Coverage of financing deficit in firms in financial distress under the pecking order theory

104 2016, XIX, 4 Finance DOI: 10.15240/ ul/001/2016-4-008 In oduc ion Capi al s uc u e has been a equen opic in i nancial li e a u e because i is one o he mos impo an decisions a i m can make. Al hough many impo an con ibu ions ha e been made in his a ea, mos o he esea ch does no include i ms in i nancial dis ess, so he i nancing decisions adop ed by hese i ms a e s ill no well known. The i nancing decisions o hose i ms a e e y impo an because mos o he s a egy decisions such as in es men s, ma ke en y, o p oduc di e si i ca ion a e conside ably a ec ed by he i nancial cons ain s aced by hem (Bowe, Fila o che , & Ma shall, 2010). O e he yea s, wo main explana ions o he capi al s uc u e o companies ha e been p oposed (Ba clay & Smi h, 2005; Flanne y & Rangan, 2006; F ank & Goyal, 2009; Mu adoğlu & Si ap asad, 2012). The i s one is he s a ic ade-o heo y, which p oposes a ade-o be ween he ax ad an ages o deb i nancing and he cos s o i nancial dis ess. Too much deb can lead o i nancial dis ess and oo li le deb can gi e ise o low e u ns on equi y. The e o e, companies selec he capi al s uc u e ha maximizes hei alue, which leads o an op imal deb le el. The second one is he pecking o de heo y, which pos ula es he exis ence o a hie a chy o i nancial esou ces, so i ms do no a ge op imum capi al s uc u es. When ou side unds a e necessa y, i ms can mainly eso o h ee sou ces: e ained ea nings, deb , and equi y. Whe eas e ained ea nings ha e no ad e se selec ion p oblem, bo h equi y and deb ha e an ad e se selec ion isk p emium because o in o ma ion asymme ies be ween manage s and in es o s. In es o s demand highe e u ns on equi y han on deb . The e o e, i companies do no ha e enough e ained ea nings o i nance hei in es men p ojec , hey will p e e deb o equi y. Al hough hese wo heo ies ha e been es ed using di e en me hodologies, he e idence is con o e sial as he empi ical esul s end o suppo he p edic ions o bo h heo ies. Some s udies highligh he impo ance o he pecking o de heo y and o he s show he ele ance o he ade-o heo y. In his ega d, Shyam-Sunde and Mye s (1999) i nd s ong suppo o he pecking o de heo y when hey analyze he ela ionship be ween ne deb issued and i nancing de i ci . Fama and F ench (2002) and Lea y and Robe s (2005) show ha i ms’ deb a ios adjus slowly o ela i ely in equen ly owa d hei a ge , which is mo e consis en wi h he pecking o de heo y. Agca and Mozumda (2004) and Lemmon and Zende (2010) p opose a conca e ela ionship be ween ne deb issued and i nancing de i ci , which enables a less s ic i nancial hie a chy o he pecking o de heo y. On he o he hand, se e al au ho s i nd e idence consis en wi h he ade-o heo y (Co ei, Fa ha , & Abug i, 2011; Flanne y & Rangan, 2006; F ank & Goyal, 2009). Besides, some s udies end o bea ou bo h heo ies. F ank & Goyal (2003) only i nd suppo o he pecking o de heo y among la ge i ms, and Lea y and Robe s (2005) show ha bo h heo ies help explain some aspec s o i nancing decisions. Finally, a lo o ecen s udies ocus hei a en ion on i ms wi h di e en cha ac e is ics, such as small, la ge amily con olled o di e si i ed i ms (González & González, 2012; La Roccaa, La Roccaa, Ge aceb, & Sma k, 2009; Pindado & De la To e, 2008; Sel a ajah & U sel, 2012). Mos p e ious s udies ha e analyzed he capi al s uc u e o heal hy i ms. Howe e , he esul s o