Warranty Reserve: Contingent Liability, Informational Signal, or Earnings Management Tool? *

Transcription

1 Warrany Reserve: Coningen Liabiliy, Informaional Signal, or Earnings Managemen Tool? * Daniel Cohen a, Masako Darrough b, Rong Huang c, Tzachi Zach d a Sern School of Business, New York Universiy dcohen@sern.nyu.edu b Zicklin School of Business, Baruch College Masako_Darrough@baruch.cuny.edu c Zicklin School of Business, Baruch College Rong_Huang@baruch.cuny.edu d Fisher College of Business, The Ohio Sae Universiy zach_7@fisher.osu.edu Firs Draf: November 2007 This Draf: January 2009 * We are graeful for commens received from seminar paricipans a Fordham Universiy, George Washingon Universiy, INSEAD, Massachuses Insiue of Technology, Temple Universiy, Universiy of Rocheser, Washingon Universiy in Sain Louis, he 2008 Four-School conference a Baruch College, and he 2008 Buron workshop a Columbia Universiy.

2 Warrany Reserve: Coningen Liabiliy, Informaional Signal, or Earnings Managemen Tool? Absrac Uilizing a daabase ha became available due o he requiremens of FIN 45, we examine he informaional role of accouning disclosures on warranies. Firs, since firms use warrany policies as a business sraegy o promoe heir producs, a warrany reserve may serve wo roles: an informaional signal regarding produc qualiy as well as a coningen liabiliy. Consisen wih his view, we find ha he sock marke recognizes ha: (1) he warrany reserve conains informaion abou firms fuure performance, and (2) he reserve is a liabiliy. Second, since warrany accruals require esimaion of fuure claims, hey can also be used as a ool of earnings managemen. Our evidence indicaes ha managers use warrany accruals o manage earnings opporunisically o mee heir earnings arges. Finally, we find ha he sock marke recognizes ha warrany liabiliies of firms ha managed earnings are underesimaed. Keywords: Warranies, Coningen Liabiliy, Produc Qualiy, Signaling, Earnings Managemen

3 1 1. Inroducion Mos durable producs are sold wih warranies. 1 A warrany provided by a manufacurer/vendor guaranees is cusomers ha a produc will provide expeced service; in he even of failure, he warrany provider would recify he produc according o he erms of he warrany policy, which can vary in duraion and scope (full or limied, labor and/or pars, repair vs. refund, ec.). A warrany is an effecive means for reducing uncerainy abou he produc s fuure performance. The role of warranies in resolving informaion-based problems has been sudied exensively in he economics (e.g., Spence, 1977, Grossman, 1981, and Luz, 1989) and markeing lieraure (e.g., Menezes and Quelch, 1990). Under informaion asymmery, manufacurers, who possess beer informaion abou a produc s expeced performance, issue warrany plans o signal produc qualiy. In he presence of imperfec informaion regarding he fuure performance of he produc, even wihou any informaion asymmery, warranies can be a means of insurance for riskaverse buyers agains produc failure (Heal, 1977). The seller may specify he condiions under which warranies are effecive, hereby encouraging he proper usage of he produc in he presence of moral hazard. Finally, warranies may be used o screen consumers of differen ypes (e.g., risk aversion) so ha he seller can price discriminae hem effecively. The accouning aspecs of produc warranies have ye o be sudied. In his paper, we fill his void in he lieraure by invesigaing he role of warrany informaion. We use a unique and comprehensive daabase of warrany disclosures ha has no been available o researchers unil recenly. Alhough firms were a libery o disclose warrany informaion volunarily, FIN 45, which ook effec saring in 2003, mandaed he disclosure of such informaion. We sudy a sample of Mos producs are sold wih eiher an express or implied warrany. An express warrany is ypically specified by a wrien warrany policy ha spells ou he erms of warrany, while an implied warrany is an implici undersanding ha he produc being sold mees he warrany of merchanabiliy, i.e., fi for sale and consumpion as represened a he ime of sale. An exended warrany may be offered by reailers for an addiional premium.

4 2 firms which disclosed quarerly warrany informaion from 2003 o We also hand-collec informaion on warrany duraions from firms annual repors for a subsample of 159 firms. Our research quesions are wofold. Firs, how does he marke inerpre accouning informaion on warranies? Specifically, we ask wheher he capial marke inerpres warrany reserves as a coningen liabiliy, an informaional signal, and/or an earnings managemen ool. 2 Second, how do managers make accrual choices regarding warrany expenses and liabiliies? Our firs research quesion examines he marke valuaion of warrany reserves. 3 If firms provide warranies as insurance, warrany liabiliies are simply coningen liabiliies: fuure obligaions o perform service if a produc fails. One dollar of warrany liabiliies is expeced o reduce firm s value by one dollar. The value of insurance is presumably capured by increased demand for he produc. However, if firms offer warranies o signal produc qualiy, warrany liabiliies can have an addiional role in providing informaion on firm value and fuure firm performance. Due o his dual naure of warrany liabiliies, we expec hem o differ from oher moneary liabiliies such as bank loans. Our empirical analysis demonsraes ha he sock marke values warrany liabiliies differenly from oher liabiliies, by placing a smaller negaive valuaion coefficien on warrany liabiliies. However, afer conrolling for analys earnings growh expecaions and he duraion of warranies, he valuaion coefficiens on boh warrany liabiliies and oher liabiliies approach negaive one. This is consisen wih he marke inerpreing warrany liabiliies as informaional signals for fuure earnings growh prospecs. 2 Throughou he paper we use he erms warrany reserve/s and warrany liabiliy/ies inerchangeably. 3 Several sudies have generally documened a negaive relaion beween oher ypes of liabiliies and marke prices (e.g., Barh, 1991; Espahbodi e al. 1991; Landsman, 1986; Mielsaed and Warshawsky, 1993; Barh and McNichols, 1994).

