Personal optioner företags skatt och skuld politiska

Personaloptioner, Bolagsskatt och Skuldpolicy John R. Graham. Mark H. Lang, Douglas A. Shackelford NBER Arbetspapper nr. 9289 Utfärdad i oktober 2002 NBER Program: PE Vi finner att personaloptionsavdrag leder till stora sammanlagda skattebesparingar för Nasdaq 100 och SP 100 företag och påverkar även företagens marginalskattesatser. För Nasdaq-företag är median marginalskattesatsen 31 procent när optionsavdrag ignoreras men faller till 5 procent när man redovisar avdrag. För SP-företag påverkar dock inte avdragsavdrag i hög grad marginella skattesatser. I enlighet med DeAngelo och Masulis (1980), är avdrag för avdrag viktiga skatteskulder som kan påverka företagets policy. Vi finner bevis som överensstämmer med optionsavdrag som ersätter ränteavdrag i beslut om bolagets kapitalstruktur. Denna bevisning förklarar delvis varför vissa företag verkar vara underlevererade. Valutakursoptioner, bolagsskatter och skuldpolicy Skriven av: John R. Graham, Mark H. Lang och Douglas A. Shackelford Vi finner att personaloptionsavdrag leder till stora aggregat skattebesparingar för Nasdaq 100 och SampP 100 företag och påverkar även företagets marginalskattesatser. För Nasdaq-företag, inklusive effekten av optioner, minskas den beräknade median marginala skattesatsen från 31 procent till 5 procent. För SampP-företag påverkar däremot inte avdragsavdrag i hög grad marginella skattesatser. Våra bevis tyder på att optionsavdrag är viktiga skattemässiga skattesköldar och att optionsavdrag ersätter ränteavdrag i beslut om bolagsstrukturstruktur, delvis förklarar varför vissa företag använder så liten skuld. I det här dokumentet undersöks bolagsskatteeffekterna av kompensera anställda med icke-kvalificerade personaloptioner. Företagen dras av skillnaden mellan nuvarande marknads - och strejkpriser när en anställd utövar ett icke-kvalificerat aktieoption. För optionsintensiva företag med stigande börskurser kan detta avdrag vara mycket stort. Vi fokuserar på effekterna av optioner för år 2000 marginalskattesatser för Nasdaq 100 och SampP 100 företag och konsekvenserna för skuldpolitiken.1 Förståelse av skatteeffekterna av optioner är allt viktigare eftersom andelen ersättning som betalats i optionsoptioner har ökade de senaste åren. Ett perspektiv på omfattningen av alternativkompensation och dess ökning över tid kan hämtas från papper som Desai (2002) som rapporterar att de fem största officerarna i de största 150 amerikanska företagen fick alternativ med bidragsvärden överstigande 16 miljarder år 2000, vilket han uppskattar är en tiofaldig ökning under årtiondet. Han uppskattar att intäkter från optionsoptioner ökar i genomsnitt 29 procent av de operativa kassaflödena 2000, en ökning från 10 procent år 1996. Dessutom har utnyttjandet av dessa optionsrätter skapat stora företagsinkomstavdrag. Sullivan (2002) uppskattar att den totala bolagsskatten sparande från avdrag för optionsoptioner hoppade från 12 miljarder 1997 till 56 miljarder år 20002 Cipriano, Collins och Hribar (2001) rapporterar att skatteavgiften är från empl oyee optionsoptioner för SampP 100 och Nasdaq 100 uppgick i genomsnitt till 32 procent av de operativa kassaflödena 2000, en ökning med 8 procent 1997. Sullivan (2002) tillägger att skatteavdrag i 2000 överskred nettoresultatet för åtta av de 40 största amerikanska företagen (bestämt av marknadsvärde): Microsoft, AOL, Cisco Systems, Amgen, Dell Computer, Sun Microsystems, Qualcomm och Lucent. Optionskompensation har dessutom spridit sig bortom tekniklagret. Företag som är så olika som General Electric, Pfizer, Citigroup och IBM har dragit in mer än 1 miljard i optionsoptionsersättning år 2000. Vår analys bekräftar att avdrag för personaloptioner minskar väsentligt företagsskattbetalningarna. Vi uppskattar att aktieoptioner under 2000 minskar företagets skattepliktiga intäkter med cirka 100 miljarder för vårt urval av SampP 100 och Nasdaq 100 företag. För SampP 100-företagen motsvarar sammanlagda aktieoptionsavdrag cirka 10 procent av aggregerad förskottsinkomst. För de Nasdaq 100-företagen (som är mer optionskrävande) överstiger de samlade avdragna aggregerad förskott. Denna studie fokuserar emellertid främst på effekten av personaloptioner på MTR och den därpå följande effekten på kapitalstrukturen. MTR är en viktig faktor för många ekonomiska beslut. I synnerhet om personaloptioner är tillräckligt stora för att påverka MTR, kan de minska värdet av ränteavdrag och ändra incitamenten att utfärda skulder. Vi finner att aktieoptionsavdrag väsentligt minskar MTR. För Nasdaq-företag utgör avdrag så stor andel av förhandsoptionsinkomsten att median MTR tumlar från 31 procent när vi ignorerar optionsavdrag till 5 procent när optionsavdrag ingår i skattesatsberäkningen. För SampP-företagen är median-MTR liten påverkad av optionsavdrag. Såsom beskrivs närmare i avsnitt I isolerar vi effekten av tre klasser av optioner på MTR: de som redan utövas, de som beviljats ​​men ännu inte utnyttjats, och de som ännu inte beviljats. Varje klass av alternativ bidrar till den totala minskningen av MTR. Vi testar sedan huruvida effekten av personaloptioner på MTR påverkar skuldpolitiken. DeAngelo och Masulis (1980) hävdar att företagen ersätter skuldsättnings - och skattemässiga avdragsskulder (till exempel optionsavdrag) när de bestämmer sin optimala kapitalstruktur. Tidigare undersökningar av denna substitutionseffekt är obetydliga (se Graham (2003) för en översyn). Några papper drar slutsatsen att högmälda företag verkar ha otillräcklig skuld i sina kapitalstrukturer. Hanlon och Shevlin (2002) påpekar emellertid att dessa tidigare studier kanske inte kan upptäcka den förväntade skuldrelationen för MTR eftersom de ignorerar skatteavdrag från aktieoptionsutövning. I vårt urval finner vi att skuldkvoter och MTR inte är signifikant parvis korrelerade när vi ignorerar optionsavdrag vid uppbyggnad av MTR. Däremot är förhållandet mellan skulder och skatter positivt och betydande efter justering för förväntade optionsavdrag. Resultatet indikerar att redovisning av de skattemässiga avdrag som är kopplade till aktieoptioner ger betydande inkrementskraft för att förklara skuldpolitiken, vilket är förenligt med chefer factoring i skatteeffekterna av optioner när de väljer kapitalstruktur. Vidare, när vi identifierar företag som verkar vara underlevererade när optionsavdrag ignoreras, finner vi att dessa företag är de som använder de flesta alternativen. Sammantaget överensstämmer vår analys med företag som handlar utan skuldskuld och skattemässiga skattelättnader vid beslut om kapitalstruktur, på det sätt som DeAngelo och Masulis (1980) föreslagit. Våra resultat kan också ge ett partiellt svar på pusslet om varför vissa företag idag använder så liten skuld (Graham (2000)) när alternativavdrag beaktas, återspeglar MTR för dessa företag ett litet skatteincitament att använda skuld, så att deras låga skuld förhållanden kan vara lämpliga. Vårt papper är relaterat till flera grenar av akademisk forskning. Den andra halvan av vårt papper är mest liknar Kahle och Shastri (2002), som undersöker om företag med stora optionsavdrag använder mindre skuld. Kahle och Shastri anser dock inte flera problem som vi tar upp. För det första beräknar de inte MTR, eller effekten av alternativ på MTR. Dessa underlåtenhet är en brist, eftersom optionsavdrag endast bör påverka kapitalstrukturbesluten i den utsträckning de påverkar MTR. För det andra, som diskuteras mer detaljerat senare, mäter de optionsavdrag med skatteförmånerna i årsredovisningen, snarare än att använda mer exakta uppgifter i aktieoptionsfotnoten (Hanlon och Shevlin (2002)). För det tredje redovisar Kahle och Shastri inte effekterna av optioner som redan beviljats ​​men ännu inte utnyttjats, ej heller alternativ som ännu inte beviljats. Slutligen behandlar Kahle och Shastri inte osäkerheten om optionsövningstidning, och mer generellt hur alternativavdrag interagerar med de dynamiska aspekterna av den federala inkomstskattkoden. Vi presenterar detaljer i avsnitt I och II som beskriver hur vi tar hänsyn till dessa ibland subtila influenser. Förutom effektiv skattesats och kapitalstrukturforskning är detta dokument relaterat till två andra forskningsgrenar. För det första undersöker en serie av papper om skatteincitament spelar en roll i form av ersättning som ett företag väljer att använda. Den tidiga forskningen på detta område var ofullständig (till exempel Hall och Liebman (2000)), men nyligen gjordes av Core and Guay (2001) att höga skattesatsföretag utfärdar färre stock options till icke-verkställande anställda, förmodligen för att företagen hellre skulle Använd traditionella ersättningsformer som leder till omedelbar avdrag. Vårt papper undersöker inte huruvida skatter påverkar valet mellan olika former av ersättning, men föreslår att företagen beaktar skatteeffekterna av ersättning vid beslut om bolagets kapitalstruktur. För det andra är vårt papper relaterat till litteraturen som undersöker hur skattecheferna optimerar företagets skattepolitik (t. ex. Scholes et al. (2002)). Vi bidrar till den här litteraturlitteraturen genom att tillhandahålla bevis som överensstämmer med skattecheferna med tanke på samspelet mellan olika företagspolicyer vid val av skattepositioner.3 I det följande avsnittet diskuteras stora konceptuella frågor som uppstår vid bedömningen av aktieoptions effekt på MTR och vår tillvägagångssätt för att ta itu med dem. Avsnitt II diskuterar vårt empiriska tillvägagångssätt i detalj och beskriver data. Avsnitt III analyserar effekten av optionsavdrag på företagens MTR. Avsnitt IV behandlar samspelet mellan optionsavdrag och företagsskuldpolitik. Avsnitt V presenterar avslutande kommentarer. I. Skattefrågor relaterade till företagsavdrag från personaloptionsoptioner Den simuleringsprocedur som vi använder för att uppskatta MTR-värden för år 2000 innehåller dynamiska funktioner i skattekoden, inklusive skattemässiga förlustavdrag och vidarebefordran (Shevlin (1990) och Graham (1996)). Förfarandet bestämmer MTR baserat på de inkrementella skatteeffekterna som är förknippade med en extra dollar av intäkterna som förvärvats år 2000. Den ökade effekten av en extra dollar av intäkter 2000 kan realiseras var som helst mellan 1998 (på grund av den tvååriga skatteförluståterbäringsperioden ) och 2020 (på grund av den 20-åriga förlustavdragsperioden), eller inte alls (om förluster är tillräckliga för att kompensera all nuvarande och framtida vinst). För att modellera framflyttningseffekten producerar vi först en framtidsbaserad prognos för framtiden genom att prognosera framtida skattepliktig inkomst (diskuterad i avsnitt II. B), framtida bidrag och övningsbeteende (avsnitt II. C) samt framtida aktiekurser (avsnitt II. D) ). Med utgångspunkt från baslinjeprognosen uppskattar vi nuvarande skattekonsekvenser förknippade med ytterligare en dollar av intäkter som förvärvats år 2000. Om skattemässiga konsekvenser uppstår 2001 eller senare, beror vi på inkrementseffekten tillbaka till årsbasis 2000 dollar. I avsnitt II. E utforskar vi frågor som rör diskontering av skatteskulder när ett företag har avdrag för aktieoptioner. För att ta fram osäkerhet om framtiden producerar vi 50 slumpmässiga basprognoser för framtiden, som var och en ger en uppskattning av MTR. Den förväntade MTR är den genomsnittliga skattesatsen bland dessa 50 uppskattningar. I avsaknad av aktieoptioner är uppskattning av MTRs relativt enkelt. Man kan använda den genomsnittliga underförstådda tillväxttakten och avvikelsen från den historiska tidsserien av skattepliktig inkomst (beräknad från förskottets inkomstjusterad skatteskuld som beskrivs närmare i nästa avsnitt) som fröparametrarna för att producera de 50 slumpmässiga utgångsprognoserna för Framtiden fram till 2020. Men förekomsten av alternativ introducerar flera viktiga frågor i standard simuleringsproceduren. Vi diskuterar dessa problem under resten av det här avsnittet. För det första kan man inte längre justera förskottsinkomst för uppskjuten skatt för att uppskatta skattepliktig inkomster, till skillnad från andra former av ersättning, återges inte teckningsoptioner normalt i förskott eller i uppskjuten skatt. När det gäller premieinkomst anses optioner i allmänhet inte som resultaträkningen och företag som väljer att inte kosta aktieoptioner minskar också