The effect of taxes on the debt policy of spanish listed companies

Clemen e-Almend os, José A.; Sogo b-Mi a, F ancisco A icle The e ec o axes on he deb policy o spanish lis ed companies SERIEs - Jou nal o he Spanish Economic Associa ion P o ided in Coope a ion wi h: Spanish Economic Associa ion Sugges ed Ci a ion: Clemen e-Almend os, José A.; Sogo b-Mi a, F ancisco (2016) : The e ec o axes on he deb policy o spanish lis ed companies, SERIEs - Jou nal o he Spanish Economic Associa ion, ISSN 1869-4195, Sp inge , Heidelbe g, Vol. 7, Iss. 3, pp. 359-391, h ps://doi.o g/10.1007/s13209-016-0147-4 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/158561 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h p://c ea i ecommons.o g/licenses/by/4.0/ SERIEs (2016) 7:359–391 DOI 10.1007/s13209-016-0147-4 ORIGINAL ARTICLE The e ec o axes on he deb policy o spanish lis ed companies José A. Clemen e-Almend os1·F ancisco Sogo b-Mi a2 Recei ed: 20 Feb ua y 2015 / Accep ed: 29 June 2016 / Published online: 15 July 2016 © The Au ho (s) 2016. This a icle is published wi h open access a Sp inge link.com Abs ac This s udy explo es he ole o axes in explaining companies’ inancing decisions. We es whe he he co po a e ax shields explana ion o capi al s uc u e is applicable o i ms lis ed on he Spanish s ock exchange o e he pe iod 2007– 2013. Taxes a e ound o be economically and s a is ically signi ican de e minan s o capi al s uc u e. Ou esul s sugges ha ma ginal ax a es a ec he deb policies o Spanish lis ed companies, and he exis ence o non-deb ax shields cons i u es an al e na i e o he use o deb as a ax shel e . Consis en wi h heo e ical expec a ions, he e is a s onge ela ion be ween deb and axa ion in less le e ed i ms. Finally, we empi ically es ima e he impac o he new hin-capi aliza ion ule pu o h by he Spanish go e nmen in 2012 on he inancing beha iou o Spanish lis ed companies. F ancisco Sogo b-Mi a g a e ully acknowledges inancial suppo om Minis y o Economy and Compe i i eness Resea ch G an ECO2015-67035P. The au ho s wish o hank he Co-Edi o , Manuel Bagues, and wo anonymous e e ees o Jou nal o he Spanish Economic Associa ion (SERIEs) o insigh ul sugges ions and ad ices ha subs an ially imp o ed his pape in many ways. We a e also g a e ul o Juan Ayuso om Banco de España and Domingo Ga cía om Bolsas y Me cados Españoles o hei help in p o iding some economic and inancial da a. Finally, we would also like o hank Pankaj Sinha, An onio Ruiz-López, Juan A. Sanchis-Llopis, Juan M. Villa-Lo a, Anna Told á-Sima s (discussan ) and he pa icipan s a he XXII Finance Fo um a Uni e si y o Za agoza in 2014 and he Resea ch Semina a Uni e si y o Valencia in 2015 o help ul discussions and use ul commen s on p e ious d a s o his pape . Any e o s a e ou sole esponsibili y. BF ancisco Sogo b-Mi a [email p o ec ed] José A. Clemen e-Almend os [email p o ec ed] 1Depa amen o de Emp esa, EDEM Escuela de Emp esa ios, Ma ina Real Juan Ca los I, Muelle de la Aduana s/n, 46024 Valencia, Spain 2Depa amen o de Economía y Emp esa, Uni e sidad CEU Ca denal He e a, Plaza Reyes Ca ólicos 19, 03204 Elche (Alican e), Spain 123 360 SERIEs (2016) 7:359–391 Ou empi ical e idence suppo s he exis ence o a ax e o m e ec , whe e companies a ec ed by in e es deduc ibili y limi a ions educe hei le e age mo e han companies ha a e no a ec ed. Keywo ds Capi al s uc u e ·Co po a e axes ·Deb ·Ma ginal ax a e · Thin-capi aliza ion ule JEL Classi ica ion C33 ·G32 ·H25 1 In oduc ion A la ge body o esea ch has examined he e ec s o co po a e axa ion. Al hough he esul s o empi ical models a y signi ican ly, he majo i y o his esea ch does ind ha , o some deg ee, axes in luence a b oad ange o co po a e inancial decisions such as inancing policy, in es men policy o co po a e eo ganiza ion and hedging.1 The magni ude o hese e ec s and hei o e all impac on he economy a e s ill unde deba e. No wi hs anding, he mos signi ican obs acle a policy make con on s in deciding on he ax ea men o co po a e deb and equi y inancing is ha he impac o axa ion on co po a e inancial policy is no en i ely unde s ood. In addi ion, G aham (2013) e iews a numbe o s udies ha sugges ha axes in luence inancing decisions; howe e , his e ec is no always s ong. Likewise, he concludes ha mo e esea ch is needed o a be e unde s anding o he in luence o axes on capi al s uc u e, pa icula ly ela ed o ime-se ies e ec s. The e o e, whe he and o wha ex en axa ion a ec s he choice o capi al s uc u e is s ill an unse led opic, dese ing u he s udy. The me a-s udy o he exis ing empi ical s udies conduc ed by Feld e al. (2013) concludes ha capi al s uc u e choices a e indeed posi i ely a ec ed by axes, an e ec which is also quan i a i ely ele an . Tax a es a e shown o be co ela ed wi h co po- a e capi al s uc u e choices, which sugges s ha i ms may inc ease alue h ough op imal deb choice. The ade-o heo y o capi al s uc u e o e s a heo e ical expla- na ion o he ela ionship be ween co po a e deb policy and axes. Speci ically, his heo y a gues ha i ms de e mine hei op imal deb a io by compa ing he p esen alue o addi ional ax sa ings and o he addi ional expec ed cos o inancial dis ess caused by a ma ginal inc ease in deb . The e has been ela i ely limi ed empi ical esea ch in o he e ec s o ma ginal co po a e ax on deb policy, despi e i s clea signi icance. In his ega d, G aham (1996a), as well as he subsequen s udies, ound ha ma ginal co po a e ax a e does in luence he deb policies o US i ms.2In coun- ies o he han he US, Alwo h and A achi (2001) conduc ed a simila analysis using a da a panel on I alian i ms and ound a posi i e ela ionship be ween i m-speci ic ma ginal ax a es and I alian i ms’ deb policy. In addi ion, Kunieda e al. (2011), Ha mann-Wendels e al. (2012) and Sinha and Bansal (2013) ob ained analogous 1A de ailed e iew o he li e a u e on he ole o axes in co po a e inance is p o ided by G aham (2008, 2013). 2A comp ehensi e su ey o ela ed li e a u e can be ound in G aham (2003,2008,2013). 123 SERIEs (2016) 7:359–391 361 esul s o Japanese, Ge man and Indian i ms, espec i ely. To he bes o ou knowl- edge, he e ha e been no empi ical s udies o da e on he e ec s o simula ed ma ginal ax a es on deb policy in Spain. In he a ea o public inance, ecen deba e abou co po a e ax e o m has ocused on he consequences o asymme ic ax ea men o equi y and deb inancing. US and Eu opean iscal au ho i ies ha e conside ed limi ing he abili y o companies o deduc in e es paymen s om axable income, as well as calling o equal ea men o equi y and deb . Some examples a e he Comp ehensi e Business Income Tax (CBIT) p oposal by he US T easu y, he Mi lees Re iew p oposals o he UK ax sys em o he Resolu ion o he ECOFIN Council Mee ing o June 8, 2010, which ecommended o Eu opean Union membe S a es he adop ion o hin-capi aliza ion ules.3The eason o his is ha he ax- a ou ed s a us o deb has educed ax e enue collec ion and supposedly encou aged a “deb bias” whe eby ax incen i es encou age companies o use ex a deb . In his ega d, i is belie ed ha excessi e use o deb inancing inc eases i ms’ p obabili y o becoming inancially dis essed and he eby exace ba es o pe haps e en causes economic down u ns. Acco ding o Mooij (2011), al hough he exis ence o deb in he capi al s uc u e did no cause he inancial c isis, excessi e le e age makes i ms mo e ulne able o economic shocks and he e o e deb bias migh ha e con ibu ed o he ex en o he c isis. The main objec i e o his s udy is o analyse he ela ionship be ween axes and deb