5 3 Our second research quesion invesigaes wheher managers exercise discreion over warrany accruals. In making accrual choices, managers may incorporae informaion abou warrany policies, or alernaively, engage in opporunisic earnings managemen. Since a warrany policy is par of an overall business sraegy, managers accouning choices regarding warranies may reflec produc qualiy and may be correlaed wih fuure performance. In he accouning lieraure, such managerial discreion has someimes been viewed as a ool o improve he informaiveness of accouning numbers (e.g., Was and Zimmerman, 1986; Bernard and Skinner, 1996; Subramanyam, 1996, among ohers). We find a significan posiive relaion beween abnormal warrany expenses, fuure sales growh, and fuure reurn on asses. This finding suggess ha firms incorporae informaion abou warrany policies, which ranslaes ino fuure firm performance, ino he warrany reserves. In addiion, we documen a posiive sock marke reacion o abnormal warrany expenses around earnings announcemens. Alhough hese resuls only represen associaions, ogeher, hey are consisen wih he hypohesis ha he marke incorporaes warrany informaion in a manner consisen wih he signaling model. Alernaively, managers migh exercise discreion over he accouning reamen of warranies as a means of opporunisic earnings managemen. Under his scenario, managers gain privae benefis from manipulaing he repored accouning numbers. These opporunisic accouning decisions can be achieved hrough changes in he assumpions and esimaes underlying warrany accruals. In paricular, we examine wheher managers use warrany accruals in order o mee shorerm financial reporing objecives. Achieving earnings arges, such as avoiding losses, avoiding earnings decreases and meeing or beaing analyss forecass, has been exensively sudied in he accouning lieraure (e.g., Burgsahler and Dichev, 1997; DeGeorge e al., 1999). In general, he consensus in prior research is ha managers care grealy abou hese benchmarks and are willing o engage in cosly earnings managemen sraegies o achieve hem (e.g., Brown and Caylor, 2005; Graham e al., 2005).

6 4 We find evidence consisen wih managers using warrany accruals o achieve specific financial reporing objecives. We documen ha firms ha achieve earnings arges repor significanly lower warrany expenses han heir counerpars. Our evidence implies ha managers use he flexibiliy in he assumpions underlying he calculaion of warrany expenses and exercise heir discreion o achieve hese financial reporing arges. Our final analysis, which combines he valuaion and earnings managemen aspecs, shows ha, afer conrolling for boh he informaion role of warrany reserves and earnings managemen incenives, he marke views warrany liabiliies similarly o oher liabiliies. Consequenly, each one dollar of warrany liabiliy reduces a firm s marke value by one dollar. We also documen ha firms ha used warrany accruals o achieve earnings arges have a sronger negaive valuaion coefficien on heir warrany liabiliies. This suggess ha invesors recognize ha he warrany liabiliies of hese firms are undersaed. Our sudy is he firs o exploi a unique and comprehensive daabase on warrany disclosures. We conribue o he exising accouning lieraure in several ways. Firs, we exend prior research on he role of accouning informaion by examining how he capial marke evaluaes warrany informaion, and wheher managers use heir discreion over accouning for warranies o incorporae informaion abou fuure firm performance. Second, we documen ha warrany liabiliies play dual roles: as a coningen liabiliy and as a signal of produc qualiy and fuure earnings growh. Third, by focusing on a specific accouning choice, which allows us o increase he power of our analysis, we specifically answer he calls made by accouning researchers (for example, McNichols, 2003) for disaggregaing empirical measures of accouning choices. Fourh, we advance he lieraure on earnings managemen by exploring wheher managers use heir accouning discreion over warrany accruals o aain financial reporing arges. This allows us o shed ligh on specific mehods ha managers use o achieve hese arges. Thus far, he evidence on hese specific mehods has been scarce. Finally, we documen ha he marke seems o ake ino accoun he possibiliy of earnings managemen in evaluaing he firm s liabiliies.

7 5 The paper proceeds as follows. In secion 2 we provide some background on he economic role and accouning reamen of warranies. In secion 3 we develop our hypoheses and in secion 4 we describe our research design. We repor our resuls in secion 5 and we conclude in secion Background 2.1 The Economic Role of Warranies In he U.S., issuing a warrany plan for consumer producs has is roos in he auomobile indusry. Consumer complains abou auomobile qualiy increased in he 1950 s and inensified he pressure on Congress o ac on behalf of consumers. In 1968, a repor issued by he Federal Trade Commission recognized he need o improve he qualiy of auomobiles, bu wen shor of mandaing warrany plans. Slowly, more manufacurers began issuing warranies for consumer producs as a sandard pracice. Ambiguiies in hese conracs, however, presened enforcemen problems and o achieve a uniform sandard in warrany conracs, Congress passed he Magnuson-Moss Ac in Alhough he Ac did no mandae issuing warranies, i required ha a warrany plan explicily describe he scope and duraion of coverage, he means o obain warrany services, and how various sae laws on warranies are affeced. Warranies became an increasingly imporan sraegic mechanism for manufacures/vendors. The economics lieraure argues ha warranies are a means o overcome informaion asymmeries regarding produc qualiy beween an informed manufacurer/vendor and uninformed cusomers. By issuing a warrany plan ha depends on an ex pos verifiable oucome ha is correlaed wih produc qualiy, he manufacurer bonds herself (and he buyer proecs himself) o is produc qualiy (Grossman, 1981). Spence (1977) posis ha manufacurers provide warranies wih beer erms o 4 Consumer producs are governed by he Magnuson-Moss Federal Trade Improvemen and he Uniform Commercial Code, which is sae specific. All commercial goods are under he Uniform Commercial Code.