skattekostnaden i resultaträkningen för att återspegla effekten av optionsavdrag.4 Vidare, till skillnad från många bokföringsskillnader , effekten av optioner tas inte upp i uppskjuten skatt eftersom skillnaden mellan skatte - och bokföringsinkomster aldrig återfaller. Som ett resultat kan ett företag konsekvent redovisa hög skattekostnad (på finansiella rapporter) och betalar aldrig några skatter (på avkastning). Tidigare forskning har vanligtvis använt resultaträkningsdata för att utgå skattepliktig inkomst och sålunda ignoreras optionsavdragsavdrag för majoriteten av företagen (eftersom de flesta företag inte bekostar optioner). Ett undantag är Kahle och Shastri (2002), som gör en justering för aktieoptioner med hjälp av redovisade skatteförmåner från aktieoptionsnummer för att justera premieinkomsten.5Hanlon och Shevlin (2002) betonar att det är problematiskt med flera anledningar att använda denna metod. För det första redovisar många företag inte separat skatteförmånen från aktieoptioner i sina finansiella rapporter. Även för Nasdaq 100, för vilka teckningsoptioner är sannolikt stora, noterar Hanlon och Shevlin att endast 63 företag redovisar skatteförmånerna från optioner i sina 1999 års bokslut. Vidare är justering av förskottsinkomst för optionsskattförmåner relativt enkel om skattepliktig inkomst är positiv, fall med skattemässiga förluster är mer komplexa på grund av effekterna av nettoförlusterna och skattevärderingsbidrag. Vi undviker dessa frågor genom att följa Hanlon and Shevlins råd och samla våra optionsavdragsdata från den detaljerade informationen om bidrag och övningar som finns i de finansiella fotnoterna. Denna information rapporteras konsekvent över företag oberoende av skattestatus. En andra unik fråga med optionsrätter är att MTR-värden i nuvarande period kan påverkas av flera klasser av optionsavdrag: De som härrör från redan utnyttjade optioner (eftersom de påverkar nuvarande nivå av beskattningsbar inkomst och eventuellt förlustavdrag), liksom de som hänför sig till överhänget av redan beviljade men inte ännu utnyttjade optioner och ej ännu beviljade optioner (eftersom dessa alternativkategorier kan skapa förluster i framtiden som påverkar aktuella MTR-värden via skattemässiga fram - och återbäringsfunktioner av skatten koda). Alla studier som vi är medvetna om överväger bara en av dessa typer av alternativ: redan utövade alternativ. Denna begränsning är acceptabel för undersökning av effektiva skattebördor som Desai (2002), Hanlon och Shevlin (2002) och Sullivan (2002). Det är emellertid viktigt att överväga alla tre alternativen när man studerar ekonomiska beslut som grundar sig på marginella skatteincitament. Alternativ som är utestående men som ännu inte utnyttjats, till exempel, skapar avdragsöverhängning i den meningen att företag kan hitta sig i positioner där det finns många utestående köpoptioner som sannolikt kommer att utnyttjas i framtiden, minskar skattepliktig inkomst och (via återbetalningsfunktionen i skattekoden) nuvarande MTR. Som ett resultat kan två företag som för närvarande ger liknande ersättningsbelopp i alternativen befinner sig i mycket olika MTR-positioner beroende på det tidigare aktiekursbeteendet och antalet alternativ som fortfarande inte utnyttjas. Vi använder fotnotinformation om utestående optioner och tidigare optionsräntebeteende för att förutse de sannolika effekterna av utestående optioner och framtida optionsrätter på nuvarande MTR. Ett tredje begreppsmässigt problem som är unikt för optionsoptionsforskningen är osäkerheten om och när inte ännu utnyttjades alternativ kommer att leda till företagsavdrag. Eftersom aktiekurserna är volatila och alternativen har långa livslängder (oftast tio år) kan nuvarande utestående optioner och framtida optionsrätter generera stora avdrag i framtiden eller inga avdrag alls beroende på aktiekursbanan. Den stokastiska karaktären av aktieoptionsavdrag kan väsentligt komplicera beräkningar av beräknade MTR och följaktligen eventuella företagsbeslut där skatter är relevanta. Aktiekursbanan och besluten om anställningsutövning är svåra att förutsäga, och för effektiv skatte - och finansplanering skulle en chef behöva faktor i sannolikheten och mängden framtida optionsavdrag. Vi genomför uttryckligen en simuleringsmetod för att överväga optionsavdrag med information om aktieoptioner, volatilitet på aktieavkastning, utdelning och förväntad avkastning för att modifiera Grahams (1996) simuleringsteknik. Vi kombinerar förväntade avdrag med simulerad framtida beskattningsbar inkomst för att komma fram till sannolikhetsvägda uppskattningar av MTR. Analysen är väldigt lik den inställning vi ser en företagschef skulle åta sig för att fatta beslut utifrån förväntade MTR. Enligt vår kunskap är vår första studie att ta det externa perspektivet för att uttryckligen införliva information om före-träningsoptioner i MTR-beräkningar. II. Empirisk tillvägagångssätt Vi studerar de företag som fanns i Standard och Poors 100 och Nasdaq 100 den 17 juli 2001 (dagen då vi började datainsamling). De utgör en betydande del av ekonomin och betalar betydande skatter.6Analys av SampP 100 företag ger insikt om traditionella, stabila industriföretag. Nasdaq 100-företagen är det mest lönsamma och stabila bland optionsintensiva högteknologiska företag. Sju företag finns i både Nasdaq och SampP, så det första provet omfattar 193 företag. Under vår MTR-analys inkluderar vi dessa sju företag i SampP-subsemplet men utesluter dem från Nasdaq-undersökningen för att undvika dubbelräkning. Vi kan inte hitta data för tre företag, vilket minskar provet till 190 företag.7 Vi begränsar provet till dessa 190 företag eftersom (i) handsamling av optionsoptionsuppgifter i finansieringsöversikten fotnoter är kostsam och (ii) vår simulering Metoden är mindre sannolikt att ge tillförlitliga resultat för små, instabila företag. Vi ser ett scenario där en chef bedömer sina företag MTR i slutet av räkenskapsåret. Vår referenspunkt är det senaste året för vilken data var tillgängliga vid starten av projektet, vilket är årsskiftet 2000 enligt definitionen av Compustat (årsskiftet från juni 2000 till maj 2001) för den stora majoriteten av provföretag. 8 Aktiekurserna vid utgången av 2000 var väsentligt under marknadshöjningar, men fortfarande över de senaste marknadsnivåerna, vilket ställer frågan om huruvida resultaten i denna studie är periodspecifika. Eftersom undersökningsperioden följer en utvidgad tjurmarknad, kan chefer inte ha tänkt på storleken på de eventuella optionsrätterna när de beviljades alternativen tidigare år. Ändå är vår karaktärisering representativ för situationen företag som befann sig i slutet av 2000, med chefer som möter MTR som liknar dem som uppskattas i denna studie.9 Mer allmänt är det tillvägagångssätt som vi utvecklar i denna studie att vara användbart under ett år för att införliva aktieoptionsavdrag i MTR-beräkningar, huruvida optionsavdrag är stora eller små under ett givet år. B. Beräkna historiska och framtida intäkter (ignorera alternativavdrag) Vi genomför en variant av simuleringsalgoritmen som används i Shevlin (1990) och Graham (1996), vilket kräver en prognos för framtida inkomster för att beräkna nuvarande MTR. Vårt förfarande förutsätter att intäkterna nästa år motsvarar inkomster i år plus en innovation. Innovationen dras från en normal fördelning med tillväxt och volatilitet beräknad från företagsspecifika historiska data. Eftersom optioner inte ger upphov till räkenskapsinkomster, innehåller vår Compustat-baserade åtgärd av historiskt resultat före skatt, justerat för uppskjuten skatt, inte effekten av aktieoptionsavdrag.10 Med förlängning innehåller inte heller våra basprognoser för framtida inkomster effekterna av optionsavdrag. Eftersom våra uppgifter härrör från bokslutet, möter vår åtgärd av skattepliktig inkomst de vanliga begränsningarna när boknummer används för att approximera skattebetalningar, inklusive bokföringsmässiga skillnader i konsolidering och erkännande av utländska vinster. 11 Vi använder Compustat data från den sista 20 år för att beräkna firmespecifik tillväxt och volatilitet. Vissa företag har extrema historiska resultatinformation som verkar otänkbara framöver. Därför bundna vi varje företag vinsttillväxt och volatilitet för att falla inom deras respektive 25: e och 75: e procentenheter bland alla företag i samma 2-siffriga SIC-kod. 12 Med hjälp av dessa tillväxt - och volatilitetsberäkningar förutspår vi förutbestämd skattepliktig inkomst för de närmaste 20 åren. C. Inklusive historiska och framtida optionsövningar Sedan 1996 har SFAS 123 krävt att företagen i bland annat ska inkludera finansiella fotnoter, a) en beskrivning av optionsvillkor, b) antal alternativ, vägt genomsnittligt aktiekurs och återstående kontraktsliv för utestående optioner vid periodens slut, c) tre års utövande, bidrag och avbokningshistorik (antal aktier och vägat genomsnittspris), och (d) Black-Scholes-värdet av optioner som beviljats ​​under perioden , inklusive de underliggande antagandena för utdelningsavkastning, riskfri ränta, årlig avkastningsvolatilitet och förväntad löptid före träning. 13 Firm har relativt litet utrymme för skönsmässig bedömning i Black-Scholes bedömningar, och fotnotformatet är i allmänhet konsekvent över företagen. För de företag med ovanliga upplysningar är våra resultat robusta mot deras uteslutning.14För illustrativa ändamål innehåller bilagan Microsofts aktieoptionsfotnot för året som slutade 30 juni 2000. Hall och Leibman (2000) konstaterar att 95 procent av samtliga optioner är nonqualified, så vi gör det förenklade antagandet om att alla alternativ som rapporteras i fotnoten är icke-kvalificerade. Fotnoten innehåller historisk övningsinformation för de föregående två och nuvarande räkenskapsåren (1998, 1999 och 2000 för de flesta av våra företag). För varje företag beräknar vi optionsavdrag eftersom antalet optioner som utövas under ett visst år gånger skillnaden mellan genomsnittlig aktiekurs för dessa optioner och aktiekursen vid övning. Vi mäter aktiekursen vid övning för ett visst år med hjälp av det genomsnittliga aktiekursen för optioner som beviljats ​​samma år.15 Att införliva historiska optionsavdrag i vår analys är enkel: vi subtraherar de historiska personaloptionsavdragen från de historiska inkomsttal som härrör från föregående avsnitt. Observera att historiska optionsavdrag kan påverka MTR 2000 genom att minska skattepliktig inkomst år 2000 och även genom att skapa en skatteminskning 1998 eller 1999 som vidarebefordras till 2000. Vi experimenterar med att även samla historiska optionsdata för 1995, 1996 och 1997 för ett slumpmässigt urval av åtta företag för att undersöka huruvida förlusterna i dessa år fortsatte framåt i 2000 tillräckligt för att påverka MTR 2000. Men kostnaden för att samla in data är stor och fördelen liten (dessa extra data påverkar knappast våra resultat) , så vi strävar inte efter att samla in alternativdata för 1998 för andra företag. Fotnoten innehåller även information om alternativ som redan beviljats ​​men ännu inte utnyttjats. För att införliva dessa framtida avdrag i vår analys gör vi antaganden om alternativt övningsbeteende. Huddart och Lang (1996) och Core and Guay (2001) rapporterar att tidigt utnyttjande av personaloptioner är vanligt, med mycket av övningen som uppträder ungefär halvvägs genom optionslivet, och att övningen tenderar att spridas smidigt över tid. Således använder vi det angivna förväntade optionslivet som vår uppskattning av när genomsnittlig övning kommer att inträffa och antar att övningen sprids smidigt under en period som börjar två år före det året och slutar två år efter det året.16 Vissa aktiekursbanor innebär att optionsoptionen är inte optimal eftersom marknadspriset ligger nära eller under aktiekursen (vår avledning av framtida aktiekursbanor beskrivs i nästa avsnitt). Därför följer vi konventet i Huddart och Lang (1996) och antar ingen övning i år om vilka alternativ som finns i pengarna med 15 procent eller mindre (om inte alternativet är vid utgången, i vilket fall vi antar allt för pengarna optioner utövas). I de fall där alternativen är out-of-the-money eller knappt i pengarna, skjuter vi upp till det första året där de är i m o ney med minst 15 procent (eller till utgången). 