inancing using panel da a on Spanish lis ed companies. Mo e speci ically, we ocus on how he deduc ibili y o deb in e es a ec s he capi al s uc u e o i ms. Ou empi ical analysis is based on a sample o Spanish lis ed i ms o he pe iod 2007– 2013. We es he hypo hesis ha companies ha e a ax incen i e o use deb inancing a he han equi y inancing because in e es paid is ax-deduc ible while di idends paid o sha eholde s a e no . Besides, we use he She lin (1990) and G aham (1996a) expec ed ma ginal ax a e app oach o examine he e ec s o ax on he deb policies o Spanish i ms. In addi ion, we es he non-deb ax shields hypo hesis which conside s o he ax shel e s di e en om he in e es allowances. In he ime pe iod analysed, he Co po a e Tax Income Law was e o med and his ac migh ha e in luenced he deb policy o Spanish lis ed companies. Fo ha eason, we es o a ax e o m e ec and conside his shock as a quasi na u al expe imen o ou esea ch. This pape con ibu es o he exis ing li e a u e on he impac o co po a e axa ion on i ms’ capi al s uc u e, u he de eloping he con ibu ions o p e ious li e a u e in di e en ways. Fi s ly, we p o ide addi ional empi ical e idence on he ela ion- ship be ween axes and deb inancing. In con as wi h o he pape s, ou measu e o le e age includes only inancial deb and di ec ly excludes o he liabili ies such as ade payables, which mainly depend on business ansac ions and no on he e ec o co po a e axa ion. Secondly, ou indings shed some ligh on his issue in he Eu opean Union, which has ecei ed li le a en ion o da e in he li e a u e. Mo e- o e , In e na ional Financial Repo ing S anda ds we e adop ed in Spain on Janua y 3Thin-capi aliza ion e e s o when a company is inanced wi h a high le el o deb ela i e o equi y. In u n, hin-capi aliza ion ules imply ha a company ha has oo much deb compa ed o equi y will be denied iscal deduc ions o pa o i s in e es paymen s, o ha pa o in e es paymen s will be eclassi ied as di idends and will no ob iously be conside ed as iscal deduc ions. 123 362 SERIEs (2016) 7:359–391 1s 2007, which allows meaning ul compa ison be ween ou esul s and hose om o he economies ha ha e also implemen ed hese in e na ional s anda ds. Thi dly, we ake in o accoun he Spanish co po a e ax e o m in 2012, as an exogenous shock, which enac ed a new hin-capi aliza ion ule limi ing he ax deduc ibili y o inancing expenses. Besides, and as simila ax e o ms ha e been conduc ed in many OECD coun ies, we belie e ha ou conclusions migh be po able o o he se ings. Applying a di e ence in di e ences app oach, we analyse he po en ial impac o he abo emen- ioned e o m. Finally, we s udy a special pe iod pa ially cha ac e ized by a se e e economic and inancial c isis ha has d ama ically a ec ed Eu opean Medi e anean coun ies such as Spain. Ou indings show ha ma ginal ax a es signi ican ly a ec he deb policy o Spanish lis ed i ms. The esul s a e consis en wi h he signi icance o co po a e axes in company inancing decisions conside ing he uniqueness o he Spanish ax p o- isions. As expec ed, he e is a s onge ela ionship be ween axes and deb policy in less le e ed companies. Fu he mo e, he exis ence o non-deb ax shields con- s i u es an al e na i e o he use o deb as a ax shel e . Rega ding he co po a e ax income e o m app o ed by he Spanish Go e nmen in 2012, we ound ha he new hin-capi aliza ion ule po en ially a ec ed 28 % o he companies in ou sample. On a e age, hese companies had highe deb a ios han hei non-a ec ed coun e - pa s, and a e he e o m, he o me g oup on a e age displayed s onge declining deb a ios as compa ed o he la e one. Ou analysis p o ides empi ical e idence consis en wi h a ax e o m e ec . The emainde o he pape is s uc u ed as ollows. The nex sec ion analyses he heo e ical amewo k o he s udy and p esen s he hypo heses o be es ed. Subse- quen ly, he Spanish co po a e ax legisla ion is desc ibed in Sec . 3, including he new hin-capi aliza ion ule. In Sec . 4, we discuss he a iables de ini ions, and explain he es ima ion o companies’ ma ginal co po a e ax a es. The ea e , Sec . 5p o ides a desc ip ion o ou sample and analyses desc ip i ely he ax da a. The empi ical model speci ica ion, econome ic me hodology and he esul s a e discussed in Sec . 6.Se - e al obus ness checks a e p esen ed in Sec . 7, and he inal sec ion d aws some concluding ema ks. 2 Theo e ical ounda ion and hypo heses de elopmen Modigliani and Mille (1963) we e he i s o in oduce he idea ha co po a e axa ion a ec s he capi al s uc u e o i ms. In pa icula , hey showed ha when co po a e income is axed and deb in e es is a deduc ible expense, i m alue can be inc eased by using deb inancing a he han unding en i ely wi h equi y. In his con ex , he inc ease in a i m’s alue is due o he deb ax shield. The ques ion o why deb inancing has adi ionally ecei ed a ou able ax ea men whe eas equi y inancing has no , seems likely o be he esul o his o ical o ces a he ime he ax ules we e being de eloped, a he han any weigh y economic easoning pe aining o con empo a y economic o business ci cums ances (S ebulae and Whi ed 2012). Ea lie empi ical a icles did no ind con incing e idence ha axa ion a ec ed i ms’ inancial policy (see o example, B adley e al. 1984;Ti man and Wessels 1988). 123 SERIEs (2016) 7:359–391 363 These discou aging esul s led Mye s (1984) o s a e in his enowned P esiden ial Add ess o he Ame ican Finance Associa ion ha “we don’ know how i ms choose hei capi al s uc u es as he e is no s udy clea ly demons a ing ha a i m’s ax s a us has p edic able, ma e ial e ec s on i s deb policy”. The me a-analysis by Feld e al. (2013) sugges s ha e y small o e en nega i e ax es ima es ound in he s udies do no accu a ely e lec deb esponse o axes. I seems di icul o conduc an e ec i e analysis o a di ec ela ionship be ween ax a es and deb policy, as mos la ge co po a ions ha e he same s a u o y ax a e.4In mos de eloped coun ies, he s a u o y ax laws do no demons a e any subs an ial a ia ion in co po a e s a u o y ax a es o e he yea s and ac oss i ms. In he absence o a ia ion in ax a es h ough ime and ac oss companies, we can only p esume a simila deb policy o each company, which is no he case, o we may end up wi h con adic o y esul s. Due o asymme ic ax ea men o co po a e p o i s and losses, he (expec ed) ma ginal ax a e may no be equal o he s a u o y ax a e. Speci ically, al hough he s a u o y ax a e is applied when he axable income o a company is posi i e, no co po a e ax is imposed when he axable income is nega i e. E en in cases whe e a company ac ually pays ze o ax in a yea due o incu ed losses, i s ma ginal ax a e may be non-ze o. In such a case he ma ginal ax a e is equal o he discoun ed alue o he axes paid on he ma ginal uni o income in he i s yea whe e he i m is expec ed o ha e posi i e axable income. Likewise, losses can ypically be ca ied o wa d and ca ied backwa d in he co po a e ax sys em, which leads o di e ences in he ma ginal ax a es. This dynamic dimension o axes makes i necessa y o o ecas u u e axable income in o de o es ima e cu en -pe iod ax a es and ax incen i es. Recognizing he exis ence o loss ca ied o wa d and ca ied backwa d in he US co po a e ax sys em, Mackie–Mason (1990) analysed he e ec s o he ma ginal ax a e on deb policy. He ound ha when a company has loss ca ied o wa d and in es men ax c edi (i.e., ano he ax shield), i is less likely o aise capi al by new deb issue. Since bo h exis ing loss ca ied o wa d and in es men ax c edi a e subs i u es o new deb issue in e ms o ax sa ings, his