8 6 signal heir firm ype (higher produc qualiy). 5 Boulding and Kirmani (1993) confirm in an experimen ha consumers learn abou produc qualiy hrough he warranies offered. Of course, warranies are also used as a markeing ool o promoe producs (Menezes and Quelch, 1990). 6 In a simple signaling game (Spence, 1977), if a separaing equilibrium exiss, a posiive relaion prevails beween firm ype and he qualiy of warrany plans. 7 If a pooling equilibrium obains, however, all firms end up wih he same warrany plan. In such a case, firm ype would no be revealed a all. Anoher possible equilibrium is a parially pooling equilibrium, which has clusers of firms wih he idenical signals. Our quesion is: how informaive is accouning informaion on warranies in valuing firms? To answer his quesion, we firs examine he disclosure requiremens on warranies and hen discuss how firm ype and accouning informaion are relaed. 5 Anoher view on warranies in economics is ha hey faciliae risk sharing beween sellers and buyers (Heal, 1977). Sellers and buyers migh be aware of he failure rae (i.e., no informaion asymmery abou produc qualiy), bu i may be impossible o deermine if a specific iem is a lemon. If warranies are provided as insurance, hen differences in warrany plans mainly reflec differen consumers aiude oward risk. In addiion, he erms of warrany plans migh specify he condiions under which he plan is honored, hereby promoing proper use of he produc. Consumers would value producs wih warranies more and would be willing o pay higher prices for hem. Coss of servicing warranies are addiional produc coss, while warrany liabiliies represen coningen liabiliies. 6 Warranies may also reflec a firm s sraegy o improve is repuaion among is cusomers. Ceeris paribus, cusomers migh infer ha a company providing producs wih beer warrany coverage is more reliable han a company providing less warrany coverage (Murhy and Djamaludin, 2002). If so, companies wih beer warrany coverage develop a sronger repuaion among cusomers regarding heir producs. In addiion, firms may use warranies o sraegically promoe fuure sales and growh even hough i is cosly o do so. Firms offer a warrany plan over a longer duraion and/or more comprehensive coverage as an effecive markeing ool (Menezes and Quelch, 1990). 7 Alhough his relaion is inuiively appealing, i is by no means he only heoreical predicion in signaling games. Of course, if a pooling equilibrium prevails, all firms offer idenical warrany plans. However, he relaion beween warrany coverage and firm ype can be negaive in a separaing equilibrium. For example, Luz (1989) derives a separaing equilibrium in which high produc qualiy is signaled wih a low warrany plan and a low produc price when consumers are subjec o moral hazard. Under double moral hazard (boh consumers and producers), he relaion beween warrany policy and firm ype can be eiher posiive or negaive, depending on he parameer values (Cooper and Ross, 1985). Gal-Or (1989) analyzes he role of warrany in an oligopolisic marke and shows ha muliple equilibria can resul; warrany/ype relaion is posiive in one, bu negaive in anoher equilibrium. Thus, in hese equilibria, he informaion conen of a warrany plan regarding firm ype is exremely limied. Given he conradicing predicions proposed by hese models, he relaion beween warrany policies and firm ype in he U.S. produc marke is, o a large exen, an empirical issue.

9 7 2.2 Accouning for Warranies Manufacurers who provide produc warranies are required o record an accrued warrany expense a he ime of sale. 8 Like many oher accruals, hese warrany expenses are esimaed based on company s projecions of fuure claims. Such warrany expenses are an imporan componen of firms selling expenses and can be subsanial in magniude. In our sample, he average warrany expense consiues abou one percen of sales and abou eleven percen of operaing income. The disclosures of warrany expenses and liabiliies were volunary unil he issuance of Financial Inerpreaion No Guaranor s Accouning and Disclosure Requiremen for Guaranees, Including Indirec Guaranees of Indebedness of Ohers (FIN 45) in 2002 (see FASB, 2002). 9 By mandaing disclosures, FIN 45 expands he informaion made available o invesors abou firms warrany accruals, claims, and liabiliies. 10 Beginning in 2003, firms provide: (1) he esimaed poenial amoun of fuure paymens under he warrany plan (warrany reserves or liabiliies), (2) he accouning policy and mehodology used in deermining he liabiliy for produc warranies, and (3) a abular reconciliaion of he changes in he warrany liabiliy for he reporing period. This deailed reconciliaion presens he beginning balance of he aggregae produc warrany liabiliy, he aggregae reducions in ha liabiliy for paymens made under he warrany plan (i.e., claims), he aggregae changes in he liabiliy for accruals (i.e., warrany expenses) relaed o produc warranies issued during he reporing period, he aggregae changes in he liabiliy for accruals relaed o preexising warranies (including adjusmens relaed o changes in esimaes), and he ending balance 8 Under he curren accouning regulaion (Technical Bullein 90-1), revenues from exended warranies are deferred and service coss are expensed as incurred. Thus, accouning informaion on warranies does no include informaion on exended warranies. 9 Prior o FIN 45, he disclosure on warrany obligaions were volunary unless he warrany liabiliies exceed 5% of oal liabiliies. FIN 45 applies o financial repors ending afer December 15, Gu (1998) documens ha prior o FIN 45, firms differ in heir volunary disclosure behavior wih respec o warrany informaion.

10 8 of he aggregae produc warrany liabiliy. Appendix A provides wo examples of warrany disclosures from he financial saemens of Dell and Wesern Digial. 2.3 Inerpreaion of Warrany Daa: A Signaling Perspecive We now discuss how one could inerpre he accouning informaion on warranies (warrany expenses, warrany claims, and warrany liabiliies) from a signaling perspecive in which a firm designs is warrany policy o signal is firm ype (produc qualiy). 11 Alhough he direc signaling mechanism is he warrany policy iself, accouning informaion on he warrany plan could also reflec firm ype. Consider he hree possible equilibria in a signaling game: pooling, fully separaing, and parially pooling equilibria. Assume firs ha warrany policies are observable wihou noise. If a pooling equilibrium prevails, clearly one canno discriminae firm ype by sudying warrany coverage. Bu, accouning informaion on warranies can reveal firm ype; inferior qualiy will resul in higher claims and higher warrany expenses. In his case, qualiy and warrany coss are negaively relaed. Nex, consider a fully separaing equilibrium, in which beer ypes provide beer warrany coverage. 12 In such a scenario, warrany policies signal produc qualiy and fully reveal firm ype. 11 If, insead, warranies are provided for insurance purpose (risk-sharing wihou any informaion asymmery beween he buyers and he sellers), we would inerpre accrued warrany expenses as a cos of providing insurance and warrany liabiliies as coningen liabiliies. The choice of insurance policy would reflec he firm s business sraegy and is buyers risk aversion, bu may be independen of firm ype. 12 For simpliciy, we assume ha warrany coverage can be characerized by is duraion and scope. Even hough scope enails differen feaures (full or limied produc replacemen, pars and labor, money back guaranee, ec.), we assume ha buyers are able o assign a sric preference ordering over (and possibly moneary values o) hese various plan feaures. Therefore, a warrany plan wih a longer warrany period and a more exensive scope of coverage is considered beer han one wih a shorer period and less scope. Since duraion and scope may be regarded as subsiues, we furher assume ha buyers are able o assign values o all possible combinaions.