hese s udies a e no di ec ly applicable o i ms in i nancial dis ess, mainly because hese i ms ha e o e in es men and unde in es men p oblems, less i nancial sou ces a ailable and a e a ec ed by bank up cy COVERAGE OF FINANCING DEFICIT IN FIRMS IN FINANCIAL DISTRESS UNDER THE PECKING ORDER THEORY Se gio San i lippo-Azo a, Ca los López-Gu ié ez, Begoña To e-Olmo EM_4_2016.indd 104EM_4_2016.indd 104 30.11.2016 16:31:4430.11.2016 16:31:44 105 4, XIX, 2016 Finance laws (Da ydenko & F anks, 2008; López, To e, & San i lippo, 2012; Gian & S ahan, 2007). The li le e idence abou i ms in i nancial dis ess is con o e sial, because he s udies do no i nd suppo o he ade-o heo y, bu hey do no p o ide conclusi e esul s abou he pecking o de heo y ei he . Fo example, Gilson (1997) i nds ha he high ansac ion cos s bo ne by i ms in i nancial dis ess p e en hem om adjus ing hei capi al s uc u e o op imum le els. In his ega d, Pindado, Rod igues and De la To e (2006), when analyzing a sample o small and medium-sized Po uguese i ms, i nd ha he i nancing decisions o i ms in dis ess do no depend on hei p e ious deb le els o on he exis ence o a ge deb a ios, and he e o e do no suppo he ade-o heo y. Liang and Ba hala (2009) pe o m a s udy on a small sample o i ms in i nancial dis ess in he Uni ed S a es, bu hei esul s a e no e y conclusi e. They i nd ha he i ms’ i nancing decisions did no seek an op imum deb a io. Howe e , hey also i nd li le suppo o he pecking o de heo y, as hei esul s show a weak ela ionship be ween i nancing de i ci and deb . The ade-o heo y p oposes ha i ms pu sue an op imal deb le el by weighing he bene i s o deb (especially deb - ela ed ax shields) and he cos s o deb (bank up cy p oblems). Howe e , many i ms canno quickly adjus hei deb in esponse o changes in hei a ge deb because hey bea ansac ion cos s. Fi ms in i nancial dis ess ha e a lo o ouble eaching hei op imal capi al s uc u e p oposed by he ade-o heo y because hey ha e high ansac ion cos s (Asqui h, Ge ne , & Scha s ein, 1994; Chou, Li, & Yin, 2010). To educe hei deb s, i ms in i nancial dis ess mus nego ia e new paymen e ms wi h c edi o s o sell asse s ha implies complica ed adjus men s. To his ega d, Gilson (1997) i nds ha dis essed i ms ha dly e e manage o educe hei deb le el in o de o each hei op imal capi al s uc u e, so hei deb a ios con inue o be high. Ano he a gumen agains he ade-o heo y in i ms in i nancial dis ess is ha hese i ms canno o en ake ad an age o he deb - ela ed ax shields. Financial dis essed i ms o en incu losses, so hey can seldom bene i om he ax deduc ibili y o in e es (Ba clay & Smi h, 2005). The e o e, hese i ms i nd i qui e ha d o s ike a balance be ween he ad an ages and disad an ages o deb i nancing. The pecking o de heo y pos ula es he exis ence o a s ic hie a chy o i nancial esou ces because o in o ma ion asymme ies be ween manage s and in es o s (Mye s & Majlu , 1984; Shyam-Sunde & Mye s, 1999). Fi ms would s a using in e nal unds, hen deb , and i nally equi y. Howe e , he imposi ion o his s ic hie a chy migh no necessa ily be applicable in i ms in i nancial dis ess o wo easons: Fi s , Shyam-Sunde and Mye s (1999) sugges ha hese i ms could co e hei i nancing de i ci by issuing equi y o selling asse s o a oid inc easing hei deb a io and/o deb es uc u ing. Mo eo e , equi y migh be he only secu i y ha ou side i nancie s o in es o s a e willing o buy; second, Chi inko and Singha (2000) show