Classifications Manipulation and Nash Accounting Standards

Transcription

1 Journal of Aountng Researh Vol. 40 No. 4 September 00 Prnted n U.S.A. Classfatons Manpulaton and Nash Aountng Standards RONALD A. DYE Reeved 19 May 001; aepted 10 Aprl 00 ABSTRACT Ths paper studes a model of lassfatons manpulaton n whh aountng reports onsst of one of two bnary lassfatons, preparers of aountng reports prefer one lassfaton over the other, an aountng standard desgnates the offal requrements that have to be met to reeve the preferred lassfaton, and preparers may engage n lassfatons manpulaton n order to reeve ther preferred aountng lassfaton. The possblty of lassfatons manpulaton reates a dstnton between the offal lassfaton desrbed n the statement of the aountng standard and the de fato lassfaton, determned by the shadow standard atually adopted by preparers. The paper studes the seleton and evoluton of aountng standards n ths ontext. Among other thngs, the paper evaluates effent aountng standards, t determnes when there wll be standards reep, t ntrodues and analyzes the noton of a Nash aountng standard, and t ompares the standards set by sophstated standard-setters to those set wth less knowledge of frms fnanal reportng envronments. 1. Introduton Fnanal reportng s, at ts roots, a proess of lassfaton: frms are gong onerns or not; transatons are reognzed n a frm s fnanal Kellogg Shool of Management, Northwestern Unversty. Paper formerly enttled Transatons Manpulaton and Nash Aountng Standards. The author thanks semnar partpants at Duke Unversty, Emory Unversty, Northwestern Unversty, Oho State Unversty, Pennsylvana State Unversty, Purdue Unversty, Stanford Unversty, the Unversty of Illnos at Chago, and the Unversty of Southern Calforna, as well as two anonymous referees, for helpful omments on prevous drafts of ths manusrpt, and the Aountng Researh Center at Northwestern Unversty for fnanal support. 115 Copyrght C, Unversty of Chago on behalf of the Insttute of Professonal Aountng, 00

2 116 R. A. DYE statements or not; leases are aptal leases or operatng leases; expendtures are assets or expenses; fnanal lams are labltes or equty, et. Many analytal studes aknowledge ths entral role of lassfaton n fnanal reportng expltly by representng the fnanal reportng proess as the produton of a partton on some underlyng state spae see, e.g., Chrstensen and Demsk [003], Demsk [1980], Dye [1985], Ijr [1975]). When ths lassfaton proess works well, fnanal reportng helps nvestors predt frms future ash flows, whh n turn affets nvestors share purhase desons. To the extent the aptal nvestors supply to frms through ther share purhases s used to fnane the frms subsequent nvestments, t follows that the lassfaton proess underlyng fnanal reportng has real resoure alloatonal effets. Ths paper ontans an examnaton of the relatonshp between ths lassfaton proess and the eonomy s produtve effeny n a model that aptures several features of the fnanal reportng envronment. Frst, the representaton of aountng standards adopted n ths paper adheres to the onventonal bnary lassfaton often found n GAAP and GAAS, and the bnary lassfaton suppresses many probablst nuanes. Eah of the examples of aountng lassfaton presented n the openng paragraph exhbts both ths bnary struture and the suppresson of probablst detals. For example, whether a frm wll reman a gong onern over a partular tme nterval s unertan untl the tme nterval passes, yet a frm s audtors ether ssue a gong onern qualfaton or they do not. Whether an expendture wll generate future benefts to a frm s often unertan n the perod the expendture s made, yet the frm must report the expendture ether as an asset or as an expense. And so on. Seond, one of the bnary lassfatons s unambguously pereved more favorably by nvestors than the other, and unertanty regardng whh of the two lassfatons s deemed approprate s resolved n the fnanal reportng proess by requrng that some typally ontext-dependent) threshold be exeeded n order to reeve the more favorable lassfaton. The examples of GAAP and GAAS n the openng paragraph llustrate ths, too. Obvously, avodng a gong onern qualfaton s preferred to the alternatve, and the lkelhood of fnanal dstress nfluenes an audtor s deson to ssue the qualfaton. Most frms prefer reognzng revenue today to not reognzng t or reognzng t wth a delay), and whether revenue gets reognzed followng a sale depends on whether the earnngs proess s onsdered to be suffently omplete and whether the proeeds from sale are measurable wth a reasonable degree of ertanty, et. Thrd, the analyss onsders the nentves of frms to engage n lassfatons manpulaton n ther attempt to seure the preferred aountng lassfaton. The fnanal reportng proess s replete wth examples of suh behavor: frms often struture lease agreements that are essentally purhase transatons so as to skrt GAAP rtera for aptal leases; frms try to nlude one-tme gans n nome from ontnung operatons; frms try to onvne ther audtors that whatever ontngent labltes they have

3 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 117 outstandng are nsuffently probable so as not to warrant nluson among ther estmated labltes, et. In the presene of these features of the fnanal reportng proess, the analyss emphaszes the dstnton between the offal or de jure) parttonng of frms or ther atvtes) desgnated by GAAP and the effetve or de fato) parttonng of frms or ther atvtes) ndued by frms lassfatons manpulaton. Classfatons manpulaton results n more frms reevng the favorable aountng lassfaton than a strt applaton of GAAP would warrant. And, sophstated nvestors adjust ther nterpretaton of a frm s fnanal statements aordngly. They reognze that the boundary delmtng the more and less favorable lassfatons enoded n a GAAP or GAAS) standard, wll as a onsequene of lassfatons manpulaton be shfted to a lower threshold, a shadow standard, demarkng the boundary between those frms who are or are not wllng to pay the ost of lassfatons manpulaton to seure the more favorable aountng lassfaton. Whle offal standards an be vewed as ether exogenous or endogenous dependng on whether one pereves the proess defnng GAAS and GAAP as exogenous or endogenous a shadow standard s always endogenous: t s the outome of an equlbrum proess nvolvng the spefaton of mutually onsstent behavor for both frms and nvestors, and t depends on eah of: the prevalng offal aountng standards, the dfferental amounts nvestors attah to frms that reeve the more or less preferred aountng lassfatons, the osts of lassfatons manpulaton, and nvestors understandng or lak thereof) of the frms produton tehnologes. The analyss demonstrates that, as nvestors learn more about frms produton tehnologes over tme, the relatonshp between the offal and shadow standards hanges. As a onsequene of ths learnng, aountng standard-setters fae a fundamental trade-off: they an hoose to hold the offal standard onstant over tme, whh auses the shadow standard to hange over tme, or they an hold the shadow standard onstant by repeatedly hangng the offal standard, but they annot hold onstant both the offal and the shadow standards. Ths trade-off appears to be new to the aountng lterature. Moreover, the paper establshes that, f standard-setters hoose to hold the shadow standard onstant, then on average they wll have to nrease the offal standard over tme. To the extent that nreases n the offal standard represented n ths paper proxy for an expanson n the set of fnanal reportng standards n GAAP, the paper provdes an explanaton for the perpetual nrease n GAAP standards. When the offal standard s onsdered endogenous, the queston arses: what are GAAP/GAAS standards desgned to maxmze? We post that standards are hosen to maxmze that expeted value of the frms subjet to the standards net of all relevant osts, nludng the osts of lassfatons manpulaton. What standards emerge as optmal are shown to depend on the extent to whh the dstrbuton of avalable projets hanges as aountng