11 9 Alhough accouning informaion reflecs he cos of providing he signal, i does no provide any incremenal informaion abou firm ype. Finally, in he case of a parial pooling equilibrium, warrany cos informaion is informaive abou firms wihin each pool wih an idenical warrany policy, bu does no provide any incremenal informaion across pools. Again, warrany policies hemselves fully reveal firm ype across policy pools. In sum, if informaion on warrany policies is observed perfecly, accouning informaion provides incremenal informaion abou firm ype if eiher a pooling or parial pooling equilibrium prevails. However, informaion on warrany policies may be imperfec and moreover, empirical measures of warrany plans are likely o be measured wih noise. 13 The necessary informaion o make an accurae assessmen of warrany plans a he firm level for each firm is simply no available. In such a case, even for a separaing equilibrium, accouning informaion is likely o provide incremenal informaion. The quesion is how warrany coss and firm ype are relaed Alhough warrany coss and firm ype are negaively correlaed in a pooling equilibrium, i is quie possible ha hey are posiively correlaed in a separaing equilibrium (or across pools in a parially pooling equilibrium). A separaing equilibrium requires a cos srucure in which he marginal cos of providing beer coverage is lower for firms wih higher qualiy producs han for firms wih lower qualiy producs (referred o as he single crossing propery ). Since buyers are willing o pay more for beer producs, higher-ype sellers will rade-off a higher cos of signal (i.e., beer coverage) agains a higher produc price. Since he coss of beer warrany plans are higher for 13 There are wo sources of noise. Firs, firms ypically provide only coarse informaion on warrany policies such as he range of warrany duraion for heir producs. Even hough we devise a mehod o evaluae various feaures of warrany plans for a specific produc and o assign a score for he warrany plans, mos of hese feaures may no be easily observable. Second, since mos firms sell many producs, o obain a perfec measure for each firm, one needs informaion on he warrany policies and he sales levels of all producs sold by he firm. However, such disaggregaed daa are no available.

12 10 lower-ype firms, i would no be worhwhile for hem o mimic higher ypes. Hence a separaing equilibrium emerges. Of course, he wors firm ype would no offer any warrany plan and repor zero warrany coss. A slighly beer firm would offer a slighly beer warrany plan and incur a sricly posiive warrany cos o separae iself from he wors ype. Thus, here will be a posiive relaion beween firm ype and warrany coss, a leas locally. However, such a monoonic relaionship migh hold globally. Therefore, under a cerain cos srucure and demand for he produc, we expec beer coverage chosen by a higher-ype firms o be more cosly han coverage chosen by lower ypes. Needless o say, beer ypes would incur more warrany coss only if hey generae sufficienly higher revenues o resul ulimaely in higher profis. Thus, warrany expenses and produc qualiy may be posiively relaed in such a separaing equilibrium. The relaion beween warrany liabiliies and produc qualiy is deermined by several variables including warrany expenses, claims, and iming of claims, and he duraion of he warrany coverage. Clearly, ceeris paribus, liabiliies will expire faser, as he duraion becomes shorer. Since warrany expenses and claims are expeced o mach over ime, for firms wih similar duraions and claim paerns, we expec he same relaion beween produc qualiy and warrany expenses o also hold for warrany liabiliies. Tha is, under cerain condiions, warrany liabiliies and produc qualiy are posiively relaed for firms wih he same warrany duraion Warrany liabiliies are deermined by warrany expenses and he claims processed during he accouning period. Consider a separaing equilibrium in which qualiy and warrany coss are posiively relaed. Recall ha plan coverage differs in scope and duraion. Then, ceeris paribus, warrany reserves would be larger for warrany plans wih longer duraion because sales from longer periods are sill under warrany. For simpliciy, furher assume a produc fails (if i fails a all) on he las day of he warrany coverage period and claims are processed nex day; hen all of he warrany expenses would be ousanding as warrany liabiliies a he end of he accouning period. If a warrany duraion is very shor, say a week, hen he maximum warrany liabiliy ha a firm would have is based on he sales during he las week, while if a warrany period is one year, he maximum warrany liabiliy would be based on he sales during he las one-year period. To he exen ha a beer warrany plan offers a longer duraion, firms wih a beer qualiy produc would have larger warrany liabiliies. Similarly, if a firm has a warrany plan wih beer scope of coverage, he warrany cos per uni