esul is consis en wi h he ade-o heo y. She lin (1990) implemen ed he Mon e Ca lo me hod using a simple linea p ojec- ion o axable income based on ac ual pas da a o simula e u u e axable income. Then, using simula ed axable income se ies and applying US co po a e ax law, he es i- ma ed he (expec ed) ma ginal ax a es o indi idual i ms. Also using his app oach, G aham (1996a) analysed he e ec s o ma ginal ax a e on US i ms’ deb policy. He ound a posi i e ela ionship be ween he i m-speci ic ma ginal ax a e and he change in deb a io. Con e sely, he e a e o he empi ical s udies using s a u o y ax a es o a e - age/e ec i e ax a es as p oxies o ma ginal ax a es (see in e alia B adley e al. 1984;T eze an 1992;Shum 1996;Sogo b-Mi a 2005;De Jong e al. 2008). These subs i u es o ax a es, howe e , a e p oblema ic in ha hey in oduce a signi ican 4S a u o y ax a es a e hose pe cen age a es es ablished by he ax law. Con e sely, ma ginal ax a es ela e o he ax a e a ibu able o he speci ic company’s ac i i y and o explici decisions ha may in ol e axes paid (o sa ed) and income ecei ed (o expenses paid) o e se e al yea s; hey can be de ined as he p esen alue o cu en and expec ed u u e axes paid on an addi ional uni o income ea ned oday. 123 364 SERIEs (2016) 7:359–391 downwa d bias in es ima es i po en ial endogenei y bias is no deal wi h. Acco d- ingly, Feld e al. (2013) s a e ha he simula ed ma ginal ax a es sugges ed by G aham (1996a) o e he ad an age o a oiding a signi ican downwa d bias in es ima ion. Fu - he mo e, G aham (1996b) and Plesko (2003) show ha he simula ion app oach is he bes a ailable p oxy o he “ ue” ma ginal ax a e. In pa icula , i is p e e able o simply using a iables ha a e assumed o be highly co ela ed wi h ma ginal ax a es, such as s a u o y ax a es, dummies which indica e whe he a i m is epo ing losses o icho omous a iables, such as hose used in Byoun (2008)o G opp (2002). Ou i s and main hypo hesis ollows di ec ly om he heo e ical a ionale and empi ical e idence discussed p e iously, and is o mula ed as: “Since highe ma ginal ax a es aise he alue o ax sa ings, ma ginal ax a es should be posi i ely ela ed o i ms’ deb policy” (Hypo hesis 1). DeAngelo and Masulis (1980) in oduced he idea o ax shield subs i u ion, which con ends ha holding in es men (and hence expec ed income) cons an , deb in e es compe es wi h o he allowable deduc ions as ax shel e . Fo example, i a mo e gene - ous ax ule inc eases he i m’s dep ecia ion allowance, hen he i m’s op imal le el o deb should dec ease due o i s lowe alue as a ax shield. The e o e, i ms can subs i u e non-deb ax shields o deb ax shields. Following his a ionale, i ms wi h a la ge amoun o non-deb ax shields will ha e lowe le els o deb han i ms wi h a small amoun o non-deb ax shields. Acco ding o he deb subs i u ion hypo hesis, he e should be a nega i e ela ionship be ween non-deb ax shields and deb usage. In his con ex , Mackie–Mason (1990) highligh s he ac ha he ax shield sub- s i u e hypo hesis o DeAngelo and Masulis (1980) is mo e applicable o i ms ha a e close o being ax exhaus ed (i.e., i ms ha ha e a high p obabili y o losing he deduc ibili y o hei ax shields). T eze an (1992) e e s o his as he ax exhaus ion hypo hesis. Mo eo e , Mackie–Mason (1990) akes issue wi h DeAngelo and Masulis (1980) by poin ing ou ha i ms wi h mo e p o i able p ojec s end o ha e la ge amoun s o bo h dep ecia ion and bo owing, and he e o e non-deb ax shields may ha e a posi i e a he han a nega i e associa ion wi h le e age. In o de o iden i y he e ec o deb subs i u ion on ax exhaus ion and p o i abili y, Mackie–Mason (1990) p oposes conside ing no only non-deb ax shields bu also he p obabili y o bank- up cy. I is likely ha non-deb ax shields a e a deb subs i u e o companies nea bank up cy and he e o e nea o ax exhaus ion. Con e sely, inancially heal hy com- panies ha a e a om ax exhaus ion may join ly exploi bo h deb and non-deb ax shields. Hence ou second hypo hesis can be o mula ed as: “Non–deb ax shields on a s and-alone basis, should be posi i ely ela ed o i ms’ deb policy and non-deb ax shields, weigh ed by he p obabili y o bank up cy, should be nega i ely ela ed o i ms’ deb policy” (Hypo hesis 2). A s eam o empi ical esea ch ha e examined he impac o axes on he inancing decisions o i ms using ax e o ms as na u al expe imen s. In his sense, changes in he ax sys em a e used as exogenous shocks o analyse whe he companies espond as p edic ed by heo y. Rep esen a i e wo k in his ield includes, bu is no limi ed o, Alwo h and A achi (2001), An (2012), Panie e al. (2013), Doidge and Dyck (2015), Faccio and Xu (2015), and Heide and Ljungq is (2015). A pa icula a ea wi hin his esea ch deals wi h he ela ionship be ween hin-capi aliza ion ules o o he 123 SERIEs (2016) 7:359–391 365 in e es deduc ion es ic ions and company capi al s uc u e decisions. Albe e ns and Su e h (2015), and D eßle and Scheue ing (2015) in es iga e empi ically he impac o in oducing a limi a ion o he in e es iscal deduc ibili y in he cou se o he Ge man co po a e ax e o m o 2008. They all ind e idence o he impac o such hin-capi aliza ion ule on companies’ deb a io; speci ically, companies ha a e a ec ed by he in e es ba ie educe hei le e age ypically mo e han companies ha a e no a ec ed. Con e sely, Blouin e al. (2014) examine he impac o hin- capi aliza ion ules ha limi he ax deduc ibili y o in e es on he capi al s uc u e o he o eign a ilia es o US mul ina ionals in 54 coun ies. In line wi h p e ious s udies, hey ca e ou a signi ican deb - educing e ec o di e en hin-capi aliza ion ules on o eign pa ne s’ deb . Based on he abo emen ioned discussion and aking he oppo uni y ha o e s he Spanish co po a e ax e o m in 2012, we s a e ou hi d and las hypo hesis: “Companies a ec ed by he new hin-capi aliza ion ule educe hei le e age a io a e he e o m mo e han hose companies ha a e no a ec ed” (Hypo hesis 3). 3 The Spanish co po a e ax se ing The egula ion o co po a e ax in Spain is con ained in he Consolida ed Tex o he Co po a e Income Tax Law, app o ed by Legisla i e Royal Dec ee 4/2004, o Ma ch 5 h, and in he Co po a e Income Tax Regula ion app o ed by Royal Dec ee 1777/2004, o July 30 h. I is wo h no ing ha Spain is poli ically o ganised as a S a e o Au onomous Regions, and wo o hese egions, namely he Basque Coun y and Na a a, ha e special ax egimes called Economic Acco d and Economic Ag eemen , espec i ely, which di e o he o he egions. Unless o he wise s a ed, we will ocus on he co po a e ax egime gene ally applied in mos o he Au onomous Regions in Spain. Co po a e ax is de e mined by he s a u o y ax a e imes axable income. The Spanish legisla o educed he s a u o y ax a e om 32.5 % o iscal yea 2007 o 30 % o iscal yea s 2008–2013. On he o he hand, co po a e axable income is de ined as he di e ence be ween pe iod e enues and pe iod expenses.5Business expenses a e deduc ible i hey a e p ope ly eco ded and suppo ed. By con as wi h o he coun ies, Spanish co po a e income ax ea s income esul ing om he ans e o asse s in he same way as o he income. Acco dingly, such income is gene ally added o (deduc ed om) egula business income o compu e he axable income. Co po a e axable income is based on he income disclosed in he inancial s a e- men s and accoun ing eco ds, adjus ed in acco dance wi h ax p inciples. The 2007 Spanish Gene al Accoun ing Plan app o ed by Royal Dec ee 1514/2007, o No embe 16 h di e en ia es be ween he cu en income ax expense (income) and he de e ed income ax expense (income). The o al ax expense o income is he sum o hese wo i ems, which should none heless be quan i ied sepa a ely. On he one hand, he cu en income ax expense is he amoun payable by he company as a esul o income ax 5The ax pe iod is he company’s business yea . The annual ax e u n mus be decla ed and he ax paid wi hin 25 days ollowing he 6 mon hs a e he end o he business yea . 