17 Framtida optionsavdrag kan påverka MTR på nuvarande sätt på två sätt. För det första om de utövas under de närmaste två åren och är tillräckligt stora för att generera en skatteminskning, kan skatteförlusten återföras till att kompensera de skatter som betalats år 2000. Denna återbetalningsbehandling kan leda till återbetalning under 2001 eller 2002 för betalade skatter år 2000, vilket minskade 2000 MTR. För det andra, för företag som inte betalar skatter år 2000 utan istället förlorar förluster, kommer framtida optionsavdrag potentiellt att öka till det belopp som överförs. Denna vidarebefordran kan försena det datum då skatter slutligen betalas och därigenom reducera (nuvärdet av) nuvarande periodens MTR. Den sista gruppen av alternativ vi anser är de som ännu inte beviljats. Som just beskrivet påverkar dessa alternativ potentiellt 2000 MTR via återbetalningar om de leder till avdrag 2001 eller 2002 (som, med tanke på våra antaganden om övningsbeteende, endast uppstår för företag med ett genomsnittligt livslängd på fyra år eller mindre) eller för närvarande icke-skattade företag, genom att skapa stora skatteförluster som kommer att överföras. Vi antar att företagen beviljar framtida optioner i ett belopp som motsvarar det genomsnittliga antalet beviljade (efter avbrott) under de senaste tre åren, vilket är en tillväxtfaktor. 18 Tillväxtfaktorn är baserad på en viss premieinkomstmängdstillväxt (begränsad mellan 25: e och 75: e procenten för inkomsttillväxten för andra företag i samma 2-siffriga SIC-kod) .19 Starkpriset för ett visst företag År nyutlånade optioner antas vara det prognostiserade aktiekursen för det aktuella året. I nästa avsnitt beskriver vi hur aktiekursen bestäms. För att införliva framtida avdrag för avdrag i vår analys subtraherar vi framtida avdragsavdrag längs en given simuleringsväg från förvalstinkomst (enligt prognos i avsnitt B). Detta ger en prognos över beskattningsbar inkomst efter redovisning av optioner. Ett alternativt tillvägagångssätt skulle vara att subtrahera effekten av optioner från all historisk data (upp till 20 års data) och sedan direkt prognostisera efteroptionsinkomst i framtiden. Tyvärr, eftersom informationen om aktieoptioner endast har krävts sedan 1996, kan vi inte justera uppskattningarna av beskattningsbar inkomst under alla tidigare år, så det här alternativet är inte möjligt. Slutligen ignorerar vi genom hela studien återprissättning, det vill säga att reducera strejkpriset för redan beviljade alternativ. I den utsträckning att företag har åtagit sig att prissätta under nedåtgående prisrörelser, skulle vår strategi leda oss att underskatta framtida avdrag för avdrag. D. Beräkning av framtida aktiekurser Vi förutser framtida aktiekurser så att vi kan projicera omfattningen av framtida optionsrätter. Vi projicerar en separat framtida aktiekursväg i samband med var och en av de 50 simuleringarna av framtida intäkter som beskrivs i avsnitten I och II. B. Genom detta förfarande kan värdet på aktieoptionerna variera med aktiekurserna (och eftersom vi länkar aktiepapper till resultat, varierar med olika inkomstsimuleringar). För att projicera framtida aktiekurser beräknar vi en förväntad avkastning för varje företag baserat på CAPMs marknadsmodell. Denna totalavkastningsberäkning kräver en företagsspecifik beta (taget från CRSP), den riskfria räntan (från aktieoptionsfotnot) och ett riskkapital på 3,0 procent (vilket överensstämmer med de senaste beräkningarna av riskpremien i Fama och French (2002) och Graham and Harvey (2002).) 20 Vi är intresserade av kapital appreciering i aktiekurs, så vi subtraherar det företagsspecifika utdelningsavkastningen från varje företags totalavkastning. Aktiepriserna tenderar att variera med resultat. Easton och Harris (1991) visar att förändringar i årslön och årlig avkastning är positivt relaterade (Pearson-korrelation på cirka 20 procent). För att införliva denna positiva empiriska koppling mellan aktieavkastning och resultat ändrar vi förväntade avkastningar för att koppla dem till de resultatprognoser som härrör från avsnitt B. Vi antar att oväntat högt intjäning åtföljs av proportionellt positiv förväntad avkastning på aktier. Tänk exempelvis på ett fall där vinsten förväntades växa till 10 procent och aktiekursen väntades öka med 12 procent. Antag att vi i en viss simulering hamnar på en väg med en ökning med 15 procent under det första året (50 procent högre tillväxt än väntat). För att länka de två serierna tilldelar vi en förväntad aktieavkastning på 18 procent på den banan för det året (50 procent högre än förväntat). Denna anpassning ändrar förväntad avkastning på ett sätt som kopplar resultat och avkastning.21 Robusthetskontroller indikerar emellertid att graden av antagen korrelation inte är särskilt viktig. När vi replikerar studien som tar hänsyn till oberoende mellan årsinkomst och årlig avkastning, är inferenser kvalitativt oförändrade (genomsnittliga skattesatser är 50 punkter högre än de som redovisas i basfallet nedan). Dessutom förändras våra kvalitativa resultat inte om vi antar en förväntad aktiekursökning om 12 procent årligen för alla företag.22 Med tanke på en förväntad avkastning av aktier, projicerar vi framtida aktiekurser genom att dra tillbaka avkastningen från en lognormal distribution. För varje år motsvarar medelvärdet av denna fördelning den förväntade avkastningen, beräknad som just beskrivet, och variansen är den som redovisas i aktieoptionerna fotnoter.23 I vår metod använder vi historiska data för att uppskatta inkomsttillväxten (som beskrivs i avsnitt B ) och en modifierad CAPM förväntad avkastning (som just beskrivet). I en robusthetskontroll använder vi Value Line-prognoser för de 131 företagen i vårt prov för vilken värdelinje tillhandahåller uppskattningar. För inkomsttillväxt årliggör vi värderingslinjens fyraåriga tillväxtberäkning av försäljningsökningen när den är tillgänglig, eller använd värdet för vinstledningsvinst när försäljningsutvecklingen inte är tillgänglig. För aktieavkastning årliggör vi avkastningen implicit i genomsnittet av de höga och låga fyra åriga målpriserna. Genom att använda dessa alternativa vinst - och aktieutvecklingsräntor ger de genomsnittliga MTR-värdena som är bara 12 punkter högre än de som vi rapporterar nedan och ingen skillnad i de övergripande kvalitativa resultaten. E. Rabattering Framtida aktieoptionsavdrag I det här avsnittet diskuteras diskonteringsräntan som vi använder för att fastställa nuvärdesskatteeffekter av optionsoptioner för MTR. Påminn om att på grund av taxkodens återbäring och framåtriktade egenskaper kan effekterna av dagens avdrag potentiellt kännas långt in i framtiden. The issue is determining what rate should be used to discount these future tax consequences. Some previous research (e. g. Graham (1996, 2000)) uses the corporate bond yield as the discount rate to determine the present value of the tax effect of various deductions (e. g. debt interest) on MTRs and firm value. This approach implicitly assumes that the tax effects of these deductions have the same risk as debt, as assumed by Modigliani and Miller (1958) for interest deductions. It seems less reasonable to discount the effects of future option deductions using the debt rate. Options generate deductions on exercise, and option exercise is correlated with stock returns therefore, options lead to higher compensation costs, as well as tax benefits, when share prices are high.24In the remainder of this section we discuss conceptually how we think that tax liabilities in a stock option world should be discounted, and we link this conceptual framework with our empirical implementation of discounting tax liabilities within the simulation procedure. To keep the discussion focused on the discount rate, we start by making several simplifying assumptions. We assume that options are cash settled, or equivalently, that firms purchase shares on the open market to deliver to employees when they exercise their options. Shares are repurchased at a fair and efficient market price, using funds that would have otherwise been invested in zero-NPV projects, so there are no dilution concerns and no change in the number of shares outstanding. We also assume that there are no incentive effects from options (and therefore option incentive effects do not cause employees to produce more in some states, nor change the cash flows or correlation of pre-tax inco m e and the m arket return). 25 Finally, we assume that no-option cash flows are positively correlated with the market, so the firms no - option cash flow beta is positive, as is the beta on no-option taxable income. Given these assumptions, how should tax liabilities be discounted for a firm that uses options as part of their compensation package (Note that the only place that we use a discount rate is within the simulation procedure, to discount the incremental future tax liability stream associated with earning an extra dollar in 2000). If a firm pays a fixed wage W, after tax income (ignoring carrybacks and carryforwards) is where CF is cash flow (before the effects of wages or options) and C is the corporate income tax rate. The min appears because tax liabilities cannot be negative. When min(CF, W)W, this becomes simply (CF-W)(1-t). For convenience, assume that wage payments are uncorrelated with stock prices. With option cash settlement and assuming that options have a negligible strike price, after-tax income is Both covariance terms in the braces will generally be positive, so the sign of the overall correlation between tax liabilities and stock price depends on whether the first covariance is larger than the second. Because cash flows are generally substantially larger than option deductions, the overall correlation between tax liabilities and stock price will typically be positive but, if the second term in the braces is large enough in absolute magnitude, the overall correlation can be negative. If the second term is small, the correlation does not differ much from the correlation in the no options case. It is an empirical matter as to whether the overall correlation is positive or negative. Using data for the firms in our sample, we determine that the correlation between tax liabilities and stock price is positive on average for the levels of these two variables, and also for percentage changes for these two series. Therefore, our argument is that the beta is positive for tax liabilities and the appropriate rate to discount tax liabilities lies somewhere between the risk - free rate and the equity rate. We show below that the implications in our paper do not change for various discount rates in this range. In the base case for this paper, to determine the present value of incremental tax liabilities associated with earning an extra dollar in 2000 (i. e. to determine the year-2000 MTR), we discount using a firm-specificequity rate. This is conservative relative to using a smaller discount rate because it will reduce the effect of changes in future tax liabilities on current-period MTRs. Discounting with an equity rate is an approximationbecause it misses the fact that option deductions are zero below some exercise price, and hence do not contain pure equity risk. It is also an approximation because it does not explicitly account for the associationbetween earnings and stock prices inherent in our approach (see Section D for details). However, these approximations likely have only modest effect because our ultimate variable of interest is the MTR, which is bounded between zero and 35 percent. 