4 118 R. A. DYE standards hange, on how well standard setters antpate the fnanal market s reaton to hanges n standards, and relatedly, on how well standard setters know the parameters of the eonomy. Sne, n prate, standard-setters may not have good knowledge of some dmensons of the eonomy, t s mportant to understand how robust the frms values are to standards that depart from those set by fully nformed standard-setters. The paper evaluates the effets of suh errors n two ways. Frst, t ontans explt alulatons of the loss n value from norretly set standards. Seond, t ompares the standards hosen by a sophstated standard setter who s fully aware of how preparers reat to a hange n standards to the standards hosen by a nave standard setter who selets standards assumng that the atons taken by preparers ths year wll be the same as ther atons last year, and hene who assumes that there wll be no reaton to a hange n standards. Whle the nave standard setter s belefs wll often be wrong that s, preparers atons ths year often wll be dfferent from the atons they hose n prevous years over tme, we show that n stable envronments, the standards hosen by nave standard setters wll onverge to a standard, dubbed a Nash standard n the followng, n whh preparers repeat ther atons over tme. Upon omparng the Nash standard to the optmal standard hosen by a sophstated standard setter dubbed a Stakelberg standard), we fnd that the Nash standard s below the Stakelberg standard. That s, we show that f standard setters do not antpate the reaton of preparers to a hange n standards, then nsuffently strngent standards emerge. Addtonal omparsons between Nash and Stakelberg standards are also made n the paper. The ontemporary aountng lterature related to ths paper s sparse. Demsk [1973], [1974] made the aountng professon aware of the dffultes n hoosng among aountng standards n general envronments where the nformaton suppled by aountng reports had wealth redstrbuton effets and where the nformaton suppled under alternatve aountng standards ould not be Blakwell-ranked. Demsk s work suggested the development of both narrower spefatons of aountng standards and the adopton of narrower effeny rtera than Pareto-optmalty; the present work represents one attempt at developng a theory of standards responsve to these onerns. Arya, Fellngham, Glover, and Shroeder [1998] have reently proposed evaluatng depreaton poles desgned to aheve effent nvestment seleton, just as the present work evaluates fnanal reportng standards from an effeny perspetve. The paper proeeds as follows: a desrpton of the base model s presented n seton. Followng that, the noton of a fnanal reportng equlbrum s presented n seton 3. The haratersts and evoluton of a fnanal reportng equlbrum are gven n seton 4. Value-maxmzng standards that assume the dstrbuton of frms subjet to the standards s fxed are desrbed n seton 5. Seton 6 presents the effets of ntrodung errors n the formulaton of aountng standards. In seton 7, haratersts of value-maxmzng standards are onsdered agan, but n ths seton,

5 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 119 the dstrbutons of frms subjet to the standards an hange as the standards themselves hange. The notons of Nash and Stakelberg standards are ntrodued and analyzed n that seton. Seton 8 summarzes the results. The appendx ontans all relevant proofs.. Model Desrpton In the base model, entrepreneur has a stand-alone projet or produton tehnology whh, for lfe yle or ash flow reasons, he wshes to sell. If the entrepreneur s produton tehnology s vable, then by nvestng $I n the tehnology after the sale, the purhasers of ths tehnology also known as nvestors n what follows) reeve β I α /α + ε n ash flows n the perod followng the sale. Here, β denotes a random produtvty parameter wth pror mean β, and ε s a mean zero error term ndependent of β. When there are multple entrepreneurs ndexed by ), we assume ε s ndependently dstrbuted aross. The salar α 0, 1) ndates the rate at whh dereasng returns to nvestment our. If the entrepreneur s produton tehnology s nonvable, then by nvestng $I after the sale, nvestors reeve no ash flows n the perod followng the sale. Produton tehnologes dffer from eah other n terms of the probablty they are vable. The probablty φ that the entrepreneur s tehnology s vable s the realzaton of some random varable φ, wth densty f φ ) and sample spae [φ l,φ u ]. Ths probablty s not observable to external nvestors, nor s t apable of beng ommunated dretly by the entrepreneur. Consstent wth the remarks made n the Introduton, we model reports produed n omplane wth an aountng/audtng standard as provdng nformaton about the realzed φ through a bnary partton on φ s sample spae. Spefally, we defne an aountng/audtng standard as a threshold probablty φ s that parttons the sample spae nto two sets W [φ l,φ s ) and B [φ s,φ u ]W worse ; B = better ). The aountng/audtng standard s brght lne n so far as t desgnates a partular probablty threshold that must be attaned to aheve the better B) lassfaton. 1 We assume that the entrepreneur s audtor, or some smlar party, admnsters the brght-lne standard: the audtor evaluates whh of the lassfatons W or B s onsstent wth the appearane of the produton tehnology, and then the audtor ssues a report of ts fndngs. Gven the report W or B), nvestors must make nferenes about φ. Ths nferene proess 1 Some prevalng aountng standards are expltly brght lne e.g., aountng for aptal and operatng leases); others are nearly so e.g., where the more lkely than not rteron s employed n the valuaton allowane for deferred tax assets, and n the aountng for nvestments, where the 0% and 50% thresholds are foal ponts, f not dspostve); and others are somewhat more vague e.g., the rtera dstngushng between ontngent and estmated labltes). Why there are suh varatons n the preson of the statement of aountng/audtng thresholds s not addressed n ths paper. I wsh to thank Lenny Soffer for suggestng the deferred valuaton allowane example mentoned above.