13 11 Having esablished possible relaions beween firm ype and warrany coss (boh warrany expenses and warrany liabiliies), we now briefly examine our sample firms o see wha sor of equilibria, if any, various indusries belong o. Perusal of warrany policies of our sample firms shows ha warrany coverage varies wihin an indusry, especially in erms of duraion. We also find ha variaions appear o be small and ha here are clusers of firms wih he same warrany duraion. Tha is, many indusries exhibi parially pooling equilibria. We herefore expec firms wih he same warrany duraion o exhibi a negaive relaion beween produc qualiy and warrany coss, while firms wih differen duraions may possibly exhibi a posiive relaion beween produc qualiy and warrany coss. 15 Alhough we view his posiive relaion as depicing a possible equilibrium in a simple signaling game, i is by no means unique. In addiion, i is possible ha our sample firms are involved in a differen game (see foonoe 6). Ulimaely, how firm ype and warrany coss are relaed is an empirical quesion, which we invesigae in he following secions. 3. Hypohesis Developmen We now develop specific hypoheses for our empirical analysis. The firs se of hypoheses focuses on warrany policies as par of an overall business sraegy (as opposed o accouning choices) and how he capial marke evaluaes accouning informaion on warranies, i.e., he warrany reserve and warrany accruals. The second se of hypoheses relaes o he accouning choices regarding would be higher. Thus, he maximum warrany liabiliies are again higher for beer qualiy firms. Of course, producs would fail hroughou he accouning period, and many claims are processed before he period end. Consider anoher simple scenario: assume ha producs fail coninuously, say uniformly during he warrany period, and claims are submied and processed insananeously. Then he ousanding warrany liabiliies would correspond o one half of he sales made during he warrany period (i.e., one half of one week sales or one half of one-year sales in he example above). Therefore, as before, he relaion beween warrany coverage and warrany liabiliies is posiive as long as all firms have he same failure/claim paern. Hence warrany liabiliies increase wih firm ype. 15 In his sudy, we use he informaion on duraion of warrany coverage as he proxy for he qualiy of warrany plans. Noe, however, his proxy is an imperfec one. Thus, for firms wih he same duraion, we canno conclude ha hey are in he same pool. I is quie possible ha he scope of he coverage differ among hese firms.

14 12 warranies. To he exen ha firms have discreion over warrany accouning, we examine if hey incorporae informaion accuraely in he accouning numbers or alernaively use he discreion o manage earnings o achieve arges. 3.1 Valuaion of he Warrany Liabiliy A produc warrany is an obligaion incurred in connecion wih he sale of goods or services ha may require furher performance by he seller afer he sale has aken place (SFAS No. 5, Accouning for Coningencies). Because of he uncerainy involved wih fuure claims, a produc warrany falls under he definiion of a coningen liabiliy. FASB requires he recogniion and disclosure of a warrany liabiliy when i is probable ha a liabiliy has incurred and he amoun of loss can be reasonably esimaed. If invesors believe ha warrany liabiliies are correcly esimaed, hey would place equal weighs on warrany liabiliies and on oher liabiliies. In his case, he sock marke values warrany liabiliies as reflecing he fuure cash flows o be paid ou. Valuaion of coningen liabiliies is complex and involves assumpions and esimaes ha are unobservable by ousiders. Several sudies invesigaed he valuaion implicaions of coningen liabiliies such as pensions (e.g., Barh, 1991; Espahbodi e al and Landsman, 1986, among ohers), reirees healh benefis (Mielsaed and Warshawsky, 1993), bank loan loss provisions (Peroni 1992; Wahlen 1994; Liu e al. 1997), and environmenal liabiliies (Barh and McNichols, 1994). In general, hey find ha coningen liabiliies are negaively associaed wih share prices. Warrany liabiliies can also capure he warrany policies signal abou produc qualiy. As discussed in secion 2.3, under a reasonable scenario, we expec firms wih beer qualiy producs o incur larger warrany expenses and have larger warrany liabiliies. Firms may ry o mimic each oher by offering idenical warrany plans. However, such a pooling equilibrium may no be susainable.

15 13 Since buyers will be able o infer he qualiy of he producs by examining warrany expenses (i.e., he higher warrany expenses, he lower he produc qualiy), a lower qualiy firm is likely o reduce he level of warrany plans. Thus, in he long run, beer firms are more likely o offer beer plans. 16 Thus, we conjecure ha he sock marke will consider he signaling value of warrany liabiliies and differeniae beween warrany liabiliies and oher liabiliies (e.g., bank loans) by recognizing he dual naure of warrany liabiliies. In paricular, he valuaion coefficien placed on warrany liabiliies is expeced o be less negaive han ha on oher liabiliies. This is because, on he one hand, he sock marke infers ha warrany liabiliies are obligaions o provide services in he fuure, bu, on he oher hand, he sock marke recognizes ha warrany liabiliies conain informaion abou produc qualiy and fuure firm performance. Therefore our firs hypohesis, saed in alernaive form, is as follows: H1: The valuaion coefficien placed on warrany liabiliy is less negaive han he valuaion coefficien placed on oher recognized liabiliies. To invesigae wheher he sock marke correcly values he rue underlying liabiliy role of warrany reserves, we examine he valuaion of warrany liabiliies afer conrolling for heir signaling role. If higher qualiy producs lead o faser fuure earnings growh, we can separae he wo roles by inroducing explicily he earnings growh expecaions of he firm. 17 Under his scenario, warrany liabiliies serving as coningen liabiliies are expeced o be valued similarly o oher liabiliies. 16 However, here are reasons why his scenario does no hold in some markes as discussed in he economics lieraure. 17 The posiive relaion beween produc qualiy and fuure accouning performance is suppored by he posiive relaion beween cusomer saisfacion and fuure performance, since cusomer saisfacion is, a leas in par, due o produc qualiy (Iner and Larcker, 1998). Furher, Nagar and Rajan (2001) provide more direc evidence, documening a negaive relaion beween produc defecs and fuure sales. Also he lieraure on Balanced Scorecard discusses he relaion beween fuure performance and produc qualiy as one form of nonfinancial performance measures (Kaplan and Noron, 1992, 1996).