123 366 SERIEs (2016) 7:359–391 se lemen s o a gi en yea . Con e sely, he de e ed income ax expense e lec s in essence he ecogni ion and se lemen o de e ed ax asse s and liabili ies. A de e ed ax asse o liabili y ep esen s he inc ease o dec ease in axes payable o e und- able in u u e yea s as a esul o empo a y di e ences and any ne ope a ing loss o ax c edi ca y- o wa ds ha exis a he epo ing da e. I s alue is compu ed wi h e e ence o inancial epo ing s anda ds o book income and ax ules o axable income. Fo ins ance, de e ed ax asse s can be c ea ed by he ax au ho i y ecogniz- ing e enues and/o expenses ou side o he imes se ou in he accoun ing s anda ds. In Spain “ ax e ec accoun ing”, which includes he concep o ne ax de e ed asse s, was i s in oduced in iscal yea 2007. As in he majo i y o de eloped economies, he Spanish co po a e ax sys em ea s p o i s and losses asymme ically and allows ca yo e o co po a e losses. The Span- ish ax code allows companies o ca y o wa d losses o o se axable income in u u e yea s, bu unlike in o he coun ies such as he US, Spanish i ms canno “ca y back” cu en losses o ecei e a ax e und o axes paid in ecen yea s. Con en ionally, inancial expenses ha e been la gely conside ed as deduc ible om a iscal poin o iew in Spain. The only es ic ion o his ule was hose si ua ions wi h excess deb wi h ela ed en i ies no esiden in he Eu opean Union, wi h he excep ion o hose s ablished in a e i o y classi ied as ax ha en. The excess deb was compu ed as he di e ence be ween he ac ual company’s indeb edness and he esul o applying a coe icien o h ee o equi y, excluding income o loss o he yea . The in e es gene a ed by he excess deb was no conside ed as iscally deduc ible; a he , i was ea ed exac ly he same as a di idend. In spi e o he p eceding ule, co po a e axpaye s could always submi o he Spanish iscal au ho i ies a p oposal o applying a highe a io; i i was accep ed, a di e en le e age h eshold would be applied.6On Ma ch 30 h, 2012, he Spanish Go e nmen app o ed se e al ax measu es wi h e ec om iscal yea s beginning om Janua y 1s , 2012.7Among such measu es, he ax e o m in oduced new ules a ec ing he deduc ibili y o inancial expenses. In pa icula , i de oga ed he o me Spanish hin-capi aliza ion egime and eplaced i by a b oade ule ha s ablishes limi a ions o he deduc ibili y o inancial expenses incu ed in excess o a gi en pe cen age o a Spanish bo owe ’s adjus ed ope a ing p o i s ega dless o whe he o no he deb is wi h ela ed pa ies. Unde he new ax egime, all ne inancial expenses (i.e. excess o inancial expenses in espec o inancial income) incu ed by a Spanish co po a e axpaye in a gi en yea ha exceed 30 % o such company’s annual ope a ing p o i s8will be non-deduc ible o co po a e ax pu poses. No wi hs anding, he e is a loo le el o he p e ious limi a ion, and i is ixed a 1 million eu os o ne inancial expenses. Hence, ne inancial expenses less o equal o 1 million eu os shall be ax-deduc ible ega dless o he le el o a company’s ope a ing p o i s in a gi en yea . 6This possibili y was no applicable o ansac ions made wi h o by pe sons o en i ies esiden in coun ies o e i o ies legally de ined as ax ha ens. 7These measu es we e included in Royal Dec ee Law 12/2012. 8They basically co espond o ea nings be o e in e es , axes, dep ecia ion and amo iza ion (EBITDA) wi h ce ain adjus men s. Fo mo e in o ma ion on his issue, e e o Royal Dec ee Law 12/2012. 123 SERIEs (2016) 7:359–391 373 No es: S a u o y Tax Ra e is he co po a e income ax a e es ablished by he Spanish co po a e ax code; MTR is he ma ginal ax a e es ima ed as pe She lin (1990) and G aham (1996a); MTREBIT is he ma ginal ax a e es ima ed using ea nings be o e in e es and axes as pe G aham e al. (1998); ETR is axes paid on p o i s di ided by p e- ax book income, excluding ex ao dina y and discon inued i ems; ETRb is axes acc ued on p o i s di ided by p e- ax book income, excluding ex ao dina y and discon inued i ems. -0.050 0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 2007 2008 2009 2010 2011 2012 2013 S a u o y Tax Ra e MTR MTREBIT ETR ETRb Fig. 1 Tax a iables, 2007–2013 losses and by he loss ca y- o wa d p o ision in he Spanish co po a e ax sys em. The s anda d de ia ion o he ma ginal ax a es is 7.68 % (7.98 % o MTREBIT), implying ha he e is mode a e a ia ion in he ma ginal ax a es o all i ms. The e o e The pe cen age o angible asse s o e o al asse s a e ages abou 11 %. The a e age size o he companies included in he sample is app oxima ely e802 million in e ms o ma ke alue o asse s. Besides, he a e age p o i abili y o ou sample amoun s o 5.51 %, and he ma ke o book a io shows a 1.61 a e age alue. We ha e calcula ed he co ela ion ma ix and, addi ionally, we ha e pe o med a mul icollinea i y es using he Va iance In la ion Fac o (VIF). Resul s a e epo ed in Table 9in he Appendix, and he low VIF alues sugges ha he e is no collinea i y among he a iables conside ed. In o de o compa e ou ma ginal ax a es a iables (i.e. MTR and MTREBIT) wi h o he adi ional ax measu es, we ha e calcula ed ETR which is he a io o axes paid on p o i s o p e- ax income, excluding ex ao dina y and discon inued i ems, as well as ETRb, whose nume a o is axes acc ued on p o i s and has he same denomina o as he p eceding a io. Figu e 1shows he ime e olu ion o he s a u o y ax a e, MTR, MTREBIT, ETR and ETRb. Du ing he pe iod 2007–2013, s a u o y ax a es emained mos ly s able. Con- e sely, om 2008 onwa ds, he e is an inc easing numbe o companies wi h losses (i.e. p e- ax book income and EBIT <0). Simul aneously, he numbe o companies wi h MTR and MTREBIT below 10 % inc eases. O e all, he e is a downwa d end in bo h MTR and MTREBIT. Figu e 2depic s he dis ibu ion o simula ed ma ginal ax a es, calcula ed using ea nings be o e in e es and axes as he base o axable income (MTREBITs) o he sampled i ms om 2007 o 2013, and an agg ega ion ac oss all yea s in he sample. The da a e eal subs an ial a ia ion in he ma ginal ax a e ac oss i ms and o e 123 374 SERIEs (2016) 7:359–391 No es: MTREBIT is he ma ginal ax a e es ima ed using ea nings be o e in e es and axes as pe G aham e al. (1998). <5% <10% <15% <20% <25% <30% 30% 0.000 0.100 0.200 0.300 0.400 0.500 2008 2009 2010 2011 2012 2013 2008-2013 <5% <10% <15% <20% <25% <30% 30% Fig. 2 Ma ginal ax a e (m ebi ) dis ibu ion ime.20 In any gi en yea , abou 2 % o he i ms ha eMTREBITs equal o he op s a u o y ax a e, oughly 10 % ha e MTREBITs below he 5 %, while he es ha e MTREBITs anging be ween 5 % and he highes a e (i.e. 30 %). The c oss-sec ional a ia ion in ax a es occu s because o he ca y- o wa d ea u es o he ax code. The ela i ely la ge pe cen age o low ax a es is due o he ac ha o e 27 % o he obse a ions in he sample ep esen i ms wi h nega i e axable income. 