6 1130 R. A. DYE s omplated beause the entrepreneur has the apablty to engage n lassfatons manpulaton, that s, the ablty to alter the appearane of the produton tehnology so as to aheve the better lassfaton. We post that an entrepreneur wth a tehnology φ <φ s an alter, at ost φ s φ ) ts appearane to qualfy for the better lassfaton. Wth ths spefaton, the farther the realzed φ s below the offal standard φ s, the more ostly t s for the entrepreneur to obtan the better lassfaton. An entrepreneur bent on lassfatons manpulaton wll do so by just enough to qualfy for the preferred aountng treatment, sne the nvestng publ only observes the reported lassfaton. An mplaton of ths s that f we researhers ould observe the fraton of produton tehnologes/projets that just met the standard φ s for the preferred lassfaton, then we would see an atom.e., a postve fraton) of projets there, even f the underlyng dstrbuton of projets φ were atomless.e., no sngle pont has a postve probablty of ourrene). Suh bunhng would be dret evdene of lassfatons manpulaton. 3 Whether an entrepreneur wll hoose to engage n lassfatons manpulaton depends upon whether the benefts from t exeed the osts. If the market value of a projet lassfed as W resp., B) sm W resp., m B ), then the alulus of lassfatons manpulaton leads to the followng optmzng behavor by preparers: lassfy φ as B f m B max { φ s φ, 0 } m W ; otherwse, lassfy φ as W. Let φ be the mnmum of the probabltes φ for whh the entrepreneur fnds t advantageous to engage n lassfatons manpulaton, 4.e., φ = φ s m B m W. 1) Ths φ = φφ s ) s what was referred to n the Introduton as the shadow standard, sne t determnes the effetve parttonng of projets ndued Two ponts should be made here: frst, obvously, no entrepreneur wth a produton tehnology that s vable wth probablty φ φ s must nur a ost to seure the better lassfaton. Seond, f an entrepreneur wth produton tehnology φ<φ s does engage n enough lassfatons manpulaton so that hs frm qualfes for the better lassfaton, I assume that the manpulaton s form rather than substane,.e., the atual probablty the produton tehnology s vable remans at ts orgnal level φ. 3 Ths s onsstent wth the bunhng doumented n Degeorge, Patel, and Zekhauser [1999] and Burgstahler and Dhev [1997]. 4 Ths equaton s governng only when the threshold φ s nsde the support of φ. If φ has support [φ l,φ u ], then a omplete spefaton of φ s: f φ s m B m W [φ l,φ u ], then φ = φ s m B m W ;fφ s m B,m W >φ u, then φ = φ u ;fφ s m B m W <φ l, then φ = φ l. In prate, the boundary ases φ {φ l,φ u } wll be unmportant, sne any offal standard that ndued φ to assume one of these boundares s tantamount to a standard n whh all projets are lassfed n the same way, n whh ase the standard and assoated lassfaton of projets) serves no alloatve funton. In suh ases, a poly of havng no standard wll be weakly superor to havng any standard.

7 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1131 by the nteraton between the offal standard φ s and preparers optmzng lassfatons manpulaton. 5 Thus, a gven offal standard φ s ndues two parttons on projets, the offal one {[φ l,φ s ), [φ s,φ u ]}, and the effetve one, {[φ l, φ), [φ,φ u ]}. 6 In the followng, onsderable attenton s devoted to the shadow standard and ts assoated effetve partton, beause the eonoms of an aountng standard are not emboded n the offal standard or partton, but rather n the shadow standard and effetve partton ndued by the offal standard. Ths ours beause the alloaton of resoures, and the market pres of projets, are determned by the shadow standard, not the offal standard. 3. A Fnanal Reportng Equlbrum So far, we have dsussed how an offal standard gets transformed nto a shadow standard through an entrepreneur s optmzng lassfatons manpulatng behavor, takng as gven the market values m B and m W assoated wth the lassfatons B and W. But, these market values are endogenous and depend upon how muh lassfatons manpulaton nvestors beleve the entrepreneur has engaged n. To defne m B and m W n an nternally onsstent fashon based on exogenous varables, we must desrbe and haraterze a fnanal reportng equlbrum relatve to a gven offal standard. We do ths presently. We start by expandng on the tme lne governng the sequene of events. Frst, an entrepreneur learns the probablty φ that hs projet s vable. Antpatng the market values m B and m W assoated wth the lassfatons B and W, the entrepreneur then dedes whether to engage n lassfatons manpulaton. The fnanal report B or W) for the projet s then released to nvestors. Competton among the, presumed rsk-neutral, nvestors results n the projet beng sold for the expeted value of the future ash flows antpated to be generated by the projet s stohast produton tehnology, net of the ost of antpated future nvestment they must make n the projet, gven the reported lassfaton. After the sale, nvestors proeed wth ther prevously antpated nvestments. Fnally, the projet s realzed ash flows are generated and onsumed by the nvestors. We now dsuss the endogenous omponents of ths event sequene n detal. 5 I wsh to thank my olleague Sr Srdhar for suggestng the name shadow standard for ths threshold. 6 Whle, as the Introduton noted, aountng lassfatons n prate are often bnary, there s nothng n the model that would prevent onsderaton of parttons wth more than two elements. I would lke to thank an anonymous referee for ths observaton. An expanded analyss ould endogenze the optmal number of elements n the partton. In suh an expanded analyss, t would typally be undesrable to have a large number of elements n the partton, even f the osts of wrtng the extra lassfatons were zero, n order to eonomze on the osts of lassfatons manpulaton. These osts typally nrease as the number of elements n the partton nrease, sne the extra elements reate more opportuntes for lassfatons manpulaton.)

