in yeaT t less revenues in year ( — 1 for firm i;

PPEi, — gross property, plant, and equipment in year ( for firm i;Aa-i — total assets in year t ~ 1 for firm i;

(i, = error term in year r for firm i;i = 1 , . . . , 23 firm index;t — 1,.. ., Ti, year index for the years included in the estimation period for firm i.

The composition of total accruals {TA,) is as follows: TA, — [ACurrent Assets, (4) — ACash, (1)] —[ACurrent Liabilities, (5)j — Depreciation and Amortization Expense, (14), where the change (A) iscomputed between time t and time ( — 1; Compustat data item numbers are indicated parenthetically.The regression equations are estimated over all available years prior to year —1.

*• Ql, Q3 are the first and third quartiles of the distribution, respectively.

remaining 6 firms. The average estimated coefficient for property, plant,and equipment is negative (—0.033), which is the expected sign becauseproperty, plant, and equipment are related to an income-decreasingaccrual (i.e., depreciation expense). The expected sign for the change inrevenues coefficient is not as obvious because a given change in revenuecan cause income-increasing changes in some working capital accounts(e.g., increases in accounts receivable) and income-decreasing changes inothers (e.g., increases in accounts payable). ** The average estimatedcoefficient for the change in revenues is 0.035, whereas the median is—0.008. The average R^ for the regression equations is 0.232.

'^^ Predictions regarding the sign of the regression coefficients are consistent with tberesults reported in Jones [1988] for regression equations estimated for four individualcomponents of total accruals: accounts receivable, inventory, accounts payable, and depre-ciation expense. Tbe coefficient for property, plant, and equipment was significantlynegative for tbe depreciation expense regression and insignificant for all otbers. Tbecoefficient for tbe change in revenues was significantly positive for accounts receivable andinventory, significantly negative for accounts payable, and insignificant for depreciationexpense. Another approach to estimating total discretionary accruals would be to developspecific models of nondiscretionary accruals for each component of accruals, as is done byMcNichols and Wilson [1988], rather than use one model to estimate nondiscretionaryaccruals for all accrual components, as is done in this paper.


Tests of the earnings management hypothesis are based on the estimateof discretionary accruals, u,p, during years —1 and 0. One method oftesting the overall significance of managers' discretionary accruals is tocompute a standardized prediction error similar to that used by Patell[1976]. For each prediction error, an estimated standard deviation, a(Uip), is calculated.^^ If the prediction errors are normally distributed,then the following ratio of the prediction errors to their standard devia-tions has a t-distribution with T, - 3 degrees of freedom:

Vip = Uip/a (u,p). (4)

The V,pS are referred to as "standardized prediction errors." FollowingPatell, the central limit theorem can be invoked to compute the followingtest statistic:

- 3)/{r, - 5)1/2

(5)

which is asymptotically distributed as a unit normal deviate if theprediction errors are cross-sectionally independent. "^ In this test, the nullhypothesis is that the average prediction error (i.e., discretionary accrual)during import relief investigations is greater than or equal to zero. Theexistence of cross-sectional correlation results in violations of the as-sumptions underlying the test statistic, and thus, any inferences basedupon the Z statistic must be made cautiously. In section 5.5, a test isconducted that addresses the cross-sectional correlation problem.

Due to the fact that two footwear cases were conducted by the ITC(i.e., in 1984 and 1985), two sets of tests are conducted: treating 1984and, also, 1985 as year 0 for the footwear industry. The results treating1984 as year 0 are reported in the body of the paper; those for 1985 arereported in the footnotes. The results are more supportive of the earningsmanagement hypothesis when 1985 is treated as year 0 for footwear.

Table 5 presents the VipS (standardized prediction errors) and relatedZvp statistics. The VipS are based on prediction errors from the totalaccruals expectations models estimated (see equations (2) and (3)) overperiods using all available data through year -2 . The Z statistics foryears —1 and 0 are —0.372 (with a one-tailed significance level of 0.356)

• The variance for the prediction error is derived by Theil [1971, pp. 122-23] as thefollowing; E[upUp'] = ir^ldp + I], where Cip equals [Xp(X'X)~'Xp'] in which X is thematrix of independent variables for the estimation period and Xp is the matrix for theprediction period. The standard error from the estimation period, -T, provides an unbiasedestimate of tr. I is the identity matrix.