6 Empi ical s a egy and esul s As Robe s and Whi ed (2013) s a e in hei su ey pape , he mos impo an and pe asi e issue con on ing s udies in empi ical co po a e inance is endogenei y. In he case o es ing he impac o axes on companies’ inancing decisions, he di icul y s ems om he ac ha any measu e o ma ginal ax a es based on ac ual balance shee da a is no exogenous. This is due o he ma ginal ax a e’s dependence on pas inancing decisions: he highe he le e age a io, he lowe he axable income and he expec ed ma ginal ax a es because o he in e es deduc ibili y. This may esul in a nega i e ela ionship be ween le e age a ios and es ima ed ma ginal ax a es e en i high axes encou age companies o use deb as a inancing ins umen . Consequen ly, he endogenei y o he ax s a us may p oduce a spu ious co ela ion be ween he le e age a io and he ma ginal ax a e, making i di icul , i no i ually impossible, o d aw causal in e ences. As discussed in Sec . 2, we ha e o mula ed h ee empi ical hypo heses o ou esea ch. The co ec es ing o hese hypo heses equi es he o e coming o he endo- genei y issue, and as such we apply wo di e en me hodologies: eg ession based on 20 This igu e and he one co esponding o he o he ma ginal ax a e a iable (i.e. MTR) u ned ou o be e y much alike. 123 SERIEs (2016) 7:359–391 375 obse ables and di e ence in di e ences app oach. Bo h o hem will allow us o es Hypo heses 1 and 2, and he la e one will be used in o de o es Hypo hesis 3. 6.1 Reg ession app oach Con en ionally, he e ha e been wo possible solu ions o he endogenei y p oblem in he empi ical li e a u e (G aham e al. 1998). The i s esembles a adi ional way o implemen ing an endogenous eg esso in econome ics, ha is, using he lagged alue o he simula ed ma ginal ax a e as an explana o y a iable. In u n, since he simula ed ma ginal ax a e based on he p e- ax income (and a e in e es ) al eady inco po a es he i m’s le e age choices, a second possible solu ion o he endogenei y p oblem uses income be o e in e es o compu e ma ginal ax a es. The e o e, his second s a egy conside s he con empo aneous alue o he ma ginal ax a e as an explana o y a iable bu simula ed on a be o e- inancing basis, i.e. wi h ea nings be o e in e es . Ou empi ical esea ch will use he la e solu ion o a oid he endogenei y o ma ginal ax a es. No wi hs anding, we will check as a obus ness es he ins umen al a iable solu ion. Ou eg ession model speci ies companies’ capi al s uc u e as a unc ion o ax and con ol a iables, which we e discussed in de ail in Sec . 4. Speci ically, we use a s a ic model o le e age because we a e no in e es ed in an economic model o he dynamic adjus men owa ds an op imal le el o le e age. As he e is an inc emen al basis o ou dependen a iable, we use—as pe G aham (1996a)— he changes in possible de e minan s as explana o y a iables, excep o he a iables MTREBIT, NDTS*RISK and RISK. The e o e, ou eg ession model equa ion is as ollows: LEVi =β0+β1·MTREBITi +β2·NDTSi +β3·NDTSi ∗RISK +β4·RISKi +β5·TANGi +β6·SIZEi +β7·PROFi +β8·TOBINSQ i +ηi+η +εi ,(5) whe e LEVi is a measu e o le e age o i m iin yea ;MTREBIT ep esen s he ma - ginal ax a e es ima ed wi h ea nings be o e in e es and axes; NDTS is he non-deb ax shield a iable; RISK is he p obabili y o bank up cy a iable; TANG is he angi- bili y a iable; SIZE deno es he na u al loga i hm o o al asse s; PROF is he e u n on asse s a io; TOBIN’S Q is he ma ke o book asse s a io; ηi ep esen s ime-in a ian unobse able i m-speci ic e ec s (e.g., managemen pe o mance, epu a ion, e c.)21; η ep esen s ime-speci ic e ec s which a e common o all i ms and may change o e ime (e.g., mac oeconomic condi ions); and εi is he dis u bance e m. Table 2shows he es ima ion esul s o ou eg ession model o bo h le e age measu es LEV1and LEV2.22 21 Lemmon e al. (2008) show ha co po a e capi al s uc u es a e cha ac e ized by an impo an i m speci ic e ec . 22 To e i y he cha ac e — ixed o andom—o he unobse able indi idual e ec s, we use Hausman es . This es ejec s he null hypo hesis ega ding he absence o co ela ion be ween he unobse able e ec s and he explana o y a iables, which leads o he selec ion o he ixed-e ec es ima o . 123 376 SERIEs (2016) 7:359–391 Table 2 Es ima ion esul s o he capi al s uc u e model om Eq. (5) Fixed–e ec eg ession coe icien s es ima ed om Eq. (5) wi h obus s anda d e o s in b acke s. Table 7in he Appendix p o ides de ini ions o all he a iables. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els Explana o y a iables LEV1LEV2 MTREBIT 2.188** (0.938) 1.078 (0.975) NDTS 1.265 (1.601) 1.536 (1.724) NDTS*RISK −0.395*** (0.073) −0.122 (0.091) RISK −0.001* (0.000) −0.001* (0.000) ΔTANG 0.244* (0.129) 0.239* (0.136) ΔSIZE 0.281*** (0.077) 0.219*** (0.076) ΔPROF 0.298* (0.1704) 0.296 (0.2475) ΔTOBINSQ −0.159*** (0.045) −0.083*** (0.017) Obse a ions 434 434 R-Squa ed Wi hin 0.200 0.121 In he case o he eg ession o LEV1, wi h he excep ion o NDTS all pa ame- e es ima es ha e he expec ed signs and a e s a is ically signi ican . To add ess he economic signi icance o he es ima ed coe icien o 2.188 on MTREBIT epo ed in Table 2, conside he impac on le e age policy esul ing om a mo emen om a e - age MTREBIT o 0.182 (see Table 1) o he maximum o he sample pe iod (0.300). All else equal, a hypo he ical i m wi h a ma ginal ax a e o 30.00 % would inc ease he use o ne deb (i.e. he change in i s deb a io) by 25.86 % poin s, compa ed o an iden ical i m wi h a ma ginal ax a e o 18.20 %. I we ge he s anda dized eg ession coe icien s, MTREBIT appea s o be he s onges p edic o in he model and a one s anda d de ia ion inc ease in MTREBIT leads o a 0.720 s anda d de ia ion inc ease in p edic ed le e age change, wi h he o he a iables held cons an . When LEV2is used as dependen a iable, he mos impo an di e ence om he p e ious esul s is ha MTREBIT is no longe s a is ically signi ican . The e o e, we pa ially con i m ou Hypo hesis 1 whe eby deb policy and ma ginal ax a es a e posi i ely ela ed. Mo eo e , he esul s in he case o dependen a iable LEV1 suppo ou Hypo hesis 2 ega ding non-deb ax shields. Con ol a iables show he ypical and expec ed coe icien signs. 6.2 Di e ence in di e ences app oach An al e na i e app oach o o e come he endogenei y p oblem o he ax s a us in capi al s uc u e esea ch, is o look o exogenous changes in ax laws, and hen analyse how companies eac o hose changes by adjus ing hei deb a ios. In pa icula , we s udy he impac o he 2012 e o m o he Spanish co po a e ax income. Acco dingly, we use a di e ence in di e ences app oach (DiD)23 and di ide ou i ms’ sample in o a ea men g oup and a con ol g oup. We design he ea men g oup by iden i ying 23 We e e o Robe s and Whi ed (2013) o an in dep h e iew o econome ic echniques aimed a add essing endogenei y p oblems, including echniques such as DiD ha ely on a clea sou ce o exogenous a ia ion. 123 SERIEs (2016) 7:359–391 377 hose companies ha a e likely o be a ec ed by he new hin-capi aliza ion ule. Fo ha eason, he classi ica ion c i e ia ela es o he ollowing: •A company is included in he ea men g oup i , in 2012, i s ne inancial expenses exceed 30 % o EBITDA, conside ing ne in e es expense su passes 1 million eu os, i is conside ed as a iscal g oup, and i s ax esidence is loca ed in any Spanish Au onomous Region excep he Basque Coun y o Na a a.24 •A company is assigned o he con ol g oup i i does no mee any o he p e ious c i e ia. Fo ins ance, a company whose ax esidence is loca ed in he Basque Coun y o Na a e will belong o his g oup albei i s inancial expenses le el. Acco ding o he p e ious ea men and con ol g oups de ini ions, 25 companies (175 obse a ions) we e classi ied as ea ed and he o he 63 companies (441 obse a ions) we e included in he con ol g oup. The e o e, app oxima ely 28 % o he companies in ou sample we e po en ially a ec ed by he new hin-capi aliza ion ule. The main cha ac e is ics o he companies o