8 113 R. A. DYE We frst desrbe how a projet s expeted ash flows are lnked to ts lassfaton and the level of nvestment made n the projet. Suppose nvestors onjeture that the offal standard φ s wll generate the shadow standard φ. If the released report s B resp., W), then nvestors alulate the probablty that the projet s stohast produton tehnology s vable to be Prvable B, φ) = E [ φ φ φ] resp., Prvable W, φ) = E [ φ φ < φ]). Thus, from outsders perspetve, the net expeted ash flows from an nvestment of $I on a projet that reeves the lassfaton R {B, W} s: Pr vable R, φ) β I α/ α I. ) Ths requres reallng that nvestment n a nonvable tehnology produes no return.) Next, we dsuss how the nvestment of $I n a projet gets determned. Sne the projet s sold to nvestors, and t s these nvestors not the entrepreneur) who make the nvestment deson, the nvestment level $I hosen must be based on nformaton avalable to them. Investors know whh of the lassfatons B or W s assoated wth the projet they purhase, and so they an predate ther hoe of $I at least on whether φ < φ or φ φ. The followng analyss proeeds by assumng that ths s all the nformaton about φ on whh nvestors base ther hoe of $I. However, t s mportant to dgress brefly to onsder what qualtatve dfferenes would emerge n extensons of the analyss were nvestors to obtan addtonal post-sale nformaton about φ. Dgresson on when aountng standards have eonom onsequenes For any suh extenson, for arbtrary φ, ether nvestors ultmate knowledge of φ vares wth what they knew about φ at the tme of sale that s, on whether φ < φ or φ φ all ths ase I) or else ther ultmate knowledge of φ s ndependent of what they knew about φ at the tme of sale all ths ase II). An example of ase I s ths: after the sale of a projet that reeved the lassfaton B resp., W), nvestors addtonally learn whether φ s realzaton belongs to the upper or lower half of the nterval [φ, φ u] resp., [φ l, φ)),.e., they learn whh element of the partton {[φ l, φ l + φ ), [ φ l + φ, φ), [φ, φ u + φ ), [ φ u + φ,φ u ]} the realzed φ belongs to. 7 Generally, the only example of ase II of whh we are aware nvolves nvestors learnng the exat realzaton of φ. 8 7 Note that ths refned partton vares wth φ and hene s onsstent wth ase I. 8 Clearly, f followng the sale, nvestors learn the exat realzaton of φ, then ther knowledge of φ at the tme of sale namely, whether φ < φ or φ φ) does not affet ther ultmate knowledge of φ. Hene, learnng the exat realzaton of φ s onsstent wth ase II. We now argue that, generally, the onverse s also true: that s, unless nvestors ultmately learn the exat value of φ, then ther knowledge of φ wll vary wth what they know about φ at the tme of sale. The argument runs as follows. For a fxed aountng standard φ s and assoated shadow standard φ, as a onsequene of the aountng report R {B, W}, nvestors know whh

9 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1133 In extensons of the present analyss, the eonom onsequenes of an aountng standard vary sgnfantly dependng on whh of ase I or ase II s applable. If ase I s applable, then as n the analyss that follows seletng an aountng standard φ s has both dstrbutonal and alloatonal onsequenes: the hoe of φ s affets φ, whh n turn affets nvestors ultmate knowledge of φ, whh n turn affets ther hoe of $I. So, both the sellng pre of the projet a dstrbutonal effet) and the amount of nvestment n the projet an alloatonal effet) result from the spefaton of φ s n ase I. In ontrast, when ase II s applable, there are no alloatonal effets assoated wth φ s, although there are dstrbutonal effets. In that ase, the spefaton of φ s affets φ and hene affets the projet s sellng pre, but t has no alloatonal effets, sne nvestors ultmate knowledge of φ s, by defnton of ase II, ndependent of φ and hene ndependent of φ s ), and so the nvestment $I does not vary wth φ s. Thus, n ase II, there are no alloatonal onsequenes of hoosng among aountng standards. Whle some of our man results 9 wll hold for extensons of our analyss nvolvng both ases I and II, the results below regardng effent aountng standards are suggestve of the knds of results that would obtan were nvestors to aqure addtonal post-sale nformaton about φ only when, as n ase I, the nformaton gleaned from the aountng reports remans nrementally nformatve relatve to whatever addtonal post-sale nformaton nvestors aqure THE DEFINITION OF EQUILIBRIUM Returnng to the model, the effent alloaton of aptal, based on nvestors knowledge of the returns on nvestment, requres maxmzng the net expeted ash flows dsplayed n ). The net expeted ash flows from the projet based on ths antpated nvestment then determne the market pres m W and m B. Fnally, these market pres are lnked to the equlbrum amount of lassfatons manpulaton through 1). When all of these element of the two element partton ={[φ l, φ), [φ,φ u ]} the realzed value of φ falls n. Suppose, as a onsequene of nformaton the nvestors dsover post-sale, nvestors also learn that φ s realzaton falls n some element of the partton ={[φ 0,φ 1 ), [φ 1,φ ), [φ,φ 3 ),..., [φ n 1,φ n ),...}. Clearly, the nformaton ontaned n the orgnal aountng report s redundant n the presene of ths post-sale nformaton f and only f,.e., f and only f s fner than partton. Equvalently, the nformaton ontaned n the orgnal aountng report s redundant f and only f φ = φ for some. If φ φ for any, we are done: The aountng lassfaton s nrementally nformatve relatve to. But, φ = φφ s ) annot equal φ for any generally, sne φ φ s ) > 0.e, there an be only a fnte number of φ s for whh φφ s ) = φ for any ). Ths proves the lam. 9 In partular, eah of the followng ontnues to be vald: the dstnton between a shadow standard and the offal standard, the mportane of lassfatons manpulaton, and the drft over tme n the relatonshp between the offal and shadow standards depted n Theorem below. 10 Case I s the pratally relevant ase when nvestors annot aqure exat nformaton about ther produton tehnology and hene φ s realzaton) followng the tehnology s sale.

10 1134 R. A. DYE ondtons hold onurrently we obtan a fnanal reportng equlbrum. Formally: DEFINITION 1. A fnanal reportng equlbrum relatve to aountng standard φ s onssts of a par of nvestment levels I B) and I W) and market values m B and m W and a shadow standard φ suh that: ) ) ) v) An entrepreneur whose projet s vable wth probablty φ engages n lassfatons manpulaton f and only f φ φ; Prvable B, φ) = E [ φ φ φ] and Prvable W, φ) = E [ φ φ < φ] For R {B, W}, the nvestment I R) attans the maxmum of: max Prvable R, φ) β I α/ α I. I For R {B, W}, the market value m R s gven by: m R = Prvable R, φ) β I R)) α/ α I R); v) φ = φ s m B m W. In bref, a fnanal reportng equlbrum onssts of a par of nvestments and market values and a shadow standard determnng how projets are lassfed, gven the prevalng aountng standard, so that: nvestment s effent gven nvestors knowledge; pres of projets are set orretly; preparers optmze when engagng n lassfatons manpulaton. 4. Charatersts and Evoluton of a Fnanal Reportng Equlbrum We start the formal analyss wth a more explt desrpton of the relatonshps among the equlbrum nvestment levels, market values, and the shadow standard. We then follow ths wth an analyss of how the fnanal reportng equlbrum evolves over tme. A smple alulaton shows that, when the aountng lassfaton s B, the optmal level of nvestment, I B) s gven by: I B) = β E [ φ φ φ]) 1 1 α, 3) whh mples that the projet has market value 11 m B = E [ φ φ φ] β I B) α/ α I B) = 1 α α β E [ φ φ φ]) 1 1 α. 4) 11 The omputaton of the market value m B runs as follows. The frst-order ondton for I B) s E [ φ φ φ] β I B) α 1 = 1. So,