® In order to apply significance tests based on the Z statistic, the prediction errors mustbe normally distributed and the covariance structure is assumed to be as follows: cov(u,,,,Ujp) equals zero when i j and equals (T,^[Cip + I] when i = j ; where Cip = [Xp(X'X)" 'Xp' jfrom the previous footnote. The denominator of the Z statistic is the sum of the variancesof the VipS. Since V,p is a (-statistic with T, degrees of freedom, the variance of V,p is (T, -3)/(T, - 5); see Theil [1971, p. 82].


TABLE 5Individual Firm Standardized Prediction Errors (Vip) and Related Test Statistics (Zvp)

from the Total Accruals Regression Models Estimated over the Period Prior to Year —1"

123456789

1011121314151617181920212223

Zvf, statistic''

Year - l "

0.534-1.218-0.623-0.514

0.097-0.114-0.211-0.128-0.115

1.641-0.795

0.1170.8940.224

-0.2030.4050.328

-0.772-0.216

1.006-1.805

0.089-0.501

-0..S72

YearO

-0.369-0.921-0.812-0.502-0.041-0.515

0.2930.293

-0.414-1.397

0.331-0.749-1.890-2.004-0.218-0.622-0.339-1.479-0.548-0.248-0.222-2.318-2.794

-3.459

Year +1

-0.519-1.806

0.5460.0120.067

-0.426-1.552-0.609

0.603-2.055-0.738

0.781-0.976-0.783

0.1710.181

-0.0621.795

-0.4830.1650.252

-0.534-0.234

-1.228

' V , is computed as i(^/(.s,(l + dp')), where Cjp = (Xp(X'X) 'Xp'] in which X is the matrix ofindependent variables for the estimation period, Xp is the matrix for the prediction period, u^ is theprediction error, p is the prediction year, and x, is the standard error from estimates of the followingregression model:

where:

TAa — total accruals in year t for firm /;u — revenues in year t less revenues in year t - 1 for firm i;a = gross property, plant, and equipment in year (ior firm i;

Au-i = total assets in year t ~ 1 [or firm c;(,t = error term in year t for firm i;1 = 1., , . , N firm index {N = 23);( = 1,. . , , 7",, year index for the years included in the estimation period for firm i.

The composition of total accruals (TA,) is as follows: TA, = [ACurrent Assets, (4) - ACash, (1)] -[ACurrent Liabilities, (5)] — Depreciation and Amortization Expense, (14), where the change (A) iscomputed between time t and time t — 1; Compustat data item numbers are indicated parenthetically.The regression equations are estimated over all available years prior to year —1.

'' Year 0 is the year the ITC completed its investigation, whereas year —1 is the previous year andyear +1 is the subsequent year. Year 0 for the footwear industry is 1984.

• The Zvp statistic is calculated as SU, V^ [Sî (T, - 3HT. - 5)]"^ where T, is the number of yearsin the estimation period.


5 1

o o o oI

o o o o o o o o o o o o o o o o o o oI t I I I I I I I I I I I { I I

I I Io o o oI I I

o oI I

o o o oI I I

o o o oI I I I

I I I

o p p p p p p p - - ;d d d o o o d d oIII I I I I

o o o oo o o o

IoI I

o o c s oo o o oI I [ II I

d o o o o o o o o o o o o o o o o o o o o o o

p p p p p p p ; ^ p p p p ;d d d d d o o d o d d d o o d d o o o o o o oI I i I I I I I I I I I I I I I I I I I


II

01C

Z

32o

=4-1

B

U3

CO

o1

,112

o(

C^

oo1

GO

oo1

CO

1—>

o1

o?

(A

'EH

H

fflo

3

COo"ô1

,071

o(

Oo1

t o

ooi

CO• ^

•—1

o1

o

O

1-tcb;_ucC

6

CC

cIS

I—

1

• ^

«;

c1

I Tccc:

c^CC

1

i r

C

1

1/Cl

"o

afz

•grai-O

CS

<NCO

oo

.057

o(

ooo

t~oo1

O)oôo1

o

c

z

IU

£o

=2

oCOoo1

.023

oI

c—• — I

oo1

ôooo

00

oo1

.—'

klnJ

1

Foo

a

oCO

ct

C

-a

cCO

55

tH! .