bo h ea men and con ol g oups a e epo ed in Table 10 in he Appendix. As expec ed, he ea men g oup shows a 13 % poin s la ge a e age deb le el han he con ol g oup. I is likely ha he ea men g oup has a highe absolu e alue o le e age han he con ol g oup. T ea men g oup companies mus ha e co espondingly high in e es expenses o ensu e ha he ax deduc ibili y es ic ion applies, while companies in he con ol g oup will no incu in such high in e es allowances. Addi ionally, ea ed companies ha e a lowe a e age angibili y, a lowe a e age p o i abili y and a lowe ma ke - o-book a io compa ed wi h con ol g oup i ms. Unde he DiD app oach, also known as “in e ac ion among dummy a iables”, we cons uc wo dummy a iables: he i s one is called TREATED which is equal o 1 i he company belongs o he ea men g oup and ze o o he wise, and con ols o pe manen di e ences be ween he ea men and con ol g oups; he second one is called TAX_REFORM which is equal o 1 o iscal yea s 2012 and 2013 (i.e. a e he implemen a ion o he co po a e ax e o m) and 0 o all he p eceding iscal yea s, and i con ols o ends common o bo h ea men and con ol g oups. The esul ing DiD model wi h he inclusion o he wo new dummy a iables is: LEVi =β0+β1·TREATEDi +β2·TAX_REFORMi +β3·TREATEDi ·TAX_REFORMi m  j=1 βj ·TAX VARIABLESji + n  k=1 βk ·CONTROL VARIABLESki +ηi+εi (6) 24 The special ax egimes in hese Au onomous Regions do no apply he same hin-capi aliza ion ule han in he es o he Spanish Au onomous Regions. 123 378 SERIEs (2016) 7:359–391 Whe e he dependen a iable, and ax and con ol a iables ha e been iden i- ied be o e in Eq. (5); ηiis he ixed e ec o company i; and εi is he esidual e m. We ha e added company- ixed e ec s in o de o cap u e all ime-cons an e ec s ha migh be esponsible o a company o be a high- o low-le e aged i m in gene al; he e o e, he TREATED dummy a iable is no iden i ied in he es i- ma ion o Eq. (6) The in e ac ion e m TREATED*TAX_REFORM equals 1 when company i belongs o he ea men g oup and yea is 2012 o 2013; con e sely, i is equal o ze o o he wise. The obse able a iables in Eq. (6) will ha e wo po en ial ou comes co esponding o he ou comes unde he ea men and he con ol g oups. The ea men e ec is he di e ence be ween he wo po en ial ou comes. In pa icula , he β3coe icien gi es he DiD es ima e o he ea - men e ec , namely he change in he deb a io o he ea men g oup and ha o he con ol g oup be o e and a e he ax e o m came in o e ec . We expec his coe icien o be nega i e, which indica es ha he deb a io o he ea ed companies declines compa ed o hei coun e pa s in he con ol g oup a e he e o m. While i is likely o see a sha p change in beha io among he ea men g oup ollowing he exogenous shock ha is he limi a ion o he ax deduc ibili y o inan- cial expenses, i is no expec ed o see such a change in he con ol g oup ha is no a ec ed by ha shock. As Robe s and Whi ed (2013) poin ou , he key assump ion o consis ency o he DiD es ima o is wha is called he ze o co e- la ion assump ion. In economic e ms, his condi ion implies ha , in he absence o he ea men , he e olu ion in he a iable o in e es (in ou case, le e age) would ha e been iden ical o bo h he ea men and con ol g oups. This assump ion is o en e e ed o as he pa allel ends assump ion because i equi es any ends in he dependen a iable o he wo g oups o companies p io o he ea men o be he same (Ang is and Pischke 2009). While le el di e ences o ou comes be ween he ea men and con ol g oups as well as common ends o ou comes o bo h ea ed and con ol companies a e easily handled by he DiD es ima o , di e en ial ends among he wo g oups will gene ally lead o inconclusi e o e oneous in e - ences. The same as wi h all endogenei y p oblems, we canno o mally es he pa allel ends assump ion. E en hough simila p e- ea men ends a e eassu ing, o exam- ple pe o ming a mean-di e ence es , hey a e no a su icien condi ion o ensu e ha he endogenei y p oblem has been amended. No wi hs anding, and as we ha e mo e han 1-yea p e- e o m da a, we can compa e he ends in he ou come a iable du ing he ime ame p io o 2012 o bo h ea men g oup companies and hei non- ea ed coun e pa s. Figu e 3illus a es he ime e olu ion o he deb le el scaled by o al asse s o bo h ea men (i.e. companies a ec ed by he ax e o m) and con ol (i.e. companies no a ec ed by he ax e o m) g oups. The mean deb le el o he ea men g oup companies inc eases sligh ly h ough- ou he whole pe iod. O e all, an inc ease o app oxima ely 2.89 % om 29.78 % in 2008 o 32.67 % in 2013 can be obse ed. Con e sely, no a ec ed o con ol g oup companies inc ease hei mean deb le el by a 2.84 % om 19.20 % in 2008 o 22.04 % in 2013. On a close examina ion, he mean le e age eaches a maxi- mum o 47.11 % (23.82 %) o ea ed (non- ea ed) companies in 2012. In e ms o 123 SERIEs (2016) 7:359–391 379 No es: he ea men g oup is de ined as companies a ec ed by he ax e o m, while he con ol g oup includes companies no a ec ed by he e o m. Di e ence is he ea men ou come minus he con ol ou come. 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 2008 2009 2010 2011 2012 2013 T ea men Con ol Di e ence Fig. 3 Deb le el ime e olu ion o ea ed s. con ol companies he di ec ion o he pos - e o m esponse, i is expec ed ha companies in he ea - men g oup educe hei le e age posi ion mo e s ongly han hose coun e pa s in he con ol g oup. As Fig. 3illus a es, he e is a signi ican dec ease in he mean deb le el o 14.44 % o he ea men g oup om 2012 o 2013, while his educ- ion was only o 1.78 % o he con ol g oup. F om a simple desc ip i e pe spec i e, Fig. 3shows ha on a e age companies a ec ed by he 2012 co po a e ax e o m educe hei deb a io a e he e o m o a much la ge ex en han hei no a ec ed coun e pa s. Nex , we o mally es whe he his di e ence is s a is ically signi i- can . Table 3depic s he es ima ion esul s o ou di e ence in di e ence eg ession model coming om Eq. 6.AsAn (2012), Kahle and S ulz (2013), and D eßle and Scheue ing (2015), we employ a ixed e ec s DiD app oach as ixed e ec s eg es- sion con ols o companies’ unobse ed and ime-in a ian cha ac e is ics ha may in luence he ou come a iable.25 Columns (1) and (2) in Table 3show he es ima es om he DiD eg ession model wi hou and wi h co a ia es, espec i ely. The in e ac ion e m, TREATED*TAX_ REFORM, cap u es o wha ex en companies in he ea men g oup adjus hei le e - age when hey a e a ec ed by he 2012 ax e o m. The coe icien o his in e ac ion 25 Hausman es ejec s he null hypo hesis ega ding he absence o co ela ion be ween he unobse able e ec s and he explana o y a iables, which leads o he selec ion o he ixed-e ec es ima o . 123 380 SERIEs (2016) 7:359–391 Table 3 Reg ession es ima ion esul s unde he DiD app oach Explana o y a iables (1) (2) (3) (4) TAX_REFORM −0.008 (0.021) 0.038 (0.028) −0.047* (0.028) −0.008 (0.036) TREATED*TAX_ REFORM −0.178** (0.069) −0.124** (0.054) 0.055 (0.062) 0.072 (0.069) MTREBIT 2.216** (1.043) 1.661 (1.076) NDTS 1.881 (1.601) 0.703 (2.640) NDTS*RISK −0.398*** (0.113) 0.225** (0.0922) RISK −0.001** (0.000) −0.000** (0.000) ΔTANG 0.231* (0.125) 0.138 (0.148) ΔSIZE 0.268*** (0.076) 0.225*** (0.069) ΔPROF 0.273* (0.178) −0.098 (0.150) ΔTOBINSQ −0.153*** (0.045) −0.076*** (0.025) Obse a ions 528 434 352 258 R-Squa ed Wi hin 0.043 0.214 0.011 0.129 Columns (1) and (2) include ixed-e ec eg ession coe icien s es ima ed om Eq. (6) wi hou and wi h co a ia es, espec i ely. Reg essions in columns (3) and (4) es ic he sample o yea s 2007–2011 and use a ake ax policy in 2011 as a placebo. The dependen a iable is deb le el scaled by o al asse s in all eg essions. TAX_REFORM is a dummy a iable ha akes he alue o 1 o da a a e he ax e o m, and TREATED is a dummy a iable ha akes he alue o 1 i he company is a ec ed by he 2012 co po a e ax e o m. Table 7in he Appendix p o ides de ini ions o he es o he a iables. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els. Robus s anda d e o s a e in b acke s e m has u ned ou o be nega i e and s a is ically signi ican a he 5 % le el in bo h eg essions. This nega i e sign implies ha hose companies a ec ed by he new hin-capi aliza ion ule educe hei indeb edness s onge han hose coun e pa s no a ec ed. Speci ically, and wi hou including co a ia es, he change o he p e- o pos - e o m deb a io is 18 pe cen age poin s lowe o he ea men g oup compa ed o he con ol g oup. A e con olling o o he ac o s, he second column o Table 3 depic s ha he a e age deb a io in he ea men g oup declines mo e by 12 pe - cen age poin s ela i e o he con ol g oup. The e o e, ou Hypo hesis 3 is con i med which can be in e p e ed as he e does seem o exis a ax e o m e ec . As a as he es o es ima es is conce ned, he signs and signi icance o he coe - icien s emain quali a i e and quan i a i ely alike o hose encoun e ed in Table 2. Consequen ly, Hypo heses 1 and 2 a e again con i med. Fu he mo e, we ca y ou a alsi ica ion es in o de o alida e whe he he e was a pa allel end o he dependen a iable be ween ea men and con ol g oups in he p e- e o m pe iod (i.e. be o e iscal yea 2012). In his way, we examine whe he po en- ial ea ed and un ea ed companies had a di e en de elopmen o hei deb a io in a placebo e o m. Speci ically, we es ic ou sample ime ho izon o yea s 2007–2011, and es ablish as a placebo a ake ax policy in 2011. Columns (3) and (4) in Table 3 epo he esul s o his es wi hou and wi h co a ia es, espec i ely. As expec ed, he DiD es ima e o he ea men e ec is nei he s a is ically and no economically signi ican ; in o he wo ds, he di e ence in le e age a ios be ween ea men and 123 SERIEs (2016) 7:359–391 381 Table 4 Es ima ion esul s o he capi al s uc u e model om Eq. (5) wi h lagged ma ginal ax a e (m ) Fixed–e ec eg ession coe icien s es ima ed om Eq. (5) wi h obus s anda d e o s in b acke s. Table 7in he Appendix p o ides de ini ions o he a iables. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els Explana o y a iables LEV1LEV2 LAGGED_MTR 1.275* (0.748) 0.299 (0.681) NDTS 1.654 (2.015) 2.742 (2.121) NDTS*RISK −0.395*** (0.099) −0.125 (0.089) RISK −0.001* (0.000) −0.001* (0.000) ΔTANG 0.319** (0.156) 0.351** (0.143) ΔSIZE 0.259*** (0.080) 0.174** (0.077) ΔPROF 0.423** (0.172) 0.430 (0.273) ΔTOBINSQ −0.155*** (0.047) −0.076*** (0.016) Obse a ions 370 370 R-Squa ed Wi hin 0.203 0.129 con ol g oups is insigni ican o e e y yea conside ed p io o he 2012 ax e o m. This implies ha he analysed companies do no di e in he inancial beha iou and all he companies eac in he same way. This inding cons i u es e idence ha he e is a pa allel end be ween he wo g oups in he yea s be o e he ax e o m. 7 Robus ness o esul s In o de o assess he obus ness o ou p e ious empi ical e idence, we pe o m h ee di e en es s. Fi s ly, in he eg ession app oach ha we ha e applied in he p eceding Sec ion, a p e-in e es income measu e o he ma ginal ax a e was used in o de o con on he endogenei y p oblem. This solu ion e ades he e ec o inancing decisions, and hus alle ia es he non-exogenei y ax s a us o companies. Ne e heless, he e s ill could exis biases in his measu e due o po en ial ea nings manipula ion om manage s in o de o in luence he ma ginal ax a e and hence, he company’s deb policy. Consequen ly, as a obus ness check we ha e ins umen ed he le e age dependen a iable wi h i s lag. Table 4p esen s he esul s o his new es ima ion and co obo a e ou p e ious indings. Secondly, he ela ionship be ween ma ginal ax a es and deb policy could be in lu- enced by he le e age s a us o he companies. Speci ically, he posi i e e ec o axa- ion on le e age should be s onge o less le e ed i ms, which p esumably ha e mo e incen i e o inc ease hei deb . In o de o es his issue we calcula e a dummy a iable (DUMMY_LEV) ha equals 1 i he deb a io is below he median and 0 o he wise. Fu - he mo e, we in oduce in he eg ession MTREBIT*DUMMY_LEV whichisanin e - ac ion e m esul ing om he mul iplica ion o he dummy a iable DUMMY_LEV and he MTREBIT a iable. Table 5con ains he es ima ion esul s and shows ha he coe icien associa ed wi h he in e ac ion e m is only s a is ically signi ican o he dependen a iable LEV1. I s posi i e sign indica es a mo e in ense posi i e e ec o axes on deb o i ms wi h a less le e ed s a us. In he sample o mo e le e ed i ms, he e ec o MTREBIT on deb policy is equal o 2.4, while among less le e ed i ms he impac is 1.1 highe , al hough his di e ence is only ma ginally signi ican . 123 382 SERIEs (2016) 7:359–391 Table 5 Es ima ion esul s o he capi al s uc u e model om Eq. (5) con olling o le e age Fixed-e ec eg ession coe icien s es ima ed om Eq. (5) wi h obus s anda d e o s in b acke s. Table 7in he Appendix p o ides de ini ions o he a iables. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els Explana o y Va iables LEV1LEV2 MTREBIT 2.383*** (0.787) 1.454 (0.964) NDTS 1.785 (1.226) 2.103 (1.493) NDTS*RISK −0.402*** (0.067) −0.138* (0.083) RISK −0.001* (0.000) −0.001** (0.000) ΔTANG 0.167 (0.144) 0.162 (0.147) ΔSIZE 0.281*** (0.071) 0.221*** (0.075) ΔPROF 0.267* (0.140) 0.270 (0.217) ΔTOBINSQ −0.133*** (0.032) −0.066*** (0.018) DUMMY_LEV −0.480*** (0.162) −0.230*** (0.081) MTREBIT*DUMMY_LEV 1.103* (0.627) 0.096 (0.434) Obse a ions 434 434 R-Squa ed Wi hin 0.307 0.187 Table 6 Es ima ion esul s o he capi al s uc u e model om Eq. (5) con olling o s a u o y ax a e Fixed-e ec eg ession coe icien s es ima ed om Eq. (5) wi h obus s anda d e o s in b acke s. Table 7in he Appendix p o ides de ini ions o he a iables. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els Explana o y a iables LEV1LEV2 TAX DIFFERENCE −1.192 (0.834) −0.096 (0.735) NDTS 1.581 (2.037) 2.768 (2.114) NDTS*RISK −0.387** (0.096) −0.120 (0.089) RISK −0.000* (0.000) −0.001* (0.000) ΔTANG 0.319** (0.156) 0.344** (0.141) ΔSIZE 0.257*** (0.079) 0.175*** (0.077) ΔPROF 0.428** (0.172) 0.434 (0.272) ΔTOBINSQ −0.158*** (0.046) −0.078*** (0.164) Obse a ions 370 370 R-Squa ed Wi hin 0.200 0.129 Thi dly, i manage s make decisions based on hei i m’s cu en s a u o y ax s a us, highe deb igu es would be obse ed o companies wi h g ea e di e ences be ween hei s a u o y ax a e and ma ginal ax a e. The e o e, we cons uc a TAX DIFFERENCE a iable as pe G aham (1996a) and Sinha and Bansal (2013), which is he esul o he s a u o y ax a e minus he ma ginal ax a e. Acco ding o his a ionale, we expec his ax di e ence a iable o be posi i ely ela ed o deb usage i companies make ax-based le e age decisions based on hei s a u o y ax a es. Con e sely, he coe icien on his a iable will be ze o o non-signi ican i i ms make ax-based le e age decisions based exclusi ely on simula ed ma ginal ax a es. We ha e used he lagged alues o his a iable, and we ha e con olled o he ma ginal ax a e in he eg ession hough i is d opped due o collinea i y. Table 6 epo s he es ima ion esul s and shows companies make deb decisions based on hei ma ginal ax a e and no he s a u o y ax a e. 123 SERIEs (2016) 7:359–391 389 Table 10 Cha ac e is ics o ea men g oup e sus con ol g oup Va iable T ea men g oup Con ol g oup Mean di e ence es Deb o asse s ma ke alue 0.328 0.200 0.128 ∗∗∗ MTREBIT 0.185 0.181 