16 14 Furhermore, we expec ha warrany liabiliies reduce share prices dollar-for-dollar once we conrol for growh expecaions. Thus, our second se of hypoheses, saed in null form, is as follows: H2: Afer conrolling for earnings growh expecaions, he valuaion coefficien placed on warrany liabiliy is equal o he valuaion coefficien placed on oher liabiliies. H2a: Afer conrolling for earnings growh expecaions, he valuaion coefficien placed on warrany liabiliy is equal o negaive one. 3.2 Managerial Discreion over Accouning for Warranies Nex, we examine wheher changes in warrany accouning informaion provides any incremenal signal abou fuure firm performance. From he perspecive of a firm, esimaion of warrany liabiliies require modeling he failure raes and he coss of recificaion acions over he warrany period (Murhy and Djamaludin, 2002). Tha is, accruals relaed o warrany expenses should reflec he esimaes of he inheren qualiy of he producs, given he warrany policy. When he qualiy of a produc improves, a firm is likely o aler he warrany policy o incorporae he change. In such a case, we expec he change in warrany expenses (referred o as abnormal expenses) o reflec he underlying change in warrany policy and serve as a harbinger of good fuure firm performance, assuming a posiive relaion beween qualiy and fuure performance. 18 Thus, abnormal warrany expenses are expeced o be posiively relaed o fuure firm performance, in cases in which produc qualiy and warrany expenses are posiively relaed (see secion 2.3). Incorporaing changes in warrany policies ino warrany expenses is likely o be a resul of alering assumpions abou, for example, expeced fuure failure raes. Thus, his process can be viewed as informaive discreion applied o repored earnings, in ha i improves how curren earnings are relaed o fuure firm performance (e.g., Was and Zimmerman, 1986; Bernard and Skinner, 1996; Subramanyam, 1996). 18 I is also possible ha he change in warrany expenses reflecs change in he sales mix of producs. When producs wih beer qualiy (wih higher margins) and beer warrany policies gain higher weighs in he sales mix, we expec o see a posiive relaion beween he change in warrany expenses and fuure firm performance.

17 15 Managers could also have incenives for ineremporal earnings managemen due o a desire o smooh income over ime. When fuure prospecs are expeced o be poor, managers can overaccrue warrany expenses in he curren period, creaing cookie jar reserves. The reserves are used o offse he fuure poor performance, by shifing income from he presen period o he fuure. If managers expec beer fuure prospecs, hen smoohing calls for under-accruing of warranies in he curren period and shifing income from he fuure o he presen. Thus, he smoohing behavior predics a negaive relaion beween curren abnormal warrany expenses and fuure firm performance, regardless of wheher he expeced fuure performance is good or bad. The associaion beween fuure performance and curren abnormal warrany expenses is expeced o be posiive under he informaional (signaling) hypohesis, while i is expeced o be negaive under he smoohing hypohesis. We use fuure sales growh and fuure reurn on asses raios as fuure firm performance merics. H3a: Fuure sales growh is posiively (negaively) associaed wih abnormal warrany expense. H3b: Fuure profiabiliy is posiively (negaively) associaed wih abnormal warrany expense. To he exen ha he sock marke can observe warrany expenses when financial saemens are disclosed (or infer informaion abou hem hrough oher means of communicaions, such as conference calls) we expec sock prices o reac o unexpeced or abnormal warrany expenses. H3c: The sock marke reacs posiively (negaively) o abnormal warrany expense around quarerly earnings announcemens. 3.3 Benchmark Beaing and Warrany Accruals We now examine he relaion beween accouning choices over warrany accruals and shor-erm managerial incenives o mee or bea earnings benchmarks. The means by which managers achieve earnings arges are numerous, and could be generally classified ino eiher accrual-based sraegies or

18 16 real earnings manipulaions. 19 Despie his broad classificaion, he specific ways in which managers mee earnings arges have been quie elusive o accouning researchers. For example, Burgsahler and Dichev (1997) do no find any srong evidence ha a paricular accouning manipulaion is responsible for benchmark beaing. Dechow e al. (2003) find no evidence ha aggregae discreionary accrual measures are associaed wih benchmark beaing. 20 In conras o he aggregae accrual evidence, several sudies examine specific accrual choices and find some evidence of earnings managemen. By limiing aenion o a specific accouning choice, hese sudies are able o poenially increase he power of he ess. 21 McNichols (2003) emphasizes he imporance of disaggregaing empirical measures of accouning choices o generae a more powerful seing. The warrany conex enables us o overcome some of he difficulies posed by aggregae accrual-based measures and direcly addresses he call for more research on his imporan aribue of he accrual accouning sysem. We hypohesize ha if firms use warrany expenses o achieve financial reporing objecives, here will be an associaion beween abnormal warrany expenses and variables proxying for reporing incenives. We focus on hree exensively-sudied earnings benchmarks: (1) avoiding reporing a loss, (2) avoiding reporing an earnings decrease, and (3) meeing analyss forecass. The evidence in he 19 Anoher way o achieve one of he imporan benchmarks advanced in he lieraure, namely meeing or beaing analyss forecass, is by managing analyss expecaions (Masumoo, 2002). 20 Based on his, hey conclude ha he kink in he repored earnings disribuion is no solely aribued o earnings managemen. They acknowledge ha one shorcoming o finding evidence of earnings managemen is he lack of saisical power in abnormal accrual models o differeniae earnings managemen a a fine level across he wo groups of firms. 21 For example, Beaver, McNichols and Nelson (2003) sudy he loan loss reserves in propery-casualy insurance companies. They find ha reserves are more undersaed in small profi firms han in small loss firms. This evidence is consisen wih firms managing he loan loss reserve o avoid losses. Furher, hey find evidence ha he loss reserve is managed hroughou he earnings disribuion bu is managed mosly by small profi firms (income increasing) and by firms wih he larges profis (income decreasing). Beay e al. (2002) provide evidence ha public banks reduce loan loss reserves o avoid reporing earnings declines. In addiion, hey show ha he higher frequency of earnings increases, relaive o earnings declines, is more prevalen in public banks han in privae banks. They aribue his o he fac ha public banks are more sensiive o beaing earnings benchmarks because heir invesors are more likely o use heurisics in judging banks performance. See also Moehrle (2002) and Dhaliwal, Gleason and Mills (2004).