11 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1135 Smlar alulatons for a projet that reeves the lassfaton W yeld: and I W) = β E [ φ φ < φ]) 1 1 α 5) m W = 1 α α β E [ φ φ < φ]) 1 1 α. 6) The equaton desrbng the shadow standard 1) lnks these two market pres together wth the ost of lassfatons manpulaton: φ = φ s 1 α α β 1 1 α E [ φ φ φ] 1 1 α E [ φ φ < φ] 1 1 α ). 7) We shall show below that, even when ths equaton annot be solved expltly, muh an be learned about the relatonshp between the offal and shadow standards. However, to exhbt a fnanal reportng equlbrum expltly, t s useful to parameterze the model n a way whh admts an explt soluton to ths equaton. We parameterze the problem by assumng α =.5 and that φ s unform on [l, u],.e., f φ ) = 1 u l, φ [l, u]. 1 For ths parameterzaton, we onlude: THEOREM 1 The Relatonshp Between Offal and Shadow Standards). Assume α =.5, φ Unform[l, u], and that the offal standard φ s s suh that the shadow standard φ s nteror.e., φ l, u)). Then, a fnanal reportng equlbrum relatve to aountng standard φ s and dstrbuton φ exsts, s unque, and s partally) desrbed by the shadow standard: and the market pres: m W = β l + φ φ = φs β u l β u l) ) and m B = β 8) ) u + φ. 9) The expresson 8) relates the threshold φ to the standard φ s, the ost of lassfatons manpulaton, the parameters of the dstrbuton of φ, and the unknown produtvty parameter β. Note that φ<φ s unless the ost of lassfatons manpulaton s nfnte, so the set of projets that reeve the m B = E [ φ φ φ] β I B) α /α I B) = E [ φ φ φ] β I B) α 1 /α 1) I B) = 1/α 1) I B) = 1 α α β E [ φ φ φ]) 1 α 1. The omputatons for m W are smlar. 1 We wrte ths n the followng as φ Unform[l, u].

12 1136 R. A. DYE better lassfaton B s always overstated relatve to what a strt applaton of GAAP would warrant, as suggested above. Whle omparatve stats an be performed dretly on the shadow standard φ, the omparatve stats that are of most nterest employ ths shadow standard to alulate the probablty that a projet wll reeve a partular lassfaton. We state these omparatve stats n the next orollary, after rewrtng the shadow standard n terms of the mean and standard devaton of φ u + l s dstrbuton. Wth µ = and σ = u l, the shadow standard an be 3 wrtten as φ = φs β σ 3 µ 1 + β σ. 10) 3 COROLLARY 1 13 The Comparatve Stats of Shadow Standards). Mantan the assumptons of the preedng theorem. The probablty that a projet wll reeve the lassfaton B resp., W ) s ) dereasng n the standard φ s and the ost of lassfaton manpulaton, and ) nreasng n β and µ. Most of these results are ntutve: as the standard φ s goes up, or as the ost of lassfatons manpulaton nreases, then the probablty of gettng the more favorable aountng treatment delnes. Also, as the expeted value of the produtvty parameter β nreases, or as the pror probablty µ that a projet s vable nreases, then the returns to the preferred lassfaton also nrease sne m β B m W ) > 0 and µ m B m W ) > 0) and so preparers wll engage n lassfatons manpulaton more often.e., for a bgger set of realzed φ s). No omparatve stat s reported nvolvng hanges n the standard devaton of φ on the probablty that a projet s lassfed as B, sne that effet s generally ambguous. 14 Next, we onsder the preedng model and assoated equlbrum as a snapshot of a mult-perod model n whh, n any gven perod, the urrent generaton) entrepreneur and urrent generaton) nvestors get to wtness and learn from what happened n pror perods when other projets were transferred between prevous generatons of entrepreneurs and nvestors. In ths settng, t s natural to nqure how the fnanal reportng equlbrum evolves over tme. We study the statonary ase, where the dstrbuton of φ s ndependently and dentally dstrbuted over tme, and wth the possble exepton of the produtvty parameter β all other parameters of the model the ost of lassfatons manpulaton, and the preferenes of the entrepreneurs and nvestors) reman onstant. The followng theorem 13 The proof of the orollary nvolves smple algebra and s not nluded n the appendx. 14 But, f the standard φ s s at or below the mean probablty µ that a frm s vable, one an onlude that the probablty a frm wll be lassfed as vable s dereasng n the standard devaton σ.

13 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1137 apples n both the ompletely statonary ase where β s onstant over tme, as well as n the partally statonary ase where β vares over tme but has some statonary omponent. To be spef, n the partally statonary ase, t suffes that there be no nformaton avalable at tme t better than the tme t value of produtvty parameter that s helpful n predtng the produtvty parameter at any tme t > t and that the tme seres β t be a martngale. More formally, f ϑ t represents all that an be known about the tme seres of the produtvty parameter up to tme t, then E [ β t ϑ t,β t ] = β t for all t > t. The followng theorem demonstrates that, even wth all ths statonarty, the relatonshp between the shadow standard and the offal standard s not onstant over tme. The theorem shows that f standard setters want to hold the shadow standard onstant over tme perhaps beause they prefer onssteny n the lassfaton of projets aross tme), then on average they wll have to nrease the offal standard over tme. That s, there must be standards reep. THEOREM Standards Creep). If the offal standard φ s s hosen n eah perod so as to ndue the same shadow standard φ and the envronment s partally or ompletely) statonary, then the expeted value of the offal standard φ s must be nreasng over tme. Ths result s suggestve of a fundamental trade-off n desgnng aountng standards: aountng standards an be hosen so that the fae value of the standard φ s s ntertemporally onstant, or aountng standards an be hosen so that the ndued shadow standard φ s ntertemporally onstant, but not both. The result s somewhat surprsng gven the assumed statonarty n the envronment. To obtan ntuton for the result, we wll onsder n the text the speal ase where α =.5 and β s ompletely statonary though the result also holds for any α 0, 1) and for the ase of tme-varyng β t s too). In ths ase, the equaton 7) defnng the shadow standard redues to: φ = φ s 1 β E [ φ φ φ] E [ φ φ < φ] ). 11) It s apparent from 11) that the shadow standard φ prevalng n a perod s affeted by the market s pereptons of β only through β. In partular, f β nreases, then the offal standard must be rased to keep the shadow standard onstant. 15 Now, onsder what happens wth the passage of tme. As tme evolves, suessve generatons of nvestors wll aqure more and more prese nformaton about the value of the unknown produtvty parameter β. Ths learnng about β wll our sne the ash flows generated by eah vable projet, β I α /α + ε, provde nvestors wth ndret nformaton about 15 Sne E [ φ φ φ] E [ φ φ < φ] > 0.