CC

ac

'ZOJ

c

TiH

if

oco

aCC

.10

C

j 2

oo

TJa;

ion

i

ei=1

c

ssio

t.t.0

b la

Inw

i

E-

5 50 5

PM

aV -a .2 aj£_§ 8

-2 "= "=

i — ce OJ

, h V H o aj

•• CO ^ • - S

- ' 3 5 CD •-' 01 _ a ^: J3 S 5 &

^ 2 ™

.2 p. M

<1 0) S

•13=*, p

ce <5E c i '

be1J

(So

rt ^

_5 C 5"

co-goC -.- ^

at^. C C- - J- . - u

— •» aI M =0

• " ca CO all> P -? O

a ft " -S

2:3 oj -S '

?.s s,s S*^ ^ 5


and —3.459 (with a one-tailed significance level of 0.0003), respectively.''Thus, year 0 provides support for the earnings management hypothesiswhereas year —1 does not.''* The Z statistic for year -1-1 is -1.228 with aone-tailed significance level of 0.109.'* Based on this test, there does notappear to be a reversal of discretionary accruals in year -1-1, which isconsistent with the descriptive statistics presented in table 3. "

Table 6 reports, by firm and industry, change in net income, changein cash flows, estimated nondiscretionary accruals, estimated discretion-ary accruals, and change in revenues scaled by lagged assets. The Wil-coxon signed-ranks test reveals that the discretionary accrual for year 0is significantly less than zero, with a significance level of 0.001. Theindustry data are presented in order to provide some information aboutthe relation between the financial variables and the ultimate ITC deci-sion. A review of the data does not result in an obvious relation betweenthe financial variables and the ITC decision. For example, the automobileindustry not only has the largest negative cash flow change, but also thelargest income-decreasing discretionary accruals, yet the industry wasnot deemed to be injured by imports (i.e., the ITC issues an unfavorabledecision). Due to the limited number of industries included in this sample,statistical testing of the relation between the various financial variablesand the ITC decision was not possible.

'' When the analysis is based on the accruals measure (rom section 4, discretionaryaccruals are also significantly negative in year 0 but the significance levels are substantiallylower ((-statistic of -2.275 with a significance level of 0.011 when 1984 is treated as year0, and -2.249 with a significance level of 0.012 when 1985 is treated as year 0). The 7.statistics are negative and insignificant for year —1 and positive and insignificant ffiryear H-l.

'*When 1985 is treated as year 0 for the footwear industry, the Z statistic tor year 0increases in absolute magnitude to —3.802 with a significance level less than 0.0001. Thisresult is consistent with the notion that the footwear industry took additional steps duringthe second investigation (1985) to decrease reported earnings. Since the Z statistic is basedon time-series regressions that control for the effect of changing revenues on accruals, thedecreases in accruals reported in table 3 do not appear to be caused entirely by decreasesin revenues.

^ When 1985 is treated as year 0 for the footwear industry, the Z statistic decreases inabsolute magnitude to —0.799 with a one-tailed significance level of 0.212.

•""Z statistics are also computed for years -5 through +1 using regression equationsestimated over periods using all available data through year —6 as the basis for obtainingthe "normal" accrual. This allowed for a wider comparison of significance levels (or the Zstatistics obtained for investigation years to those obtained in noninvestigation years. TheZ statistics for years ~5, -4 , —2, and - 1 are negative and relatively small compared totheir standard errors, with Z statistics ranging from -0.274 to -1.021. The Z statistic foryear 0 is also negative but more significant (-3.137 with a significance level of 0.0008) thanfor any of the other years, lending support to the earnings management hypothesis. The Zstatistic is positive for year - 3 but is not statistically significant. The Z statistic for yearH-l is negative and relatively small (—1.552) which does not indicate that managers reversedtheir income-decreasing accruals in year -t-l. The results of this estimation are con.sistentwith those reported in table 5.


5.2 TEST FOR MODEL MISSPECIFICATION

Scatter plots of the regression equation residuals do not exhibit anonlinear relation between abnormal accruals and changes in revenue;therefore, the negative residuals for year 0 do not appear to be the resultof this type of model misspecification. In order to obtain additionalinformation regarding possible model misspecification for periods inwhich there are large decreases in revenues, the accruals expectationsmodel is estimated for 459 firms that are not in the ITC investigationsample. The 459 firms represent all firms not in the ITC investigationsample having 25 years of data available on Compustat (1961-85). Theresiduals for 1980 through 1985 are each divided by the standard errorfrom the estimated regression model resulting in 2,754 V^ values. TheVitS are divided into deciles based on the size of the change in revenuesscaled by assets.