19 17 lieraure regarding hese benchmarks suggess ha managers view meeing or beaing hem as very imporan. In paricular, based on heir survey, Graham e al., (2005) conclude ha: CFOs believe ha earnings, no cash flows, are he key meric considered by ousiders. The wo mos imporan earnings benchmarks are quarerly earnings for he same quarer las year and he analys consensus esimae. Meeing or exceeding benchmarks is very imporan. (p. 5) They also wrie: Several performance benchmarks have been proposed in he lieraure such as previous years or seasonally lagged quarerly earnings, loss avoidance, or analyss consensus esimaes. The survey evidence indicaes ha all four merics are imporan: (i) same quarer las year (85.1% agree or srongly agree ha his meric is imporan); (ii) analys consensus esimae (73.5%); (iii) reporing a profi (65.2%); and (iv) previous quarer EPS (54.2%). According o Brown and Caylor (2005), analyss forecass have become he mos imporan benchmark o bea since he mid-1990s. This evidence is consisen wih a long lis of archival sudies ha find a endency of firms o repor earnings paerns consisen wih incenives o mee or bea benchmarks. We examine wheher firms appear o have managed warrany accruals o mee he hree alernaive benchmarks. For each of he hree benchmarks, we define suspec firms as hose firms ha are more likely o have used warrany expenses o mee one of he hree benchmarks. Specifically, we idenify firms whose pre-managed earnings numbers fall shor of he arge benchmark, bu whose pos-managed numbers exceed he arges. Abnormal warrany expenses are used o compue pre-managed earnings. Thus, we compare abnormal warrany expenses of hese firms o hose of a se of non-suspec firms. Our hypohesis, in alernaive form, is summarized as follows: H4: Firms ha were jus able o exceed an earnings benchmark will repor lower abnormal warrany expenses for ha quarer compared o oher firms. 3.4 Valuaion of Warrany Liabiliies Combining Growh Expecaions and Earnings Managemen Incenives

20 18 As we noed earlier, he sock marke valuaion of warrany liabiliies may reflec hree aspecs: (i) a coningen liabiliy; (ii) informaion abou he firm s produc qualiy and fuure performance ha is incorporaed in he reserves; and (iii) an earnings managemen componen ha relaes o managers incenives o mee or bea earnings benchmarks. In secion 3.1, we hypohesized (H1) ha he repored warrany liabiliies as a whole, are valued less negaively han oher liabiliies. We hen hypohesized (H2 and H2a) ha afer conrolling for he informaion role of warrany liabiliies, which encapsulaes earnings growh expecaions, hey are valued similarly o oher liabiliies. We now incorporae earnings managemen incenives ino our valuaion framework. Firms wih incenives o mee or bea earnings benchmarks may engage in upward earnings managemen by opporunisically cuing down warrany expenses. This leads o an under-esimaion of warrany liabiliies. If invesors correcly infer ha warrany liabiliies are undersaed by hese firms, hey will adjus he underesimaed warrany liabiliies by placing a larger negaive coefficien on hem. Therefore, we expec a more negaive coefficien on warrany liabiliies for firms wih incenives o mee or bea earnings benchmarks. Our hypohesis, saed in alernaive form, is as follows: H5: For firms ha jus exceeded an earnings benchmark using warrany accruals, he valuaion coefficien placed on he warrany liabiliy is more negaive han he valuaion coefficien placed on oher liabiliies. Finally, we expec ha afer conrolling for earnings managemen incenives and growh expecaions, he marke values warrany liabiliies similarly o oher liabiliies. The valuaion coefficiens on warrany liabiliies and oher liabiliies would be close o negaive one. Thus, we sae our hypoheses in null forms as follows: H6: Afer conrolling for earnings growh expecaions and earnings managemen incenives, he valuaion coefficien placed on warrany liabiliies is equal o he valuaion coefficien placed on oher liabiliies. H6a: Afer conrolling for growh expecaions and earnings managemen incenives, he valuaion coefficien placed on he warrany liabiliy is equal o negaive one.

21 19 4. Research Design: Proxies for abnormal warrany expenses and claims In our analyses we use hree proxies for quarerly abnormal warrany expenses and quarerly abnormal warrany claims. Our firs proxy is based on he seasonal change in warrany expenses or claims, adjused for he seasonal change in sales. In calculaing his proxy we assume ha he level of warrany expenses (or claims) is proporional o sales, i.e., WEXP = SALES where WEXP SALES j, 4 j, 4. Thus, abnormal warrany expenses in our ime-series seasonal model (ABWEXP) are: (Time-series model) ABWEXP _ TIME j, WEXP WEXP j, j, 4 TA j, 4 SALES * SALES j, j, 4 We obain quarerly observaions of each variable () and use as a benchmark he same variables in he same quarer in he previous year (-4). In his model we conrol for growh in a firm s operaions, which is one of he imporan deerminans of warrany accruals. Marquard and Weidman (2004) uilize a similar model in a differen conex. In a similar way, we compue he abnormal (or unexpeced) claims made during a paricular period as: (Time-series model) ABCLAIM _ TIME j, CLAIM CLAIM j, j, 4 TA j, 4 SALES * SALES j, j, 4 This will be a more direc measure of changes in produc qualiy. Our second proxy is an indusry-adjused measure based on membership in a common wodigi SIC code group. For each quarer, we compue he mean level of he raio of expenses (or claims) o sales, excluding he firm for which we calculae he measure. We require a leas en firms in he indusry group. We consider he deviaion from he indusry mean as our proxy for he

22 20 indusry-adjused abnormal warrany expenses (or claims). Thus, abnormal warrany expense in our indusry model is: WEXP (Indusry model) j, WEXP j, ABWEXP _ INDUSTRY j, AVERAGE SALES j, SALES j, Similarly, abnormal claims are defined as: (Indusry model) CLAIM j, CLAIM j, ABCLAIM _ INDUSTRY j, AVERAGE SALES j, SALES j, OTHER _ FIRMS OTHER _ FIRMS Our hird proxy considers he duraion of warranies in calculaing indusry-adjused abnormal warrany expenses (or claims). For each indusry-quarer, we classify observaions ino a low, medium, or high-erm group if he warrany duraion falls below indusry median, equals o he indusry median or exceeds he indusry median, respecively. We hen compue he mean level of he raio of warrany expenses (or claims) o sales for each indusry-quarer-erm group, excluding he firm for which we calculae his measure. Finally, we ake he deviaion from he indusry-quarererm mean as our proxy for abnormal warrany expenses or claims.