14 1138 R. A. DYE β s realzed value. To desrbe ths formally, we reall the notaton ϑ t ntrodued above. It represents all that wll be known about the produtvty parameter β up through perod t. If today s perod 0, then ϑ t s unertan for any t > 0 whereas ϑ 0 s known. In ths notaton, today s resp., perod t s) expeted value of the produtvty parameter β s denoted by E [ β ϑ 0 ] β resp., E [ β ϑ t ]). In any perod t > 0, the offal standard φt s and the shadow standard φ t prevalng n perod t wll be onneted to eah other through the followng ounterpart to 11): φ t = φt s 1 α E [ β ϑ t ]) E [ φ φ φ α t ] E [ φ φ < φ t ] ). 1) Analogous to what 7) demonstrated n perod 0, equaton 1) demonstrates that the shadow standard φ t prevalng n perod t s affeted by the market s pereptons of β at that tme only through the expetaton E [ β ϑ t ]). In fat, 1) tells us more: t tells us that to hold the shadow standard onstant as E [ β ϑ t ]) nreases, the offal standard φt s must nrease. Ths fat drves the theorem: sne more and more nformaton aumulates about β as tme passes, the expeted ondtonal varane of β alulated today), E [Var β ϑ t ) ϑ 0 ], wll derease the further nto the future we look.e., the hgher t s). That s, E [Var β ϑ t ) ϑ 0 ] = E [ E [ β ] ϑ t E [ β ϑ t ] ] ϑ0 = E [ β ] [ ϑ0 E E [ β ϑ t ] ] ϑ0 dereases as t nreases, and so E [E [ β ϑ t ] ϑ 0 ] nreases as tme passes, and hene the offal standard φt s must be adjusted upwards to keep φ onstant. Thus, the antpaton of learnng about β s responsble for an nrease n the expeted value of the offal standard over tme. Moreover, note that ths effet wll be renfored f the ost of lassfatons manpulaton were to delne over tme, perhaps beause of the nformaton dffused through nvestment bankers about how a preferred aountng treatment an be aheved I would lke to thank Mary Barth for the observaton that the ost of lassfatons manpulaton mght, under some rumstanes, nrease over tme. Whle the dffuson of nnovatons by nvestment bankers aross frms would tend naturally to redue over tme as the text mentons), nreased sutny by SEC offals of prates that they onsdered to be volatve of the sprt of some standards ould ause these osts to nrease over tme. Of ourse, f ths happened, then the nteraton between the learnng effets dsussed n Theorem and hanges n would lead to ndetermnant effets on the tme seres evoluton of the offal standard φ s. I also want to thank Lenny Soffer for hs observaton that, apart from the behavor of nvestment bankers, one mght expet to delne over tme beause of frms onstantly pushng the envelope to determne what aountng treatments qualfy as beng n aordane wth GAAP. One a frm, or a olleton of frms, dsovers that no objeton gets rased to what prevously was regarded as an aggressve aountng treatment of some transaton, then that

15 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS Value-Maxmzng Standards When the Dstrbuton of Avalable Projets Is Fxed The analyss so far has taken the offal standard φ s as exogenously gven. In ths seton, as well as throughout muh of the remander of the paper, we onsder endogenzng the hoe of the offal standard. In makng the offal standard endogenous, we must endow the standard setters empowered to hoose the standard wth an objetve funton. In ths seton, we presume that standards are hosen to maxmze the expeted value of a projet, before the projet s lassfed, net of the expeted ost of lassfatons manpulaton, whle holdng the dstrbuton of projets fxed. 17 Choosng a standard to maxmze ths objetve n part nvolves tradng off the osts of two knds of errors: mslassfyng a vable projet as nonvable, and mslassfyng a nonvable projet as vable. The standard setters must also aount for the relatve frequeny of vable and nonvable projets n the populaton, as well as the ost of lassfatons manpulaton. One agan we restrt attenton to the parameterzatons α =.5 and φ Unform[l, u]. The expeted value of a projet net of the ost of lassfatons manpulaton s gven by: Pr φ φφ s ) ) m W + Pr φ > φφ s ) ) m B φ s φφ s ) φ s φ ) f φ )dφ. 13) The last term s the expeted ost of lassfatons manpulaton. When φφ s ) l, u), these expeted osts ntegrate to: 18 u l) β 4 φφ s ) + u + l ). 14) 8 Sne the shadow standard φφ s ) nreases n φ s, we see that the expeted osts of lassfatons manpulaton nrease as the standard φ s nreases. It aggressve treatment subsequently beomes the benhmark aganst whh even more aggressve reportng of transatons wll be ompared n the future. 17 In later setons, we allow the dstrbuton of projets to hange as the aountng standard hanges. 18 Reall from Theorem 1 that φ s and φ are related through the equaton φ 1 + β ) u l) + β 4 u l ) = φ s. Hene, we an express the expeted ost of transatons manpulaton as: 1 φ s u l φ φs φ)dφ = u l) φs φ) = = u l) β 4 8 φ + u + l ). ) β φ u l) + β ) u l) 4 u l )