The first two columns of table 7 report the average change in revenuescaled by lagged assets {AREVu/Au-i) and V,; for each decile. Visualinspection of AREVu/A^t-i and V,t across the deciles does not reveal anysystematic relation between the two variables. A systematic relationbetween the two variables might indicate that the nondiscretionaryaccruals model is misspecified. Since the year 0 mean change in revenuesscaled by lagged assets for the ITC investigation sample (—0.188 fromtable 3) falls within decile 1 in table 7, it is also important to comparethe average Vu for decile 1 to all other deciles to provide some evidencethat the nondiscretionary accruals model is not misspecified for extremedecreases in revenues.

Paired comparisons for the mean V,tS by decile are computed and theresulting significance levels are reported in table 7. The question ofinterest is whether the mean Vu for the decile containing the largestdecrease in revenues (decile 1) differs from the other deciles. If decile 1is found to differ significantly from the other deciles, it could indicatetbat the accruals expectations model is inappropriate when changes inrevenues are large and negative. The results in table 7 indicate that themost significant difference between decile 1 and any of the other decilesis significant at the 0.171 level. The mean Vu for decile 1 is greater thanthe means for deciles 3 and 5 and less than the means for the otherdeciles. Decile 3 has the smallest mean V^ value and is the only decilethat differs from other deciles at significance levels of 0.10 or less. Thisanalysis provides some evidence that the significant negative Z statisticin year 0 for the ITC sample is not due to the relative inability of theexpectations model to predict accruals during periods of severe economicdownturn.

5.3 SENSITIVITY ANALYSES

Sensitivity analyses, not reported here, of the results reported in table5 were conducted to ascertain the effect various companies and industries


> ^

,9 5:

a.

..a

o•nCD

d

ol > )

d1

COc-lCO

COl> -o'1

t o

i n

•nO l

d1

i r t

d1

l O

CO0 0

d1

COi nd1

CD1X>

O

d1

i r tCO

d

00cn

o

d1

T—1

d1

!MT-H

&

IHrt

. 2

;cil

ecil

e

"3 aj

• .

ta lo

rs c-l

^ 'a.=2 S

fir

J 3U51OJ

M

"Z00

ho

po

+

+

o o o o cp o

GO C- -^ 00 rt CD uOCO C-- in CO o CO rto ^ 4 •—I --H -rt o O0 0 0 0 0 0 0

0 0 0 0 0 0 0I

C J C S O O O Of I I I

OOI I

o oI I

o o o o oI I I I I

C0rt0o o oI I I

(CCO(-J10

oo

0 t5

a if c

c -1 a

,2 . : Q. o c

S.S-Sn S «

11 11


had on the Z statistic. If firm 23 (an auto company) or the entire autoindustry is excluded from the analysis, the resulting Z statistics arenegative and significant at the 0.002 and 0.005 levels, respectively. If thefootwear industry is omitted from the sample because a second investi-gation was conducted in this industry in 1985 (year +1 in table 5), theabsolute magnitude of the Z statistic increases (—3.771) for year 0 anddecreases (—0.562) for year +1.

5.4 ALTERNATIVE TESTS

Table 8 presents results for two alternative tests. As discussed insection 3, managers may have greater incentives to manage earnings ifthey are petitioners or if the ITC investigation is being conducted underthe general escape clause. The Z statistic for the petitioning firms foryear 0 (—2.833) is smaller in absolute magnitude than for the entiresample. The lower Z statistic may be due to a smaller sample size. Thus,the average VpS (Vp) are compared to the entire sample rather than theZ statistics. In year 0, Vp is greater in absolute magnitude for thepetitioners (—0.800) than it is for the entire sample (—0.760), indicating

TABLE 8Alternative Tests of the Earnings Management Hypothesis

'Ivp Average VDescription — _ — _

Year - l " Year 0 Year +1 Year - 1 Year 0 Year 4-1

Al] companies, see table 5

(n = 23)' -0.372 -3.459 -1.228 -0.082 -0.760 -0.270Include only petitioners

(n = 14) -0.264 -2.833 -1.097 -0.074 -0.800 -0.310Include only escape clause

cases (n = 18} -0.788 -2.772 -1.058 -0.196 -0.690 -0.263

• The Zvp statistic is calculated as Sil, V^ [Sf , (T, - 3)(7', - 5 ) ^ ^ where r, is the number of yearsin the estimation period. V^ is computed as u^/(s,(l + C^*)), where Cip — [Xp(X'X)~'Xp'] in which Xis the matrix of independent variahles for the estimation period, Xp is the matrix for the predictionperiod, u^ is the prediction error, p is the prediction year, and s, is the standard error from estimates ofthe following regression model:

where:

TA,, = total accruals in year ( for firm i;AREVU = revenues in year ; less revenues in year I — 1 for firm (";

PPE,, = gross property, plant, and equipment in year ( for firm i;Ai,-, = total assets in year t — 1 for firm i;

1,1 ~ error term in year ( for firm i;i = 1 , . . , , 23 firm index;£ = 1 , . . . , T(, year index for the years included in the estimation period for firm I.

The composition of total accruals (TA,) ia as follows: TA, — [ACurrent Assets, (4) — ACash, (1)] —[ACurrent Liabilities, (5)] — Depreciation and Amortization Expense, (14), where the change (A) iscomputed between time ( and time t — 1; Compustat data item numbers are indicated parenthetically.The regression equations are estimated over all available years prior to year —1.

'' Year 0 is the year the ITC completed its investigation. In this table, year 0 for the footwear industryis 1984.

" The number of observations included in the Z statistics is n.


that discretionary accruals for the petitioners are more income-decreas-ing. When the sample is limited to general escape clause investigationfirms, both the Z statistic (-2.772) and Vp (-0.690) for year 0 are smallerin absolute magnitude. As such, the results do not indicate that thesefirms made more income-decreasing accruals than firms being investi-gated under the countervailing duty and antidumping statutes.

5.5 PORTFOLIO TEST

As noted earlier, cross-sectional correlation results in violations of theassumptions underlying the Z test statistic. Since the firms are clusteredby industry and, within industry, by time there is potential for cross-sectionally correlated accruals. One way to address the problem of cross-sectional correlation is to group the firms by industry and to analyze thediscretionary accruals for the industries. One such method of testingportfolio prediction errors is set forth in Mandelker [1974]. The residuals(i.e., discretionary accruals) from equation (2) are averaged across allfirms within an industry for each time period during the regressionestimation period.*^ The estimated standard deviation of these averageresidual terms is computed for each portfolio (industry). Average predic-tion errors are computed for each portfolio for year 0 and standardizedby the estimated standard deviation for the portfolio as follows:

/p fpf ( 6 )

where:

uSfp — average standardized prediction error for portfolio / at time p;Ufp = average prediction error for portfolio / at time p;Cf = estimated standard deviation of the portfolio / residuals;/ = portfolio (industry);p = time period.

A t-statistic is computed to test whether the average prediction errorsare different from zero as follows:

where:

uSp — average of us/,, across all portfolios / at time p;S = estimated standard deivation of uSp, (S = 1);rij = number of portfoUoa.

•" The regression equations are estimated over the same number of time periods for eachfirm within an industry, although the number varied across industries. The number ofperiods included in the estimation is limited to the number of observations for the firm ineach industry with the fewest available observations. When the same estimation period isused to compute Z statistics for each firm in an industry, the results are -0.754, -3.290,and —0.953 for year ~ 1 . year 0, and year +1, respectively (footwear year 0 is 1984). Theseresults are similar to those using the maximum number of available observations for eachfirm, which are reported in table 5.


The resulting year 0 overall i-statistic for all five industries is —5.008,whicb is significant at less than the 0.001 level. When eacb industry isomitted from the t-statistic calculation (one at a time), the resultingfour-industry t-statistics range from —3.635 to —5.035. The portfolio testsindicate that after the problems related to cross-sectional correlation aremitigated by grouping firms into industry portfolios (wbich also resultsin grouping by year), the discretionary accruals in year 0 are stillsignificantly income-decreasing. Tbe results also indicate that tbe signif-icant i-statistics are not the result of a single influential industry.

In summary, the empirical tests using total accruals indicate thatdiscretionary accruals are income-decreasing in year 0, providing supportfor the earnings management hypothesis. Discretionary accruals are notsignificantly different from zero in years —1 and -t-l.