23 21 5. Empirical Resuls 5.1 Daa and Sample FIN 45 inroduced new disclosures abou warrany accruals, warrany claims, and liabiliies associaed wih firms warranies. We obain hese daa for he years The sample firms are drawn from he se of manufacuring firms ha are expeced o have significan warrany expenses. We also hand collec informaion abou he duraion of warranies from 10-K s of a subse of our sample firms ha belong o indusries ha have more han en firms in our sample. We describe our sample consrucion in Table 1. The original file conains 14,510 firmquarer observaions covering 889 unique firms. Of hese, we eliminae 516 observaions belonging o 36 firms for which we could no obain valid Compusa idenificaion informaion. We furher delee 4,473 observaions for which warrany expenses and claims are missing. In he analyses ha require informaion abou abnormal warrany expenses, we lose up o 3,278 addiional observaions, depending on wheher we use a ime-series or indusry-based model o compue abnormal warrany expenses. Thus, he number of observaions in our analyses varies beween 9,521 and 4,521, depending on he required variables. We also conduc addiional analyses on a subse of firms for which we obain informaion abou he duraion of warranies. We require ha hese firms belong o indusries wih a leas en firms o ensure ha we obain a reliable benchmark agains which o evaluae each firm s warrany erms. This requiremen, as well as he exisence of informaion abou warrany duraion, reduces he sample in hese analyses o 1,651 observaions spanning 159 firms. The sample firms originae from several indusries, bu as manufacuring firms, hey concenrae in a number of groups. As repored in Table 2, abou 70 percen of firms belong o hree indusry groups: manufacurers of indusrial machinery and equipmen (196 firms, 24.3% of sample 22 We hank Eric Arnum of Warrany Week for his help (

24 22 firms), manufacurers of elecronic and oher elecric equipmen (198 firms, 24.6% of sample firms), and manufacurers of insrumens (165 firms, 20.5% of sample firms). Warrany expenses in hese indusries range beween 1.45% and 1.82% of sales. Since hese indusries consis of a large number of firms, we also collec informaion abou heir duraion, which we repor in he las column of Table 2. In Panel A of Table 3, we provide summary saisics ha describe our sample firms. We measure all variables on a quarerly basis by aking averages from he firs quarer of 2003 o he fourh quarer of For some of he variables, we also provide, for comparison purposes, heir values for firms in he S&P 500 index. Our sample firms are dispersed in size, and he average firm is of medium size. The average (median) marke capializaion of our sample firms is $3.2 billion ($678 million), alhough here is large variaion, wih an iner-quarile range of $208 million in Q1 o $2.2 billion in Q3. The average quarerly sales of firms in our sample is $639 million. The average (median) book-o-marke raio is 0.47 (0.42) compared o 0.42 (0.38) of he S&P 500 firms, indicaing ha our sample firms exhibi similar growh as he index firms. Our sample firms quarerly ROA is, on average, 0.8%. ROA before warrany expense is on average 1.2%. This is comparable o 1.5% ROA for S&P 500 firms. Turning o informaion abou warrany expenses, he average (median) warrany expense is $8.54 ($1.16) million. I comprises abou 1.4% of sales and 1.5% of oal expenses. However, he average (median) raio of warrany expenses o he absolue value of ne income is 54.8% (13.1%), indicaing ha for many of our sample firms, he effec of managing warrany expenses could be economically significan. Finally, we find ha he liabiliy for fuure warrany services comprises, on average, abou 4.1% of sample firms oal liabiliies. Panel A of Table 3 shows ha abnormal warrany expenses comprise abou 0.016% of oal asses (median is 0.005%). The indusry-adjused warrany expense is 0.088% of oal asses (median is 0.394% of oal asses). The average deviaion of warrany expenses from is benchmarks is small, which is no surprising since, absen of produc qualiy changes or addiional facors, warrany

25 23 expenses are expeced o say around he benchmark level. This also suggess ha our benchmark models are reasonable. The average (median) quarerly warrany claims is $7.35 million ($1.15 million). These claims consiue abou 1.3% of curren sales. Similarly, he abnormal claims cener around zero, indicaing ha our benchmarks are reasonable proxies of expeced expenses. In Panel B of Table 3 we repor correlaions of key variables. We focus on he warrany variables. There is a negaive correlaion beween he fracion of warrany liabiliies on firms balance shee and firm size, measured as eiher marke capializaion, sales or oal asses. Furher, warrany liabiliies are posiively correlaed wih analyss forecased growh. Examining he abnormal warrany expenses, we find ha hey are posiively correlaed wih he book-o-marke raio. 5.2 Sock Marke Valuaion of Warrany Liabiliies We firs invesigae wheher and how warrany liabiliies are relaed o firm s equiy marke prices. We esimae several models ha include a firm s marke price as he dependen variable, and various componens of balance shee iems as well as ne income as explanaory variables. We use shares ousanding as he deflaor. Our empirical specificaions are derived from he Ohlson (1995) model. They are consisen wih prior research on valuaion of pension liabiliies (Landsman, 1986; Barh, 1991; Barh e al., 1992), liabiliies on reirees healh benefis (Mielsaed and Warshawsky, 1993), and environmenal liabiliies (Barh and McNichols, 1994). Specifically, we esimae several variaions of he following model for firm i and ime : P ASSET 0 NI 5 1 i. NI 6 WLIAB 2 i. * Q NI 1 7 OTHER _ LIAB 3 * Q 2 NI 8 * Q i. 3 ANALYST _ GR 4 i. (1) where P is sock price, ASSET is oal asses per share, WLIAB is he warrany liabiliy per share, OTHER_LIAB is oal liabiliies excluding he warrany liabiliy per share, ANALYST_GROWTH is analys long-erm earnings growh forecass as repored in IBES, and NI is earnings before exraordinary iems per share. To conrol for earnings seasonaliy, we include Q 1, Q 2 and Q 3 as indicaors for he firs hree fiscal quarers.