how a hie a chy o deb and equi y is no necessa ily ollowed s ic ly when i ms ace a es ic ion on hei deb capaci y, a common si ua ion o i ms expe iencing di i cul ies. All in all, i ms in i nancial dis ess equen ly ha e o use all o hei a ailable i nancial esou ces o co e hei i nancing de i ci and ha e mo e and mo e di i cul ies o issue deb . This implies ha i ms in i nancial dis ess inc easingly u n o sou ces o unds o he han deb issues as hei i nancing de i ci g ows. The e o e, he ela ionship be ween ne deb issued and i nancing de i ci es ablished by he pecking o de heo y cease o be linea and become conca e quad a ic. This quad a ic ela ionship migh well explain he con o e sy abou he capi al s uc u e o i ms in i nancial dis ess. The main con ibu ion o his s udy is o es a po en ial conca e quad a ic ela ionship be ween ne deb issued and i nancing de i ci in i ms in i nancial dis ess, which has no been s udied p e iously. I his quad a ic ela ionship exis s, he i nancing decisions o i ms in i nancial dis ess will be di e en o m he i nancing decisions o heal hy i ms, so he o me s will no ollow he s ic i nancial hie a chy p oposed by he pecking o de heo y due o hei speci i c si ua ion. Ano he impo an con ibu ion o his s udy is ha , di e en om p e ious esea ch pape s, we also analyze he p obabili y o issuing equi y. I i ms in dis ess do no ollow he s ic hie a chy o he pecking o de heo y, a g ea p obabili y o issuing equi y can be expec ed han in heal hy i ms. In his s udy we include heal hy and dis essed i ms, so we a e able o es he quad a ic ela ionship in bo h se s o i ms and EM_4_2016.indd 105EM_4_2016.indd 105 30.11.2016 16:31:4430.11.2016 16:31:44 106 2016, XIX, 4 Finance compa e hei di e en i nancing beha io . Also, he me hodology used allows us o o e come some limi a ions o p e ious s udies. In he i s analysis, he Sys em GMM me hodology o panel da a is used, which enables con olling o he model’s indi idual he e ogenei y and he exis ence o po en ial p oblems o endogenei y. Subsequen ly, in he s udy o he p obabili y o issuing equi y, we use, o he i s ime in his kind o s udies, a new He e ogeneous Choice Models (HCM) me hodology de eloped by Williams (2009) applied o a logis ic unc ion. This me hodology allows us o a oid he bias caused by he di e ences in he deg ee o esidual a ia ion be ween heal hy i ms and i ms in i nancial dis ess. P e ious s udies do no conside hose di e ences, so hei esul s could be biased. The analysis is pe o med on a sample o 3,337 lis ed i ms om Ge many, Canada, he Uni ed S a es, F ance, I aly and he Uni ed Kingdom om 1995 o 2006. The inclusion o hese coun ies co e s a b oad spec um o ins i u ional en i onmen s. The sample pe iod ends in 2006 o a oid he biases o he i nancial c isis. The esul s indica e a quad a ic ela ionship be ween i nancing de i ci and ne deb issued o i ms in i nancial dis ess. This ela ionship is conca e, so ha as he i nancing de i ci inc eases, he ne deb issuance p opo ion dec eases. Howe e , he i nancing decisions o heal hy i ms ollow a linea ela ionship a he han a quad a ic one. Finally, he second analysis shows ha i ms in i nancial dis ess ha e a g ea e p obabili y o issuing equi y, which suppo s ou esul s ega ding he exis ence o a conca e quad a ic ela ionship. Thus, equi y i nancing could be an al e na i e o deb issuance as a sou ce o unds o i ms in i nancial dis ess. The s uc u e o he s udy is as ollows: The sample used is desc ibed in Sec ion 1. Sec ion 2 p esen s he model and main esul s in ela ion o he exis ence o a quad a ic ela ionship. I also desc ibes he analysis o he p obabili y o issuing equi y and displays he esul s. We i nish wi h he conclusions and he e e ences. 