26 This i m plies that any err o rs we m ake in discounting will have an attenuated effect on our MTR estimates (because the MTR cannot vary outside of the range from zero to 35 percent, no matter how we discount). To ensure that our results are not sensitive to the discount rate, we conduct several sensitivity analyses. Technically, option deductions could be discounted as options rather than pure equity. Therefore, weimplement an approach based on the contingent claims valuation outlined in Schwartz and Moon (2000). Specifically, we assume an earnings risk premium of two percent per year, grow stock prices at the risk freerate, and discount everything at the risk-free rate. In another set of robustness checks, we follow our standard simulation approach but discount using very high (e. g. double the CAPM market-model discount rate)and very low (e. g. the risk-free rate) discount rates. The empirical results indicate that the discounting assumption has only a second-order effect on the estimated MTR. For example, doubling the discount rate reduces the estimated MTR 120 basis points relative towhat we report below and does not change the qualitative results. The Schwartz and Moon (2000) approach reduces the estimated MTR by 100 basis points. All other robustness checks on the discount rate lead tosmaller changes in the estimated MTR. While conceptually important, the choice of discount rate only has a modest effect on our empirical estimates of the MTR. This reflects the fact that the magnitude of historic, current and very near-term option deductions are the dominant effects on current MTRs, more so than distant option deductions (for which the discount rate would be more important). III. Empirical Analysis of the Effect of Option Deductions on Corporate MTRs A. Descriptive Statistics Table I presents descriptive statistics for the stock option disclosures of the SampP 100 and Nasdaq 100 samples. For both groups, the average expected option life is close to five years, although it is slightly shorter forNasdaq firms, consistent with the higher volatility for Nasdaq firms, possibly coupled with risk aversion, precipitating early exercise. Not surprisingly, given GAAP reporting requirements, the risk-free rate is verysimilar for the two samples, equaling approximately 6 percent. The small difference in the risk-free rate for the two samples probably reflects differences in year-ends (because risk-free rates should be similar forfirms with common year-ends), with non-calendar year-ends more common for Nasdaq firms. Dividend yield averages 1.5 percent for SampP 100 firms with most firms paying dividends. Conversely, few Nasdaq 100 firms pay dividends the mean dividend yield is 0.1 percent and the 75th percentile is zero. Annual stock return volatility is higher for Nasdaq 100 firms, with a mean volatility of 75 percent versus 36 percent for the SampP firms. The volatility of returns is important because it affects the probability that stock price appreciates greatly, which would lead to large option deductions in good scenarios. Table II summarizes firm characteristics. Not surprisingly, the market capitalization of the typical SampP 100 firm is roughly five times larger than that for Nasdaq 100 firms. However, there is substantial overlap betweenthe two distributions, with the 75th percentile of Nasdaq firms being one-third larger than the 25thpercentile of SampP firms. The difference in size between the two subsamples is more pronounced for total assets, reflecting the fact that Nasdaq valuation is based more prominently on intangibles and growth options. In terms of profitability, the median return on assets (ROA) is quite similar for the two samples, and is actually a little higher for the Nasdaq firms (4.9 percent) than for the SampP firms (4.7 percent). The 75thpercentilesare also similar for the subsamples. However, the dispersion of profitability is higher for Nasdaq firms, with a much higher proportion reporting losses. In fact, the 25th percentile ROA is 3.4 percent for the Nasdaqfirms versus 1.5 percent for the SampP firms. Nasdaq firms tend to use less debt in their capital structure, with a mean (median) debt ratio of 6.7 percent (1.0 percent) versus 17.5 percent (13.4 percent) for the SampP firms. Both samples have average betas of approximately one, although the SampP firms are slightly below one while the Nasdaq firms have betas slightly above one. Figure 1 summarizes the overall effect of option deductions on the year-2000 corporate MTR (i. e. the effect of all historic and future exercises). The histogram shows MTRs for all 190 firms in our sample, with andwithout the effects of options. Options cause a significant shift in MTRs. Before options, 24 percent of the sample face MTRs of less than 10 percent while after considering options, 35 percent face such rates. Similarly, before options 65 percent of the sample firms face MTRs above 30 percent as compared with 46 percent after factoring in options. In the next two sections, we analyze the effects of options separately for SampP and Nasdaq firms, and break out the effects by historic versus future exercise activity. B. Tax Effects for SampP 100 Companies Table III presents evidence on the effects of option deductions on MTRs, segregated by sample. The first row contains estimated MTRs for fiscal year-end 2000, produced using standard tax deductions and deferredtaxes to infer taxable income, but before taking stock options into account. This computation is comparable to the one used in Graham (1996), with the only differences being that we bound income growth andvolatility to lie within the 25thand 75th industry percentiles and that we discount the tax consequences of option deductions with the cost of equity. The median MTR for the SampP 100 firms in 2000 is the top statutoryrate of 35 percent while the mean is 29 percent, which is consistent with prior studies that show clustering at the upper end of the statutory rates. The 25th percentile MTR is 32 percent, reflecting the fact that mostSampP 100 firms face relatively high tax rates. However, the 5thpercentile is zero, consistent with a few SampP 100 firms not expecting to pay any taxes over a 23-year period (e. g. after carrying losses in 2000 back two years to 1998 and forward 20 years to 2020). The next three rows of Table III illustrate the impact of stock option deductions on MTRs. Recall that there are several groups of stock option deductions: already exercised (second row: MTR w exercised options), already granted but not yet exercised (third row: MTR w current grants), and notyet granted (fourth row: MTR w future grants). For the SampP 100 sample, we find that incorporating stock options into the simulations has relatively little effect on the MTRs. In the fourth row of Table III, when alloption deductions are considered (including future grants and future exercises), the median MTR is still 35 percent. For the 25th percentile, the estimated MTR drops to 26 percent from 32 percent. The fifth row of Table III summarizes the change in MTRs brought about by option deductions (MTR w future grants). Inferences are the same. Options materially reduce MTRs for only about one-fourth of SampPfirms. When we consider all options, the mean reduction is 1 percent. Among the firms with the largest drop in tax rates, the 25th percentile MTR falls 1 percent and the 5thpercentile MTR decreases 5 percent. Even though employee stock option deductions do not substantially reduce the MTR for many SampP 100 firms, the deductions have a noticeable effect on corporate tax liabilities. The bottom two rows of Table IIIpresent gross deductions expressed in dollar terms and as a percentage of earnings before tax. The mean SampP firm had 640 million of option tax deductions in 2000. With 99 firms in the sample, this implies totaldeductions of 63.4 billion. With aggregate pretax earnings of approximately 349 billion for SampP 100 firms, stock option deductions represent nearly one-fifth of aggregate pretax income. Option deductions are 4 perce n t of pretax inco m e for the m edian fir m. 12 perce n t for the 7 5 th percentile, and 111 percent for the 95thpercentile. To summarize, SampP 100 firms substantially reduce their tax liabilities through deductions for nonqualified, employee stock options. However, while option deductions reduce tax rates for some firms, the tax savings do not translate into significantly lower MTRs for the typical (highly profitable) SampP 100 firm. Though option deductions slash their tax bills, only about one-fourth of SampP 100 firms have enough deductions to (i) fully offset the current years pre-option income and also eliminate the past two years of taxable income, (ii) generate losses in 2001 and 2002 that can be carried back to fully offset income in 2000, or (iii) for currently nontaxable firms, delay when tax consequences are realized for year-2000 option deductions. One or more of these conditions must be met for option deductions to reduce MTRs. C. Tax Effects for Nasdaq 100 Companies Options dramatically affect the MTRs of Nasdaq 100 companies. The median MTR before options is 31 percent and the mean is 20 percent (see the bottom panel in Table III), suggesting that Nasdaq firms have relatively high MTRs before the effects of options, though not as high as the MTRs of SampP 100 firms. For the median firm, just considering historic exercises reduces the MTR from 31 percent to 15 percent. Incrementally considering options that are already granted but not yet exercised reduces the median MTR from 15 percent to 8 percent. Considering all forms of option deductions, including those from future grants, reduces the m edian MTR all the way down to 5 percent. Considering all deductions, the 75 th percentile drops from 35 percent to 26 percent, indicating that option deductions affect most Nasdaq 100 firms. The proportion of Nasdaq firms with a MTR less than 0.05 increases from 33 percent to 50 percent. This increase implies that half of the Nasdaq 100 firms anticipate paying very little in corporate taxes from 1998 (the beginning of the two-year carryback period for 2000 losses) to 2020 (the end of the carryforward period for 2000 losses). Overall, the mean (median) decrease in MTRs is eight (two) percent. The size of the decline is limited by the fact that MTRs are bounded below by zero. In 2000, the median Nasdaq 100 firm enjoyed option-related tax deductions of 173 million, with a mean of 388 million. Aggregating across the 91 firms in our Nasdaq sample, the resulting deductions total about 35billion. This figure is striking because it is larger than the 13 billion of aggregate earnings before taxes and option deductions for the Nasdaq sample in 2000. Note that these large deductions do not eliminate all taxesfor the Nasdaq 100 because some firms have pre-option income that exceeds option deductions and others have deductions that expire unused however, it does indicate the enormous magnitude of the optiondeductions. Figure 2 summarizes the effect of options on the MTRs of Nasdaq firms. Before options are considered, 52 percent of Nasdaq firms face MTRs exceeding 0.30 after considering options, only 18 percent do. Almost 60percent of the Nasdaq 100 face post-option MTRs below 10 percent and almost 30 percent face MTRs of approximately zero. If one were to ignore option deductions, these figures imply that most Nasdaq companieswould reap substantial tax advantages from tax shields, such as interest. After considering option deductions, only a minority of Nasdaq firms has much of a tax incentive to finance with debt. IV. Empirical Analysis of the Effect of Option Deductions on Debt Policy The preceding section indicates that the effects of stock options on MTRs can be substantial, especially among option-intensive companies. These substantial effects imply that option deductions might affect corporate policies for which the MTR is an important decision variable. In this section we explore whether the effect of option deductions on MTRs is important to corporate debt policy decisions. This investigation has the potential to help explain why some firms appear to use too little debt when the effects of option deductions are ignored. A. Univariate Analysis of Debt Policy Table IV presents Pearson and Spearman correlations between pre-interest MTRs and various measures of debt in the capital structure, specifically, debt-to-market value, debt-to - assets, and interest-to-marketvalue. We examine pre-interest MTRs because Graham, Lemmon, and Schallheim (1998) show that corporate tax status is endogenously affected by debt policy. That is, when a firm uses debt, the associated interestdeduction reduces taxable income and can also reduce the MTR, which induces a spurious negative correlation between debt ratios and tax rates. This endogeneity can be avoided by using pre-interest MTRs (that is, tax rates based on earnings before interest and tax) when examining the relation between debt ratios and tax rates. The first row (column) in Table IV displays the Pearson (Spearman) correlation between the debt variables and conventional pre-interest MTRs (MTR wo options), i. e. before the effects of interest and