6. Conclusions

The results of the empirical tests reported here support the earningsmanagement hypothesis suggesting that managers make income-decreas-ing accruals during import relief investigations. Discretionary accrualsare more income-decreasing during the year the ITC completed itsinvestigation (year 0) than would otherwise be expected. Tests of theearnings management hypothesis are based on firm-specific expectationsmodels used to estimate "normal" total accruals. These models allow forchanges in nondiscretionary accruals that are caused by changes ineconomic conditions. The expectations models developed here representan attempt to improve upon the measures of discretionary total accrualsused in prior research; specifically, time-series models are developed toestimate total nondiscretionary accruals and cross-sectional tests of theearnings management hypothesis are applied to the resulting discretion-ary accruals measure.

In addition to providing evidence that managers manage earningsduring import relief investigations, the results of this study may proveuseful to regulators at the ITC. The ITC may benefit by taking intoaccount the evidence provided herein that managers appear to be makingincome-decreasing accruals during import relief investigation periods. Ofcourse, tbe ITC relies on several factors in making their injury decisionswhicb may reduce tbe problem of relying on the reported earningsnumbers. ^

^ For example, injury itself is necessary but not sufficient to obtain a favorable injurydecision. The injury must be shown to have been caused by foreign imports.


I

<X

<

o

O""^

"SCL)

• 2

g

*

03

o

2td

Cd

OfPdOS

'-âi ; .

• • t

ra0)

J: ;+ j

+jra

V3

ine

a0)I D

ur

Sera

QJ "^ CDH -£3 i-C

5

4-H " r t CT"

^ -5 o

oj JH

rt . 3 ^ .iS

ra" -g a

+ j CC O) J 3

-tJ'ra

-- £

ra -a

bJD

S

ra i

O

o

otH

o

a>o

TJ

oI—I

QJ=

« "S ra S-^ ."t: +ffl > CO CO

aj *.j -J

' .S

n TO

"ra rt

oCfiCOcuI

3CJ

-a

•-. o

otDCO11

mZo

<

Cd>Z

COp<_ ]

<

w

T3

r l+ j

o

<utH

' " — ^

CO

o"

CH _ , !^ j ^

O> '""C N t4_icfl O

o o

I"^ -r)

o 13

z .25 +2

&] <


o o

3 O

c *-•

.S ffl§•£•

c3 ^

O ^ ? Ô 3 ^

^ 'E flCC c! ^be ^ "CS 'c3 ai

TO +-> - *

m fl "U3 O ô " ôj fc: flrt O C

203

a> O H-fl -d .S

oft

a -E o ^ flO O * ^ CJ ''^

^ a "cj "^ flo • - J*S > 'Si).

tJ I fl TO >>S .fl 'S ^ i3

S H S O

^ 2 «H .S

S f H x:

flCC

J3

n. tH 1

OJ

5 CJf I -.fl

O

•H +: O >•I C -rt .Jj

^ 2 "co «& a fl - s

l. r t ^ Iwr j .rt TO

•5

'-t-j .rtra be

•n fl

4H O

o aCD ' ^ ^

Cd

s a

fl.2OJ

tZ!

CC

Zo

IGA

Tl

H

zO2

-,J2

O

fl

735

(

T3flCC .

o>> ^CC .-1fl <ii-i

-IP

ING

*a

O

\NT

i:

7-

5TA

ND

I1F

OU

T!

IEW

0

ta

t

iIL

ING

IIT

ER

VJS

) C

OU

^

1f H

The

o

and

751

of 1

97£

04{b

)ts

Act

tion

s 1

eem

en

o

CO

ZO d

as

O

UR

P

<

cazCdo

a*J=l-t.i

^ - <O

(MCOCQ

.2


fl

o

fH Ol

'Z

TO S fl

CO

•S

o

Xi—i

Q2;PH

ai

Of

en p,

g Odj

<D CC

4< .ZjO Q

2 c

fl ^

fl O

a o oc3

OCJ

T 3flCC

<ucd

S o

t; oo +.

g ^ H

S 4 4

O OJ 'cc ' ^

fi fi 'H.P

^

Qfl

(- cc-2 acd cda ^

, o

il

K COOH CO

ai oo -x^5 <u

I

O

o

td

we-zHH

CdOQ-

z

IGR

AW

<OH

d

rric

• 1

CMCM

flO


APPENDIX B

Information Collected in the Producer's Questionnaire for ITC FinalInvestigations

A. Practical Annual Capacity''^B. ProductionC. End-of-Period InventoriesD. Purchases

1. Imports2. Other Purchases

E. Shipments1. Intracompany and Intercompany Transfers2. Domestic Shipments3. Export Shipments4. Amounts Made to End Users and Distributors