16 1140 R. A. DYE an be shown that these osts are also always nreasng n the produtvty parameter β. 19 Further, the expeted ost of lassfatons manpulaton s never monoton n. 0 The graph n fgure 1 below llustrates how these osts hange as the parameter hanges. 1 The offal standard φ s and the assoated shadow standard φφ s ) that maxmze the expeted value of a projet net of the expeted ost of lassfatons manpulaton are detaled n the followng theorem. THEOREM 3 Value-Maxmzng Standards when the Dstrbuton of Projets s Fxed). When α =.5 and φ Unform[l, u], and the standard φ s s hosen to maxmze the expeted value of a projet net of the expeted ost of lassfatons manpulaton, then ) when > 3l + u) β, the optmal shadow standard s φ = u + l u l) β + u l) β = µ 1 3σ β 1 + 3σ β and the offal standard assoated wth ths shadow standard s φ s = u + l = µ; ) the net expeted value of the projet s strtly nreasng n the ost of lassfatons manpulaton, for > 3l + u) β. ) When 3l + u) β, the optmal standard s no standard,.e., φ = φ s = l. Several parts of the theorem are of nterest. Frst, one mght expet that, sne the probablty a projet s vable s unformly dstrbuted, the optmal standard would result n half the projets reevng the lassfaton B and half reevng the lassfaton W. But, aordng to the theorem, even though φ s = µ, more than 50% of all projets are lassfed as B. Ths follows sne the reported lassfatons are determned by the shadow standard φ, not the offal standard φ s, and φ<µ.the explanaton for ths asymmetry s the presene of, and the aountng for, the osts of lassfatons manpulaton n the standard setters objetve funton. Sne the osts of lassfatons manpulaton are nurred only when the better B) lassfaton s reeved, the optmal standard that nets out these osts of lassfatons manpulaton s set below the mean. It an be shown that, were the objetve u l 19 That s, u l) β 4 8 φs β β u l) + u + l ) s nreasng n β. 0 Ths last lam s easest to see by wrtng the expeted ost of transatons manpulaton dretly n terms of φ s. Omttng the algebra, these osts an be shown to be expressable as u l) β 4 8 φs + u + l 1 + β u l) ). Now, for any postve fnte, under the preedng ondtons, these expeted osts are postve. Sne, as approahes 0, the osts approah zero, and as approahes nfnty, the osts also approah zero, the nonmonotonty follows. 1 I would lke to thank Madhav Rajan for pontng out an error n the onstruton of ths graph n a prevous verson of ths manusrpt.

17 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1141 FIG. 1. Plot of expeted ost of lassfatons manpulaton as a funton of ; φ [0,1]; φ s =.5; β = 1. Unform desgned to maxmze the expeted value of a projet gross of these expeted osts of lassfatons manpulaton, then the optmal shadow standard s set at the mean, and the offal standard s set above the mean. But, sne the osts of lassfatons manpulaton are real osts, standard setters should aount for them when desgnng standards, and so, other thngs equal, standards should be set n part to eonomze on these osts. Seond, t s nterestng to note from ) that, even though the standard s hosen to maxmze the expeted value of a projet net of the ost of lassfatons manpulaton, the ost does not enter nto the determnaton of the optmal standard φ s, as long as the ost exeeds 3l + u) β. However, as part ) reports, when the ost s low 3l + u) β ), standards serve no funton. For suh low s, nontrval standards φ s > l) generate osts of lassfatons manpulaton, yet they fal to dsrmnate among projets: all entrepreneurs, regardless of ther projet s realzed φ, hoose to have ther projet lassfed as B. Ths may be the eonoms underlyng SFAS and other standards that purposefully do not dsrmnate between eonomally dfferent expendtures. Part ) observes that nreases n the ost of lassfatons manpulaton always nreases the net expeted value of a projet. Whle, holdng projets market pres fxed, entrepreneurs may prefer to redue the ost of obtanng the better lassfaton, ex ante they are always worse off by havng these osts delne, sne the market wll antpate ther subsequent explotaton of the redued osts of gettng the preferred lassfaton. Hgh osts of lassfatons manpulaton.e., hgh s) ommt entrepreneurs to engage n less manpulaton, whh s ex ante effent. The theorem also yelds several omparatve stats. Aordng to the theorem, the optmal standard φ s equals the expeted probablty that a projet s vable, and so t nreases as ths mean probablty nreases. φ s s otherwse ndependent of the parameters of the fnanal reportng envronment, a

18 114 R. A. DYE fat that wll be mportant n the analyss of errors n standard setter s belefs n the next seton. Fnally, the optmal shadow standard φ s: below the mean probablty that a projet s produton tehnology s vable; nreasng n the mean and dereasng n the standard devaton of the dstrbuton of φ ; and dereasng n the expeted value of the produtvty parameter β. These are all testable mplatons of the theorem. Before onludng ths seton, we dsuss the robustness of the results n ths seton to varatons n the dstrbuton of φ. Sne the unform dstrbuton s a speal ase of the beta dstrbuton, one way to evaluate the robustness of the above results s by omparng them to orrespondng results generated by other beta dstrbutons. Whle members of the beta famly other than the unform dstrbuton are dffult to deal wth analytally, numeral plots of these ases provde some nsght nto these robustness questons. Fgures a d plot four denstes drawn from the beta lass, and Fgures 3a 3d plot the net expeted value of a projet net of the expeted ost of lassfatons manpulaton) as a funton of the shadow standard that de fato parttons the better B) and worse W) aountng lassfatons. The beta dstrbutons depted n Fgures a and b are symmetr, and as areful sutny of Fgures 3a and 3b reveals, the optmal shadow standards orrespondng to these symmetr dstrbutons are lose to, but slghtly below, the ommon) mean.5) of these dstrbutons. In ontrast, the beta dstrbuton n Fgure resp., Fgure d) s skewed rght resp., left), and upon areful srutny, Fgure 3 resp., Fgure 3d) reveals that the optmal value of the shadow standard s lose to but to the left resp., rght) of the dstrbuton s mean whh s 5/7 resp., /7) for the dstrbuton n Fgure 3 resp., Fgure 3d)). Generalzng from these fgures, t appears that a useful heurst s to have the shadow standard lose to the mean of φ s dstrbuton, and that f φ s ether symmetr or skewed rght, then the optmal value of the shadow standard s slghtly below the dstrbuton s mean, onsstent wth the results presented above: n these ases, the ost of lassfatons manpulaton pulls down the shadow standard below the mean of φ s dstrbuton. But, f φ s dstrbuton s skewed suffently left, then, notwthstandng the extra ost of lassfatons manpulaton t ndues, the optmal shadow standard and, a fortor, the optmal offal standard) may be to the rght of the dstrbuton s mean. 6. Errors n Spefyng Standards The analyss n the preedng seton assumed that standard setters know the eonom envronment n whh the sale of projets was onduted: they know the parameters of the dstrbuton u and l) generatng vable projets; they know the ost of lassfatons manpulaton; they have the same belefs about the expeted value of the produtvty parameter as do other market partpants, et. Whle the atual partpants n an eonomy have strong fnanal nentves to aqure suh a sophstated understandng of the envronment n whh they work, t s not lear that standard