options. For allthree measures, for both Spearman and Pearson correlations the coefficients vary in sign and are insignificant (except for the Pearson correlation on interestvalue, which is significant but has the wrong sign). Thesecorrelations provide little evidence that capital structure is correlated with MTRs for our sample when we ignore options deductions. The second row and column show the relation when the computation of pre-interest MTRs is modified to include all employee stock option deductions (MTR w future grants). The relation is positive for all three debtvariables. For the Spearman correlations, the correlations range from 0.25 to 0.34 and are always significant at the 0.01 level. These results are consistent with managers making financing and compensation decisionsjointly, considering the effect of options on MTRs. 27 The third row and column present the correlations between the change in pre-interest MTRs resulting from options (MTR w future grants) and the other variables. Two points are worth noting. First, the correlation between the decrease in rates and the post-option MTRs is strongly positive, indicating that options have a significant effect on MTRs. Second, the decrease in rates is positively correlated with the amount of debt in the capital structure. This correlation implies that firms that use options intensively enough to reduce their MTR use relatively little debt, which is consistent with firms trading off options and interest deductions. B. Regression Analysis To further assess the relation between option deductions, MTRs and debt, Table V presents tobit regressions with debt-to-value as the dependent variable.28We use the tobit method because the debt ratio equalszero (i. e. is left-censored) for 17 firms in our sample. Since determining a debt ratio for a financial institution is problematic, we delete the 40 firms that have a primary or secondary division that is financial (2-digit SICcode between 60 and 69). For deletion, we require that the financial division contribute at least 10 to total firm revenue. This process leaves 150 firms (down from the 190 included in Section II). The first two columns of Table V are univariate and regress debt-to-value on MTR wo options and MTR w future grants, respectively. Like the correlation coefficients presented in Table IV, the coefficient on the MTRvariable, when all stock options are ignored, is insignificant. The coefficient on the MTR variable, when stock options are considered, is significantly and positively correlated with the debt ratio at the 0.01 level (seecolumn 2). In addition to being statistically significant, the coefficient estimate on the MTR variable is economically large. For example, consider the predicted debtvalue ratios for firms at the 25th and 75thMTR w future grantspercentiles (MTRs of 2.3 and 35.0, respectively). We gauge economic significance using the slope coefficient estimate of 0.23, the intercept of 0.05, and a tobit adjustment factor of 0.88 that accounts for the effectof using a censored normal distribution (Maddala (1983)). The implied debtvalue ratio is 0.049 (the 23rddebtvalue percentile) for a firm at the 25th MTR percentile, versus 0.115 (the 61st debtvalue percentile) for afirm at the 75th MTR percentile.29In other words, moving from the 25th percentile to the 75th percentile in the MTR distribution, the implied amount of debt in the capital structure more than doubles, from wellbelow the debtvalue median to well above. A number of nontax factors can affect debt policy, and it is important to control for these potential influences in a multivariate analysis. Controlling for such influences helps isolate tax effects and minimizes thepossibility that the tax variable proxies for some other factor. For example, financially weak firms face lower tax rates and also might face barriers to borrowing and therefore use options to save cash. It seems unlikelythat this condition drives the correlation between debt and tax rates because if the issue is simply that less profitable firms are less able to obtain debt financing, the relation between debt and MTRs before optionsshould be significant, but it is not. However, to ensure that differences in financial health do not drive our results, we include controls for financial strength in the regression: operating cash flow divided by assets andthe quick ratio. We also control for three other factors that are commonly thought to drive debt policy (see Rajan and Zingales (1995)): growth options, asset tangibility, and firm size. Firms with extensive growth options might useless debt to avoid the underinvestment problem (Myers (1977)). Shareholders of a firm with risky fixed claims in its capital structure will potentially underinvest by forgoing positive NPV investments because projectbenefits might accrue to the firms existing bondholders this problem is likely to be more severe among growth firms. Therefore, we expect firms with growth options, which we measure with research and development expense divided by sales, to use less debt. In contrast, firms with more tangible assets, as measured by property, plant and equipment divided by total assets, are less subject to underinvestment and informational asymmetry problems, and also have more assets to collateralize, and therefore can use more debt. Finally, largerfirms are thought to have better access to debt markets, which allows them to borrow more. We therefore expect a positive relation between debt ratios and firms size, which we measure with sales revenue. Note that data are missing for at least one of these explanatory variables for three observations, so the regressions that include control variables have 147 observations. Finally, though not shown in the tables, everyregression specification includes five industry dummy variables based on 2-digit SIC codes. We choose these five industries by performing a regression that includes a dummy for each 2-digit SIC code, and thenretaining the five that are significant: SIC codes 26 (paper and allied products), 40 (railroads), 48 (communications), 49 (utilities), and 78 (amusements). The third through sixth columns of Table V report results for tobit regressions that include tax rates and the control variables. To reduce any potential effect of endogeneity between debt policy and the explanatory variables, we use the lagged values of the control variables. The coefficients on the control variables have the correct signs and are generally significant. These estimated coefficients indicate that firms with many tangible assets use more debt but firms with substantial growth options (as measured by RampD) use less debt. Also, consistent with a pecking - order view (Myers and Majluf (1984)), firms with more cash flow use less debt. Finally, large firms use more debt than do small firms. More importantly for this study, in the third column, the control variables increase the significance of the pre-option tax rate, although it is only marginally significant at conventional levels (p-value of 0.07). In thefourth column, the coefficient on the tax rate that includes the effects of historic option deductions (MTR w exercised options) is larger and more significant than the no-options tax rate (p-value of 0.03). In the nexttwo columns, coefficients on the tax rates that consider the effects of currently granted options (fifth column) and also future option grants (sixth colu m n) are both significant at the 0.01 level. 30 The incr e asing statistical significance of the tax variables highlights the influence of stock option deductions on MTRs and debt policy. The rightmost column of Table V presents a specification that includes the control variables, the tax rate variable that ignores options, and the difference between the no-options tax rate and the MTR w future grants. By using two tax variables, we are able to examine the effects on debt policy oftraditional tax effects separately from the incremental effect of options. In this specification, the MTR wo options tax variable is significant at the 0.01 level, and the incremental effect of options is significant at the0.06 level, and both coefficients have the expected sign. The fact that the coefficient on MTR wo options becomes significant in the presence of the MTR wfuture grants variable is striking because it suggests thatthe effect of non-option factors is strengthened once options are accounted for. Further, the coefficients on the MTR wo options and MTR wfuture grants variables are similar suggesting that both option - relatedand non-option-related tax effects are of comparable importance in determining debt policy. Thus, we conclude that taxes affect capital structure decisions for reasons unrelated to, as well as directly related to, deductions that result from employee stock options. C. Robustness Checks of Regression Results We perform a number of robustness checks that consist of adding additional control variables or estimating the regressions on subsets of the data (see Table VI). Though the estimated coefficients are not shown in Table VI, the control and industry dummy variables from Table V are included in all of the Table VI specifications. First, we examine the tax variable based on Value Line growth estimates and stock price forecasts, rather than using historical data to estimate income growth and the CAPM to estimate stock returns. The leftmost column of Table VI indicates that the Value Line tax variable coefficient is 0.21 (and significant at the 0.01 level), which is identical to the base case results in Table V. Second, we include an SampP dummy variable (second column of Table VI). Suppose that our results are explained by differences between Nasdaq and SampP firms. Nasdaq firms may have low debt because of a nontaxeffect (e. g. perhaps because they have substantial growth options) and a low tax rate (possibly because growth firms often are currently or have recently been unprofitable). SampP firms may have high debt ratios andhigh tax rates. If so, then including an SampP dummy could cause the tax variable to be insignificant. In fact, the tax variable is less significant when the SampP dummy is included but it is still significant (p-value of 0.06). The third column summarizes the results of including stock volatility as a right-side variable. Firms with volatile returns might be considered risky and therefore have higher costs of debt and borrow less. The sign ofthe volatility coefficient is negative and consistent with this hypothesis but it is not significant. Importantly, the tax variable is still positive and significant even when the stock volatility variable is included as acontrol. The fourth column shows the results when a control variable measuring the dollar value of deductions, scaled by assets, is included. The purpose of this control is to rule out the possibility that the debt ratio is relatedsolely to a firms option intensity. The positive coefficient on the tax variable (p-value of 0.08) provides some assurance that the effect of the options on the MTRs has incremental value beyond merely identifyingoption-intensive firms. The fifth column of Table VI uses debt minus cash as the dependent variable. This allows negative debt for firms that have large cash holdings but very little or no debt, such as Microsoft. Because the dependentvariable is no longer censored at zero, we estimate the model with OLS. Again, the tax coefficient is positive and significant in this alternative specification. The sixth through tenth columns of Table VI show the results from performing the main regression specification on different subsets of data. The intent of these five specifications is to investigate whether thesignificant tax results might be driven primarily by the contrasting behavior of two types of firms (e. g. unprofitablelow-taxlow-debt versus profitablehigh - taxhigh-debt), or whether the tax effects also occur forsubsets of somewhat homogeneous firms for which theory predicts there should be tax effects. The sixth column investigates the 130 firms that report positive debt. We test whether option-affected tax rates provide a positive incentive to use debt for these firms. The tax coefficient in the sixth column (from an OLS regression) indicates that high tax rate firms do indeed use more debt than low tax rate firms. In the seventh