F. Employment and WagesG. Income-and-Loss and Other Financial Information

1. Methods of Allocation2. Income from Operations3. List of Unusual or Nonrecurring Expenses4. Asset Valuation: Original Cost and Book Value5. Capital Expenditures6. Capital and Investment^—Actual and potential negative effects of

imports on firm's growth, investment, and ability to raise capital7. Research and Development Expenses

H. Prices/. Competition from Imports—Price Suppression/DepressionJ. Competition from Imports—Lost Sales

REFERENCES

DEANGELO, L. E. "Accounting Numbers as Market Valuation Substitutes: A Study ofManagement Buyouts of Public Stockholders." The Accounting Review 61 (July 1986}:400-20.

. "Managerial Competition, Information Costs, and Corporate Governance: The Useof Accounting Performance Measures in Proxy Contests." Journal of Accounting andEconomics 10 (January 1988): 3-36.

DOBSON, J, M. Two Centuries of Tariffs: The Background and Emergence of the U.S.International Trade Commission. Washington, D.C.: Government Printing Office, 1976.

HEALY, P , "The Impact of Bonus Schemes on the Selection of Accounting Principles."Journal of Accounting and Economics 7 (April 1985): 85-107.

" Companies must report ail information by establishment, and if 85% or less of theestablishment is devoted to the product line in question, the information must also heprovided by product line. An establishment is defined by the ITC as "each facility in theUnited States in which product A is produced, including auxiliary facilities operated inconjunction with (whether or not physically separate from) such production facilities." Forpreliminary investigations, less detailed information is collected in parts B, C, D, E, and Gof the questionnaire. This information was obtained from a sample copy of a Producer'sQuestionnaire provided by the ITC.


JONES, J. "The Effect of Foreign Trade Regulation on Accounting Choices, and Productionand Investment Decisions." Ph.D. dissertation. University of Michigan, 1988.

KAPLAN, R. S. "Comments on Paul Healy: Evidence on the Effect of Bonus Schemes onAccounting Procedure and Accrual Decisions." Journal of Accounting and Economics 7(April 1985): 109-13.

. "Developing a Financial Planning Model for an Analytical Review: A FeasibilityStudy." In Symposium on Auditing Research III, pp. 3-34. Urbana-Champaign: Universityof Illinois, 1979.

KMENTA, J . Elements of Econometrics. New York: Macmillan, 1986.LIBERTY, S. E., AND J. L. ZIMMERMAN. "Labor Union Contract Negotiations and Account-

ing Choices." The Accounting Review 61 (October 1986): 692-712.LIPE, R. C. "The Information Contained in the Components of Earnings." Journal of

Accounting Research 24 (Supplement 1986): 37-64.MANDELKER, G. "Risk and Return: The Case of Merging Firms." Journal of Financial

Economics 1 (1974): 303-35.MCNICHOLS, M., AND G. P. WILSON. "Evidence of Earnings Management from the

Provision for Bad Debts." Journal of Accounting Research 26 (Supplement 1988): l-'M.PATELL, J. M. "Corporate Forecasts of Earnings per Share and Stock Price Behavior:

Empirical Tests." Journal of Accounting Research 14 (Autumn 1976): 246-Y6.PELTZMAN, S. "Toward a More General Theory of Regulation." Journal of Lau: iind

Economics 19 (August 1976): 211-40.RAYBURN, J. "The Association of Operating Cash Flows and Accruals with Security

Returns." Journal of Accounting Research 24 (Supplement 1986): 112-33.THEIL, H. Principles of Econometrics. Santa Barbara, Calif.; Wiley, 1971.UNITED STATES CONGRESS (93d). Trade Act of 1974, Public Law 93-618, January 3, 1975.UNITED STATES CONGRESS (96th). Trade Agreements Act of 1979, Public Law 96-39.

July 26, 1979.UNITED STATES INTERNATIONAL TRADE COMMISSION. Investigation Number TA-201-32.

Publication no. 905, August 1978.— . Investigation Number TA-201-44, Publication no. 1110, December 1980.~—. Investigation Number TA-201-55, Publication no. 1717, August 1985.

. Annual Report, Publication no. 1847, December 1985.

Date post:	13-May-2017
Category:	Documents
Upload:	mikeraul
View:	216 times
Download:	1 times

Jones 1991

Documents