19 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1143 FIG.. Plots of beta denstes for varous parameterzatons. setters have orrespondng nentves. Indeed, one mght argue that many lobbyng efforts are desgned to prevent standard-setters from aqurng suh knowledge. It s mportant, onsequently, n any dsusson of standards, to onsder the robustness of the standards wth respet to errors

20 1144 R. A. DYE FIG. 3. Plots of net expeted value of projet as a funton of the shadow standard φ for varous beta dstrbutons when = 5 and β = 1. n the standard setters knowledge. Ths seton dsusses these robustness questons. To evaluate errors n standards, we ontnue to assume that market partpants preparers and nvestors) know the orret values of the eonomy s

21 CLASSIFICATIONS MANIPULATION AND NASH STANDARDS 1145 parameters, and we onsder what happens when standards are set norretly. That s, regardless of how standard setters arrve at the spefed offal standard φ s, the shadow standard φ, and the sellng pres m W and m B are set ratonally aordng to 8) and 9). We further assume that, as before, standard setters hoose standards that, based on ther nformaton, are pereved to maxmze the expeted value of a projet net of the expeted ost of lassfatons manpulaton. That s, f the standard setters beleve φ Unform[l e, u e ], then they set φ s at φ s = ue + l e. In ths settng, note that standard setters an reman gnorant or wrongheaded onernng both the ost of lassfatons manpulaton and the expeted value β of the produtvty parameter β, and t makes no dfferene to the onstruton of the standard. Implementaton of the optmal standard s robust wth respet to these knds of errors. And, f the standard setters are wrong about the support of the dstrbuton,.e., about l e and u e, the only sense n whh ther error matters to the formaton of the standard s the error n the mean: e = ue + l e u + l. To evaluate the onsequenes of the error e, we alulate the dfferene n the maxmum expeted value of a projet had no error n standards spefaton ourred to the maxmum expeted value of a projet n the presene of ths erroneously set standard. We refer to ths dfferene as the absolute robustness of a standard subjet to error e. The relatve robustness of a standard subjet to error e s the rato of ths dfferene to the maxmum value of a projet when the standard s onstruted wth no error.) THEOREM 4. The Robustness of Standards to Errors n Standard Setters Belefs.) Suppose α =.5 and φ Unform[l, u]. If the objetve n settng a standard s to maxmze the expeted value of a projet net of the expeted ost of lassfatons manpulaton, then the robustness of a standard wth error e = ue + l e u + l s gven by: error absolute robustness relatve robustness e β e 4 + β u l)) 8e 5l + 6lu+ 5u ) β u l)3. Note that the robustness of a standard s quadrat n the error e. Ths has two mplatons. Frst, the frst-order effet of small errors s d β zero n both absolute and relatve terms.e., e de 4 + β u l)) e=0 = 0 and d 8e de 5l + 6lu+ 5u ) β u e=0 = 0). Seond, the dreton of the error does not matter: overestmates l)3 of the mean u + l are equally ostly as underestmates and ve versa). 3 It an also be shown that the absolute resp., The proof of ths theorem, whh nvolves smple, but lengthy, algebra manpulatons, s omtted. 3 Pror to obtanng ths result, one mght have guessed that the presene of and onern over lassfatons osts mght have ndued an asymmetry between the effets of overstated errors and understated errors: errors of overstatement lead to hgher standards whh n turn would seem to lead to hgher lassfatons osts, whh would seem to make overestmates of standard setters more ostly than underestmates.

22 1146 R. A. DYE relatve) robustness of a standard s dereasng resp., nreasng) n β,so larger values for the produtvty parameter aentuate the loss n the expeted value of a projet due to an erroneously set standard, but ths loss dereases n perentage terms. In ontrast, t an be shown that nreases n the lassfatons ost parameter adversely affet both the absolute and relatve loss due to errors n a standard s spefaton. 7. Value-Maxmzng Standards When the Dstrbuton of Avalable Projets Is Endogenous So far, the paper has examned determnants of effent aountng standards, the evoluton of standards, and the effets of errors n standards n a settng where the real resoure alloaton effets of standards derve from ther mpat on the sze of the nvestments n the set of avalable exstng projets. Ths mght be onsdered an analyss of the ex post effets of standards, sne the set of avalable projets s taken as gven that s, not responsve to hanges n the aountng standards. But, n prate, we mght expet there to be ex ante effets of hanges n standards as well, sne the behavor of entrepreneurs and others nvolved n the produton of projets, IPOs, new ventures, and the lke may respond to hanges n aountng standards. Ths happens n prate: What deals spn-offs, takeovers, nvestments, leases, et.) are ompleted are affeted by prevalng aountng standards, and as standards hange, both the knd and struturng of these deals hange. We now formally aount for these responses by allowng the dstrbuton of projets to hange as aountng standards hange. The standards that emerge n ths settng depend on the extent to whh standard setters antpate the dependene of the now) endogenous dstrbuton of projets on the hosen standards. We wll onsder two types of standard setters: sophstated also known as Stakelberg leaders ) and nave also known as, followers ). A sophstated standard setter orretly antpates how preparers wll respond to hanges n standards. Neessarly, a sophstated standard setter knows all the detals of the eonom envronment n whh standards are set. In ontrast, a nave standard setter s presumed to know nothng about the eonom envronment other than what an be dedued from observng preparers past behavor. Moreover, a nave standard setter beleves that preparers do not alter the dstrbuton of ther projets n reaton to a hange n standards. More detal onernng the behavor of both nave and sophstated standard setters wll be presented below. Gven these assumed dfferenes n sophstaton, t s lear that f the only purpose of the analyss were to ompare the expeted performane of these two types of standard setters, then there would be no need for a formal analyss. Whatever the objetve used to alulate the performane of standards, a nave standard setter wll never selet a better standard than