column, we examine tax effects for the 120 firms that were profitable in 2000, to make sure that our overall results are not driven strictly by profitablehigh-tax firms using more debt than losslow-tax firms, perhaps for nontax reasons (like accessibility to debt markets). The next two columns further explore the accessibility of debt markets by considering firms that have an SampP bond rating (100 firms in column eight) or have an investment grade bond rating (72 firms in column nine). For all three subsets of these firms we find a positive and significant tax variable. Finally, in the rightmost column we examine the 101 firms that have annual growth in taxable income of at least 3.6 (the sample mean). Again, the tax variable is positive and significant. Overall, the results in Tables V and VI indicate that taxes exert a positive effect on the use of debt and that options use exerts a negative effect. These results are robust to a number of different specifications andsubsamples. D. The Relation Between Stock Option Deductions and Debt Conservatism The preceding sections link stock options and debt policy by documenting improved statistical power in detecting tax effects when MTRs incorporate option deductions. In this section we examine a direct measure ofdebt conservatism and test whether firms that appear to have the most unused debt capacity (when option deductions are ignored) use option deductions to reduce tax liabilities. Graham (2000) develops a measureof debt conservatism that he refers to as kink. Kink measures the proportion by which a firm could increase interest deductions without experiencing reduced marginal tax benefits for interest deductions. Forexample, consider a firm with EBIT of 2 million or more in every state of nature. If this firm has interest expense of 0.5 million (and we ignore carryforwards and carrybacks), it has a kink of 4.0 because it could quadruple interest deductions and still enjoy the full tax-reducing benefit of interest deductions in every state. (That is, even if it quadruples interest, the firm will not experience a tax loss in any state, so all taxbenefits are enjoyed in the current year). Graham notes that many large profitable firms, which presumably face small costs of debt financing, have large kinks and appear to potentially be underlevered. Grahamsanalysis, however, does not incorporate option deductions. We calculate kink for our sample firms based on pre-option income. (For computational reasons, we restrict the maximum kink to 8.0, as in Graham (2000)). The median (mean) kink is 8.0 (5.3) for our sample, whichappears to indicate debt conservatism. However, we uncover evidence consistent with conservative firms (i. e. those with large kinks) substituting option deductions in place of interest. The Pearson correlation inTable IV between kink and reduction in MTR is 0.23 (significant at 0.01 level), indicating that option deductions have the largest effect on MTRs for firms with large kinks (i. e. firms that appear to have the mostunused debt capacity when option deductions are ignored). Similarly, the Pearson correlation between option deductionsvalue and interestvalue is 0.28, which is consistent with firms substituting between optiondeductions and interest. Finally, when we recalculate kink based on EBT that subtracts options deductions, the mean kink falls to 4.3 from 5.3 (though the median kink remains at 8.0). The fact that the mean kink fallsby one-fifth indicates the importance of the economic effect of stock option deductions on capital structure. Overall, this evidence is consistent with firms that appear debt conservative (when options are ignored) using option deductions heavily in place of interest. However, the large mean kink of 4.3 (even after optiondeductions are considered) indicates that employee stock option deductions offer only a partial explanation for the conservative use of debt. Additional research is needed to more fully understand the apparentlyconservative debt policy at many firms. The tax deduction for nonqualified employee stock options is unusual. The company has little control over its timing or amount. Instead, the corporate deduction is delayed until employees choose to exercise. Theamount of the deduction is determined by the firms stock price years after the options are granted. This paper develops an approach for evaluating the complex and uncertain tax benefits associated with employeestock options, impounds the corporate tax savings in MTRs, and assesses the effects of the option deduction on debt policy. Incorporating option information from financial statement disclosures into Grahams (1996) MTR simulations, we compute MTRs that take account of option deductions. We then compare these firm-specific rates withcompanies debt levels in an attempt to assess the relation between tax shields associated with leverage and tax shields associated with option compensation. We find that employee stock options substantially reduce corporate taxes for both the industrial SampP 100 and the high-technology Nasdaq 100. For the more option-intensive Nasdaq 100, stock options dramatically reduce estimated MTRs, with the median rate tumbling from 31 percent to 5 percent. Consistent with the concerns raised in Hanlon and Shevlin (2002), our findings raise doubts about the usefulness of conventionalMTRs, which ignore stock option deductions. Unfortunately, developing MTRs that impound option deductions from public sources is costly because the option data must be hand-collected from financial statements. Because scholars, policymakers, practitioners, and analysts, among others, need MTRs for option-intensive companies, future research should consider developing a low-cost method of estimating MTRs thatincorporates the effects of stock option deductions. We document a positive relation between leverage and post-option MTRs. Moreover, we find that firms that use little debt also use options extensively. These results provide at least a partial explanation forconservative debt usage at highly profitable, option-intensive firms, such as Microsoft and Dell. By presenting evidence that options provide an important non-debt tax shield that substitutes for interest in the spiritof DeAngelo and Masulis (1980), this paper extends our understanding of the role of taxes in financial decisions. Cipriano, M. D. Collins, and P. Hribar, 2001, An empirical analysis of the tax benefit from employee stock options, Working paper, University of Iowa. Core, J. and W. Guay, 2001, Stock option plans for non-executive employees, Journal of Financial Economics 61, 253-287. DeAngelo, H. and R. Masulis, 1980, Optimal capital structure under corporate and personal taxation, Journal of Financial Economics 8, 3-29. Desai, M. 2002, The corporate profit base, tax sheltering activity, and the changing nature of employee compensation, Working paper, Harvard University. Easton, P. and T. Harris, 1991, Earnings as an explanatory variable for returns, Journal of Accounting Research 29, 19-36. Fama, E. and K. French, 2002, The equity premium, Journal of Finance 57, 637-659. Graham, J. R. 1996, Debt and the marginal tax rate, Journal of Financial Economics 41, 41-73. Graham, J. R. 2000, How big are the tax benefits of debt Journal of Finance 55, 1901-1941. Graham, J. R. 2003, Taxes andcorporate finance: A review, Review of Financial Studies 16, forthcoming. Graham, J. R. and C. R. Harvey, 2002, Expectations of equity risk premia, volatility, and asymmetry from a corporate finance perspective, Working Paper, Duke University. Graham, J. R. M. Lemmon and J. Schallheim, 1998, Debt, leases, taxes, and the endogeneity of corporate tax status, Journal of Finance 53, 131-161. Hall, B. and J. Liebman, 2000, The taxation of executive compensation, in James Poterba, ed.: Tax Policy and the Economy 14 (MIT Press, Cambridge, Massachusetts). Hanlon, M. and T. Shevlin, 2002, Accounting for the tax benefits of employee stock options and implications for research, Accounting Horizons 16, 1-16. Huddart, S. and M. Lang, 1996, Employee stock option exercises: An empirical analysis, Journal of Accounting and Economics 21, 5-43. Kahle, K. and K. Shastri, 2002, Firm performance, capital structure, and the tax benefits of employee stock options, Working Paper, University of Pittsburgh. Maddala, G. S. 1983, Limited Dependent and Qualitative Variables in Econometrics (Cambridge University Press, Cambridge, U. K.). McDonald, R. 2002, The tax (dis)advantage of a firm issuing options on its own stock, Working Paper, Northwestern University. Modigliani Franco, and Merton Miller, 1958, The cost of capital, corporate finance and the theory of investment, American Economic Review 48, 261-297. Myers, S. 1977, Determinants of corporate borrowing, Journal of Financial Economics 3, 799- Myers, Stewart, and Nicholas Majluf, 1984, Corporate financing and investment decisions when firms have information that investors do not have, Journal of Financial Economics 13, Plesko, G. 2003, An evaluation of alternative measures of corporate tax rates, Journal of Accounting and Economics forthcoming. Rajan, R. G. and L. Zingales, 1995, What do we know about capital structure choice Some evidence from international data, Journal of Finance 50, 1421-1460. Scholes, M. Wolfson, M. Erickson, M. Maydew, E. and T. Shevlin, 2002, Taxes and Business Strategy: A Planning Approach (Prentice Hall, Upper Saddle River, N. J.). Schwartz, E. and M. Moon, 2000, Rational pricing of internet companies, Financial Analysts Journal 56:3, 62-75. Shevlin, T. 1990, Estimating corporate marginal tax rates with asymmetric tax treatment of gains and losses, Journal of the American Taxation Association 12, 51-67. Sullivan, M. 2002, Stock options take 50 billion bite out of corporate taxes, Tax Notes, March 18, 2002, 1396-1401. Microsofts Stock Option Plan Footnote for the year ended June 30, 2000 The Company has stock option plans for directors, officers, and employees, which provide for nonqualified and incentive stock options. Options granted prior to 1995 generally vest over four and one-half years and expire 10 years from the date of grant. Options granted during and after 1995 generally vest over four and one-half years and expire seven years from the date of grant, while certain options vest either over four and one-half years or over seven and one-half years and expire after 10 years. At June 30, 2000, options for 341 million shares were vested and 734 million shares were available for future grants under the plans. Stock options outstanding were as follows: For various price ranges, weighted average characteristics of outstanding stock options at June 30, 2000 were as follows: The Company follows Accounting Principles Board Opinion 25, Accounting for Stock Issued to Employees, to account for stock option and employee stock purchase plans. An alternative method of accounting for stock options is SFAS 123, Accounting for Stock-Based Compensation. Under SFAS 123, employee stock options are valued at grant date using the Black-Scholes valuation model, and this compensation cost is recognized ratably over the vesting period. Had compensation cost for the Company8217s stock option and employee stock purchase plans been determined as prescribed by SFAS 123, pro forma income statements for 1998, 1999, and 2000 would have been as follows: The weighted average Black-Scholes value of options granted under the stock option plans during 1998, 1999, and 2000 was 11.81, 20.90, and 36.67. Value was estimated using a weighted average expected life of 5.3 years in 1998, 5.0 years in 1999, and 6.2 years in 2000, no dividends, volatility of 0.32 in 1998 and 1999 and 0.33 in 2000, and risk-free interest rates of 5.7 percent, 4.9 percent, and 6.2 percent in 1998, 1999, and 2000. Descriptive Statistics on Option Characteristics All variables are from the Black-Scholes option valuation assumptions in the company financial statement footnotes. Expected life is years from grant until average exercise. The risk-free interest rate is the rate on zero-coupon U. S. government issues with remaining term equal to the expected life of the options. Dividend yield is dividends as a percentage of share price. Annual return volatility is the standard deviation of the continuously compounded rates of return on the stock (i. e. standard deviation of the difference in the natural logarithm of stock prices). Descriptive Statistics on Firm Characteristics Asset is total assets. Market equity is the value of common equity at fiscal year-end. Return on assets is net income divided by assets. DebtValue is total debt divided by the market value of the firm. Beta is themarket-model beta as reported on CRSP. Effect of Employee Stock Option Deductions on Marginal Tax Rates This table summarizes the effect of option deductions on corporate marginal tax rates (MTRs) for all 190 firms for which we can calculate tax rates. MTR wo options is a simulated MTR, assuming there are no employeestock option deductions, based on earnings before tax (EBT). A simulated MTR accounts for the tax-loss carryback and carryforward features of the tax code. MTR w exercised options is the simulated rate except thathistoric deductions from options exercised in 1998, 1999, and 2000 are subtracted from EBT. MTR w current grants is the simulated MTR, with historic deductions and future deductions associated with already grantedoptions deducted from EBT. MTR w future grants is the simulated MTR, with historic deductions, future deductions for already granted options, and deductions for not-yet-granted options deducted from EBT. MTRw future grants is MTR w future grants minus MTR wo options, so a negative number indicates that option deductions lead to a reduction in the tax rate. 2000 Stock Option Deductions is the dollar figure (in millions)of option deductions in 2000. 