0.004 NDTS −0.007 0.007 −0.014 NDTS*RISK −1.429 −0.078 −1.351 ∗∗ RISK 7.556 3.062 4.494 TANG 0.064 0.126 −0.062 ∗∗∗ SIZE 20.374 20.553 −0.179 PROF 0.019 0.069 −0.050 ∗∗∗ TOBIN’S Q 1.298 1.734 −0.436 ∗∗∗ This able compa es he means o key a iables be ween companies assigned o he ea men g oup and companies included in he con ol g oup. The o al sample o 88 companies (616 obse a ions) is dis ibu ed be ween 25 companies (175 obse a ions) o he ea men g oup and 63 companies (441 obse a ions) o he con ol g oup. NDTS and NDTS*RISK ha e been escaled by a ac o o 100. Table 7p o ides de ini ions o he a iables. The las column on he igh hand side show he esul o a es i he mean alues be ween he ea men g oup and he con ol g oup a e s a is ically equal. Supe sc ip as e isks indica e s a is ical signi icance a 0.01(***), 0.05(**) and 0.10(*) le els Re e ences Albe e ns S, Su e h C (2015) The e ec o axes on co po a e inancing decisions – E idence om he Ge man in e es ba ie . Wo king Pape , No. 8, Cen e o Tax and Accoun ing Resea ch Al man E (1968) Financial a ios, disc iminan analysis, and he p edic ion o co po a e bank up cy. J Finance 23:589–609 Alwo h J, A achi G (2001) The e ec o axes on co po a e inancing decisions: e idence om a panel o I alian i ms. In Tax Public Finance 8:353–376 An Z (2012) Taxa ion and capi al s uc u e: empi ical e idence om a quasi-expe imen in China. J Co p Finance 18:683–689 Ang is JD, Pischke JS (2009) Mos ly ha mless econome ics: an empi icis ’s companion. P ince on Uni- e si y P ess, P ince on Ba holdy J, Ma eus C (2011) Deb and axes o p i a e i ms. In Re Financ Anal 20:177–189 Blouin J, Co e JE, Guay W (2010) Ha e he ax bene i s o deb been o e es ima ed? J Financ Econ 98:195–213 Blouin J, Huizinga H, Lae en L, Nicodème G (2014) Thin capi aliza ion ules and mul ina ional i m capi al s uc u e. Wo king Pape , WP/14/12, In e na ional Mone a y Fund B adley M, Ja ell G, Kim E (1984) On he exis ence o an op imal capi al s uc u e: heo y and e idence. J Finance 39:857–878 Byoun S (2008) How and when do i ms adjus hei capi al s uc u es owa d a ge s. J Finance 63(6):3069– 3096 Con os G (2005) An essay on he e ec s o axa ion on he co po a e inancial policy. In: 98 h Annual Con e ence on Taxa ion, Na ional Tax Associa ion, Miami, Flo ida, USA De Jong A, Rezaul K, Thuy TN (2008) Capi al s uc u e a ound he wo ld: he oles o i m- and coun y- speci ic de e minan s. J Bank Finance 32:1954–1969 DeAngelo H, Masulis R (1980) Op imal capi al s uc u e unde co po a e and pe sonal axa ion. J Financ Econ 8(1):3–29 Doidge C, Dyck A (2015) Taxes and co po a e policies: e idence om a quasi na u al expe imen . J Finance 70(1):45–89 D eßle D, Scheue ing U (2015) Empi ical e alua ion o in e es ba ie e ec s. Discussion Pape , No. 12-046, ZEW Faccio M, Xu J (2015) Taxes and capi al s uc u e. J Financ Quan Anal 50(3):277–300 123 390 SERIEs (2016) 7:359–391 Fama EF, F ench KR (2012) Capi al s uc u e choices. C i ical Finance. Re iew 1(1):59–101 Feld LP, Heckemeye JH, O e esch M (2013) Capi al s uc u e choice and company axa ion: a me a-s udy. J Bank Finance 37:2850–2866 F ank MZ, Goyal VK (2009) Capi al s uc u e decisions: which ac o s a e eliably impo an ? Financ Manag 38:1–37 Go don RH, Lee Y (2001) Do axes a ec co po a e deb policy? E idence om U.S. co po a e ax e u n da a. J Public Econ 82:195–224 G aham JR (1996a) Deb and he ma ginal ax a e. J Financ Econ 41:41–73 G aham JR (1996b) P oxies o he ma ginal ax a e. J Financ Econ 42:187–221 G aham JR (2000) How big a e he ax bene i s o deb ? J Finance 55(5):1901–1941 G aham JR (2003) Taxes and co po a e inance: a e iew. Re Financ S ud 16:1075–1129 G aham J R (2008) Taxes and co po a e inance. In: Eckbo B E (Ed) Handbook o Co po a e Finance, ol 2, Chap e 11. Empi ical Co po a e Finance. Else ie , No h-Holland, pp 59–133 G aham J R (2013) Do axes a ec co po a e decisions? A e iew. In: Cons an inides G M, Ha is M and S ulz R M (Eds) Handbook o he Economics o Finance, ol 2, Pa A, Chap e 3. Else ie , No h-Holland, pp 123–210 G aham JR, Mills L (2008) Using ax e u n da a o simula e co po a e ma ginal ax a es. J Accoun Econ 46(2–3):366–388 G aham JR, Lemmon ML, Schallheim JS (1998) Deb , leases, axes, and he endogenei y o co po a e ax s a us. J Finance 53(1):131–162 G opp R E (1997) The e ec o expec ed e ec i e co po a e ax a es on inc emen al inancing decisions. Wo king Pape , No 46, In e na ional Mone a y Fund G opp RE (2002) Local axes and capi al s uc u e choice. In Tax Public Finance 9(1):51–71 Ha mann-Wendels T, S ein I, S ö e A (2012) Tax incen i es and capi al s uc u e choice: e idence om Ge many. Discussion Pape , No 18, Deu sche Bundesbank Heide F, Ljungq is A (2015) As ce ain as deb and axes: es ima ing he ax sensi i i y o le e age om S a e ax changes. J Financ Econ 118(3):684–712 Kahle KM, S ulz RM (2013) Access o capi al, in es men and he inancial c isis. J Financ Econ 110:280– 299 Ko JK, Yoon S (2011) Tax bene i s o deb and deb inancing in Ko ea. Asia Pac J Financ S ud 40:824–855 Kunieda S, Takaha a J, Yada H (2011) Japanese i ms’ deb policy and ax policy. Discussion Pape Se ies No 2011-11, Hi o subashi Uni e si y, Tokyo Lemmon ML, Robe s MR, Zende JF (2008) Back o he beginning: pe sis ence and he c oss-sec ion o co po a e capi al s uc u es. J Finance 63(4):1575–1608 López-G acia J, Sogo b-Mi a F (2014) Sensi i i y o ex e nal esou ces o cash low unde inancial con- s ain s. In Bus Re 23:920–930 Mackie-Mason JK (1990) Do axes a ec co po a e inancing decisions? J Finance 45(5):1471–1493 Modigliani F, Mille MH (1963) Co po a e income axes and he cos o capi al: a co ec ion. Am Econ Re 53(2):433–443 Mooij R A (2011) The ax elas ici y o co po a e deb : a syn hesis o size and a ia ions. Wo king Pape , WP/11/95, In e na ional Mone a y Fund Mye s SC (1984) The capi al s uc u e puzzle. J Finance 39:575–592 Panie F, Pé ez-González F, Villanue a P (2013) Capi al s uc u e and axes: wha happens when you (also) subsidize equi y? Unpublished Manusc ip , S an o d Uni e si y Plesko GA (2003) An e alua ion o al e na i e measu es o co po a e ax a es. J Accoun Econ 35:201–226 Robe s M R, Whi ed T M (2013) Endogenei y in empi ical co po a e inance. In: Cons an inides G M, Ha is M and S ulz R M (Eds) Handbook o he Economics o Finance, ol 2, Pa A, Chap e 7. Else ie , No h-Holland, pp 493–572 She lin T (1990) Es ima ing co po a e ma ginal ax a es wi h asymme ic ax ea men o gains and losses. J Am Tax Assoc 12:51–67 Shum PM (1996) Taxes and co po a e deb policy in Canada: an empi ical in es iga ion. Can J Econ 29:556–572 Sinha P, Bansal V (2012) Algo i hm o calcula ing co po a e ma ginal ax a e using Mon e Ca lo simula- ion. MPRA Pape No 40811, Uni e si y Lib a y o Munich, Ge many Sinha P, Bansal V (2013) Capi al s uc u e puzzle: he in e ela ionship be ween le e age, axes and o he mic o economic ac o s. MPRA Pape No 49878, Uni e si y Lib a y o Munich, Ge many 123 SERIEs (2016) 7:359–391 391 Sogo b-Mi a F (2005) How SME uniqueness a ec s capi al s uc u e: e idence om a 1994–1998 Spanish da a panel. Small Bus Econ 25:447–457 S ebulae IA, Whi ed TM (2012) Dynamic models and s uc u al es ima ion in co po a e inance. Found T ends Finance 6:1–163 Ti man S, Wessels R (1988) The de e minan s o capi al s uc u e choice. J Finance 43:1–19 T eze an R (1992) Deb inancing and ax s a us. J Finance 47:1557–1568 Welch I (2011) Two common p oblems in capi al s uc u e esea ch: he inancial-deb - o-asse a io and issuing ac i i y e sus le e age changes. In Re Finance 11(1):1–17 123