2000 Deductions Pretax Income is 2000 deductions divided by pre-option EBT. The columns show the mean and standard deviation across all sample firms, as well as the 5 th. 2 5 th. 5 0 th. 7 5 th. a n d 9 5 th percentiles. Correlations Between MTRs and Leverage for Combined Sample of SampP 100 and Nasdaq 100 Firms in 2000 Pearson (Spearman) correlations between corporate MTRs and various measures of debt policy appear above (below) the main diagonal. MTR wo options is a simulated MTR, assuming there are no employee stockoption deductions, based on earnings before tax (EBT). MTR w future grants is the simulated MTR, with historic deductions, future deductions for already granted options, and deductions for not-yet-granted optionsdeducted from EBT. MTR w future grants is MTR w future grants minus MTR wo options. Debt-to-value is total debt divided by the market value of the firm, where market value equals book assets minus bookequity plus market equity. Debt-to-assets is total debt divided by total assets. Interest-to - value is debt interest divided by market value. Deductions-to-value is the dollar amount of option deductions in 2000 dividedby market value. Kink in the proportion by which interest could be increased before the value of incremental interest deductions would begin to fall. Kink is calculated as in Graham (2000) using pre-option earnings. Ahigh value for kink can be interpreted to mean that a firm has unused debt capacity (ignoring the effect of option deductions). These correlations are for the 150 firms included in the regression analysis. , , means statistically different from zero at the 0.01, 0.05, and 0.10 levels, respectively. Significance for the tax variables tests whether the correlation coefficient equals zero versus the alternative that the coefficient isgreater than zero. Tobit Regressions of Debt-to-Value on Marginal Tax Rates and Control Variables Results are from cross-sectional regressions using data from 2000. The dependent variable is Debt-to-value (total debt divided by the market value of the firm, where market value equals book assets minus book equity plus market equity). MTR wo options is a simulated MTR, assuming there are no employee stock option deductions, based on earnings before tax (EBT). MTR w exercised options is the simulated rate exceptthat historic deductions from options exercised in 1998, 1999, and 2000 are subtracted from EBT. MTR w current grants is the simulated MTR, with historic deductions and future deductions associated with alreadygranted options deducted from EBT. MTR w future grants is the simulated MTR, with historic deductions, future deductions for already granted options, and deductions for not-yet-granted options deducted from EBT. MTR w future grants is MTR w future grants minus MTR wo options. PPampEAssets is property, plant, and equipment divided by total assets. Quick ratio is cash plus receivables, the sum divided by current liabilities. Cash Flow is operating cash flow divided by total assets. RampD is research and development expense divided by sales. Sales is sales revenue. Five significant 2-digit SIC code dummies are included in all specifications but are not shown in the table. Regression coefficients and P-values (in parentheses) are shown. More Regressions of Debt-to-Value on Marginal Tax Rates and Control Variables Results are from cross-sectional regressions using data from 2000. The dependent variable is Debt-to-value (total debt divided by the market value of the firm, where market value equals book assets minus bookequity plus market equity). MTR w future grants is the simulated MTR, with historic deductions, future deductions for already-granted options, and deductions for not-yet-granted options deducted from EBT. MTR wfuture grants (Value Line) is the same simulated tax variable, based on stock price and growth projections from Value Line. SampP dummy is an indicator variable that takes on a value of one for SampP firms and zero for Nasdaq firms. Stock Volatility is the volatility of stock returns. Option DeductionsAssets is the dollar value of tax deductions from employee stock options divided by total assets. Though not shown in the table, eachregression includes PPampEAssets, Quick ratio, Cash Flow, RampD, Sales, and five 2-digit SIC code dummies. Regression coefficients and P-values (in parentheses) are shown for the tax variable(s) and new controlvariables. The rightmost five columns summarize regressions that include, respectively, only firms that have nonzero debt, earnings greater than zero, an SampP bond rating, an investment grade bond rating, and annualgrowth larger than the mean growth in taxable income for the sample (3.6). The four leftmost columns include all firms with nonmissing values for the explanatory variables. The regressions are all tobitspecifications, except for the Dep vardebt-cash and Debtgt0 columns, which are OLS. debt-cash allows negative debt because the dependent variable is total debt minus the firms cash holdings, the quantitydivided by the market value of the firm. Corresponding author: Fuqua School of Business, Duke University, Durham NC 27708-0120, (919) 660-7857 (phone), (919) 660-8038 (fax), john. grahamduke. edu We appreciate excellent research assistance from Courtney Edwards, Allison Evans, Laura Knudson, and Julia Wu and insightful comments from an anonymous referee, Alon Brav, John Core, Richard Frankel, DavidGuenther, John Hand, Mike Lemmon, Ed Maydew, Hamid Mehran, Vikas Mehrotra, Dan Rogers, Richard Sansing, Jim Schallheim, Jake Thomas, Mike Weisbach, workshop participants at the University of Colorado, Cornell, Duke, MIT, the University of North Carolina, Wharton and audience participants at the 2003 American Accounting Association and American Finance Association meetings. Bob McDonald and Terry Shelvinscomments were especially helpful. All data are publicly available. Lang was visiting the University of Queensland when the first draft of this paper was completed. Graham acknowledges financial support from theAlfred P. Sloan Research Foundation. 1We use a Scholes et al. (2002) MTR that accounts for the present value of current and future tax consequences associated with changes in todays income. 2It is important to note that this amount does not imply a reduction in overall tax revenues because it fails to take into account the increase in individual tax burdens associated with option exercise. In particular, employees exercising nonqualified options face potential tax obligations for the difference between the market and strike price at the time of exercise. 3Strictly speaking, our results are consistent with managers trading off interest and option deductions in 2000. In other years, when option deductions are less important, tax planners may accelerate non-option deductions. It would be interesting for future research to investigate whether managers trade off non-option deductions with interest in eras where option deductions are less prominent. 4 State m ent of Financial Accounting Standards (SFAS) 123 per m its fir m s t he choice of either expensing stock options on the income statement or disclosing in the footnotes the effect stock options would have had if expensed. In 2000, it was extremely rare for a firm to expense stock options on the income statement, with the vast majority of firms opting for footnote disclosure. If a firm opted not to expense options, it was not permitted toreduce tax expense for the deductions related to option exercise. The underlying logic was that, since the original charge did not reduce pretax income, the tax benefit at exercise should not decrease tax expense. 5Tax benefits from option deductions are sometimes explicitly reported on two financial statements: the statement of cash flows and the statement of shareholders equity. However, tax benefits from options are not always reported as a separate lineitem and instead are often aggregated with another item on these statements. 6In 1998, the most recent year for which IRS data are available, the firms in our sample had tax expense equal to more than one-third of the taxes paid for the entire corporate sector. 7Of the three missing companies, two are foreign companies (Erickson and Checkpoint). The other (JPM) is not listed on Edgar for unspecified reasons. 8In the sample, 124 firms have December 2000 year-ends, and 22 have year-ends between September and November 2000. Another 20 have year-ends in 2000 earlier than September, and in eight of these cases we use 1999 data because the year-end is in May (and 10-Ks for fiscal year 2000 were not available when we collected the data). Finally, the remaining 24 companies have year-ends between January and May 31, 2001. 9To estimate the effects of the stock market run-up, we perform a robustness check in which we assume that historic stock prices and returns, as well as historic grant and exercise prices, are only half what they actually were. Even with dampened stock prices, the sheer number of options granted and exercised is such that this robustness check produces a mean tax rate that is only 40 basis points higherthan the base case tax rate we report below. 10Stock option deductions can show up in our pre-option measure of taxable income if they affect deferred taxes. This should only occur when option deductions contribute to tax loss carryforwards (Hanlon andShevlin (2002)). Due to data limitations, we are unable to determine the extent to which this occurs in our sample. Therefore, in our main analysis we assume that option deductions do not affect deferred taxes. Wealso perform an unreported robustness analysis in which we do not adjust income for deferred taxes, thereby guaranteeing that options do not affect our pre-option earnings figure. Relative to the base case resultsreported below, the mean tax rate is 70 basis points lower in this no deferred taxes adjustment analysis but the qualitative implications are unchanged. 11See Plesko (2003) for a comparison of the actual MTR based on the tax return versus estimated tax rates based on financial statement data, such as the simulation tax rate used in this paper. Note that Pleskos analysis ignores potentially important dynamic features of the tax code, such as tax loss carrybacks and carryforwards, by using a static tax return tax rate as the benchmark. Nonetheless, Plesko concludes that, of the various tax variables he considers, the simulated tax rate is the most highly correlated with tax return tax rates. 12This approach is consistent with the common procedure of using industry inputs when calculating a firms cost of capital. Note that our qualitative results do not change if we do not bound growth rates andvolatility to lie within the respective industry interquartile ranges, nor if we set each firms growth and volatility equal to industry medians. 13Specifically, SFAS 123 states that the fair value of a stock option (or its equivalent) granted by a public entity shall be estimated using an option-pricing model (for example, the Black - Scholes or a binomial model)that takes into account as of the grant date the exercise price and expected life of the option, the current price of the underlying stock and its expected volatility, expected dividends on the stock , and the risk-freeinterest rate for the expected term of the option. Appendix B of SFAS 123 provides detailed guidance on estimating the inputs into the valuation formula, and firms are required to disclose the assumptions used invaluation. 14Most companies with multiple plans combine all plans into one aggregate disclosure. In the 12 cases in which firms separate information across plans, we aggregate shares and use weighted averages of variablessuch as share price and expected term to exercise. Similarly, exercise decisions are disclosed separately for 13 sample firms (e. g. cancellations separated from forfeitures or reloads separated from new grants), and Black-Scholes assumptions are disclosed separately for 15 firms (e. g. different expected lives for executives relative to non-executive employees). Again, we aggregate the disclosures and use a weighted average of the variables, weighted by the number of options in the respective plan. Twenty-eight companies disclose a range for Black-Scholes assumptions, and five disclose a range of exercise prices rather than a weighted average, perhaps reflecting the fact that they use different assumptions for different groups of employees. In these cases, we use the midpoint of the range because sufficient detail is not available to calculate a weighted average. Finally, eight firms disclose dividends per share rather than dividend yield. In these cases, we compute dividend yield based on year-end share price. In total, 73 firms report in one of these nonstandard formats. If we exclude these 73 firms, the mean tax rate increases by approximately 150 basis points, but the overall implications of our study do not change. 15For example, using the Microsoft footnote disclosure in the appendix for the year ended June 30, 2000, we estimate the 2000 tax deduction for stock options to be 13,925,340,000, which is the product of the 198 million options exercised and the difference in the weighted average grant price of 79.87 and the weighted average strike price of 9.54. 16We do not explicitly incorporate vesting schedules because the stock option footnotes are often vague and indicate a range of vesting periods. Further, our use of expected lives should incorporate the effects of vesting. To get a sense for the typical vesting schedule, we gather the available information from the option footnotes. The average vesting period (using the midpoint when a range is indicated) is 3.5 years for our sample firms, and most firms indicate that vesting occurs ratably over time, typically beginning within the first year. As a result, our assumption that option exercise is spread over the period beginning two years priorto and ending two years following the expected life (4.8 years on average) seems consistent with the likely vesting schedules. Huddart and Lang (1996) suggest that exercise is common immediately following vestingdates. On another note, it is possible that in 2000 the expected option life that companies report in the footnotes is low by historic standards, due to the bull market of the 1990s, which may have encouraged earlyexercise and shorter option lives. To investigate how a longer expected life would affect our results, we perform a robustness check in which we add two years to the expected life of all options. The mean estimatedtax rate in this analysis is only 20 basis points higher than what we report below, and overall qualitative results are unchanged. 17For example, the Microsoft footnote disclosure in the appendix reports a weighted average expected life of 6.2 years, and an expiration of 10 years, for options granted in 2000. Thus, we assume the optionsgranted in 2000 will be exercised evenly over the period from 2004 to 2008 if they are in the money by at least 15 percent during those years. If they are not in the money by 15 percent, exercise is deferred until thefirst year in which they are in the money by 15 percent. In 2010 (the presumed date of expiration), all options that remain outstanding are exercised if they are in the money by any amount. 18For example, the Microsoft footnote disclosure in the appendix reports grants (cancellations) of 138 (25) million in fiscal year 1998, 78 (30) in 1999, and 304 (40) in 2000. We assume that fiscal year 2001 grants are141.7 million (i. e. 173.3 million (the mean of 1998, 1999, and 2000 grants) less 31.6 million (the mean of 1998, 1999, and 2000 cancellations)) times a growth factor. 19In unreported analysis, we find qualitatively similar results when we perform our calculations based on sales revenue growth, rather than income growth. Sales growth rates are typically much larger than incomegrowth rates in our sample, so we use the latter so that our future options grant numbers are conservative. 20In a robustness check, we use an estimated risk premium of 8.1 (the Ibbotson historic average). This premium leads to a mean tax rate that is 40 basis points lower than the base case mean reported below. Allresults are qualitatively similar whether we use an 8.1 or a 3 risk premium. 21While we directly link growth of earnings and expected stock prices, we do not directly link realized future earnings and stocks prices. That is, we use the realized draw for earnings growth on a given earnings pathfor a given period (15 percent in our example) to determine the mean expected stock price growth for that period on the associated stock price path (18 percent in our example). However, on top of that mean, we layer a variance based on the past returns series and draw a return from that distribution. The resulting realized return can be substantially different from 18 percent because of high return variances. In fact, thecorrelation of simulated earnings and simulated stock prices is approximately 15 percent in our analysis, which puts our simulated correlation in line with that observed empirically by Easton and Harris. 22A related issue is the potential that management makes decisions based on unrealistic or optimistic expectations of future returns. We do not believe that reasonable alternative management beliefs would greatly alter our results. For example, if we set the expected return to 15 and halve the variance of expected returns to capture optimistic managerial beliefs, the mean tax rate falls by only 13 basis points relative towhat is reported below. 23Since the annual stock price is based on log returns, implied prices cannot be negative. Note also that if we assume that volatility is 25 for all firms (rather than using the volatility firms report in the footnotes), the mean tax rate is only 10 basis points different from that reported below in the base case. 24While the per-share option deduction is directly the result of stock price appreciation, the correlation between option deductions and contemporaneous-year returns is likely to be well below one for at least two reasons. First, options are typically exercised in about the fifth year of their lives and the per-share deduction is determined by the multi-year return, so the current year return is a relatively small part of thededuction. Second, the number of options exercised is a function of many factors beyond current year return (e. g. prior exercise, cancellation, marketstrike ratio and liquidity concerns), so the current year returnmay be high but exercise low because employees opt not to exercise. 25 W e thank Terry Shevlin for pointing out these incentive possibilities. W e thank Bob McDonald for suggesting the basic framework that we discuss next. 26We thank Bob McDonald for pointing this out. 27This interpretation is consistent with our conversations with tax managers at several high - technology companies. Although these firms appear profitable based on their income statements, the managers indicatethat debt is not particularly attractive because the company pays little in taxes. Similarly, this result may explain why Microsoft and Dells derivatives trading is not as tax-inefficient as implied by the effective tax ratesreported in their financial statements (McDonald (2002)). 28A potential concern is that share price movements can affect both the debt-to-value ratio and stock option deductions (and hence MTRs). To investigate this issue, we also estimate the regressions with debt-to-assets replacing debt-to-value. Consistent with the high correlation between debt-to-value and debt-to-assets in Table IV, regression results for debt-to-assets are qualitatively similar to those for debt-to-valuethough weaker statistically. 29 The calculation is 0.88 x (0.055 0.23 x 0.023) 0.049 and 0.88 x (0.055 0.23 x 0.350) 0.115. 30 The adjusted-R2 is 60 percent in an OLS version of the regression in the sixth column. Previously published by the Duke University, June 2003 We find that employee stock option deductions lead to large aggregate tax savings for Nasdaq 100 and SampP 100 firms and also affect corporate marginal tax rates. For Nasdaq firms, including the effect of options reduces the estimated median marginal tax rate from 31 percent to 5 percent. For SampP firms, in contrast, option deductions do not affect marginal tax rates to a large degree. Our evidence suggests that option deductions are important nondebt tax shields and that option deductions substitute for interest deductions in corporate capital structure decisions, explaining in part why some firms use so little debt. FORTHCOMING in the Journal of Finance in 2004 Employee Stock Options, Corporate Taxes and Debt PolicyEmployee Stock Options, Corporate Taxes, and Debt Policy We find that employee stock option deductions lead to large aggregate tax savings for Nasdaq 100 and SP 100 firms and also affect corporate marginal tax rates. For Nasdaq firms, including the effect of options reduces the estimated median marginal tax rate from 31 to 5. For SP firms, in contrast, option deductions do not affect marginal tax rates to a large degree. Our evidence suggests that option deductions are important nondebt tax shields and that option deductions substitute for interest deductions in corporate capital structure decisions, explaining in part why some firms use so little debt. Copyright 2004 by The American Finance Association. Om du har problem med att hämta en fil, kontrollera om du har rätt program för att visa den först. Vid ytterligare problem läs IDEAS hjälp sida. Observera att dessa filer inte finns på IDEAS-webbplatsen. Var tålmodig eftersom filerna kan vara stora. Eftersom tillgången till det här dokumentet är begränsat kanske du vill leta efter en annan version under Relaterad forskning (nedan) eller söka efter en annan version av den. Article provided by American Finance Association in its journal The Journal of Finance . Volume (Year): 59 (2004) Issue (Month): 4 (08) Pages: 1585-1618 References listed on IDEAS Please report citation or reference errors to. or. if you are the registered author of the cited work, log in to your RePEc Author Service profile. click on citations and make appropriate adjustments. Myers, Stewart C. Majluf, Nicholas S. 1984. Corporate financing and investment decisions when firms have information that investors do not have , Journal of Financial Economics. Elsevier, vol. 13(2), pages 187-221, June. Graham, John R. 1996. Debt and the marginal tax rate , Journal of Financial Economics. Elsevier, vol. 41(1), pages 41-73, May. Myers, Stewart C. Majluf, Nicols S. 1945-, 1984. Corporate financing and investment decisions when firms have information that investors do not have , Working papers 1523-84. Massachusetts Institute of Technology (MIT), Sloan School of Management. Eugene Fama F. Kenneth R. French, undated. The Equity Premium. quot , CRSP working papers 522, Center for Research in Security Prices, Graduate School of Business, University of Chicago. Eugene F. Fama Kenneth R. French, 2002. The Equity Premium , Journal of Finance. American Finance Association, vol. 57(2), pages 637-659, 04.