1. Sample and Da a To es he exis ence o he quad a ic ela ionship, we use a sample o non- i nancial i ms lis ed on he s ock exchange in Ge many, Canada, he Uni ed S a es, F ance, I aly and he Uni ed Kingdom. The inclusion o hese coun ies allows co e ing companies ope a ing unde di e en ins i u ional en i onmen s wi h a b oad spec um o bank up cy sys ems. This p e en s ha hese ci cums ances condi ion he analysis by con olling o he coun y. Fo each coun y, we ha e an unbalanced panel o i ms wi h in o ma ion a ailable o a minimum o se en consecu i e yea s be ween 1995 and 2006. To calcula e he second-o de se ial co ela ion es , undamen al o gua an eeing he obus ness o he es ima ions made ia he Sys em GMM me hodology, da a o each company o a leas ou consecu i e yea s is equi ed. In addi ion, o calcula e ce ain a iables in ou analysis, we equi ed a iables lagged h ee yea s. We es ic he sample pe iod o end in 2006 so ha ou esul s a e no a ec ed by he i nancial c isis. A e he onse o he i nancial c isis, he i ms’ i nancing beha io could be condi ioned mo e by he a ailabili y o unds in he economy and he dis up ion o he i nancial sys ems han by he i ms’ si ua ion, which could ha e gi en ise o a bias in ou esul s. The economic- i nancial in o ma ion o each i m is om he Da aS eam da abase, o he Thomson Financial Se ices g oup. The mac oeconomic in o ma ion is ob ained om he Wo ld Bank’s Wo ld De elopmen Indica o s da abase and OECD s a is ics. Tab. 1 shows he empo al and coun y dis ibu ion o he i ms o he six coun ies included in he analysis. By including only lis ed companies, he numbe o i ms aded on each o he secu i ies exchanges condi ions he size by coun y. Howe e , he able shows ha he sample size, o all yea s and coun ies analyzed, is adequa e o pe o ming he analysis. Since he i nancial dis ess si ua ion is no di ec ly obse able, we employ wo di e en p oxy measu es o dis inguish he i ms in i nancial dis ess. Fi s , we use he Z-Sco e model (Al man, 1968). The Z-Sco e model is: Z = 1.2*X1 + 1.4*X2 + 3.3*X3 + + 0.6*X4+1*X5 (1) whe e X1 is he wo king capi al o o al asse s a io; X2 is he e ained ea nings o o al asse s a io; X3 is he ea nings be o e in e es and axes o o al asse s a io; X4 is he ma ke alue equi y o book alue o o al liabili ies a io; X5 is sales o o al asse s a io. EM_4_2016.indd 106EM_4_2016.indd 106 30.11.2016 16:31:4530.11.2016 16:31:45 107 4, XIX, 2016 Finance The alue o Z-sco e has he ollowing in e als. Values highe han 2.99 a e conside ed he “sa e zone”, and i means ha he possibili y o company’s bank up cy is e y low. Values be ween 1.81 and 2.99 a e conside ed he “g ey zone” o “zone o igno ance”, because o he suscep ibili y o e o classi i ca ion. Values below 1.81 a e conside ed “dis ess zone”, and i means ha he possibili y o a company’s bank up cy is high. So, we iden i y i ms in i nancial dis ess when hey a e si ua ed in he “dis ess zone”, when hey ha e in a pa icula yea a Z-sco e less han 1.81. Second, we use he O-Sco e o classi y i ms in i nancial dis ess (Ohlson, 1980). The O-Sco e is based on Ohlson’s p edic ed bank up cy p obabili ies p, ollowing his speci i ca ion: (2) yi = – 1.32 – 0.407 * SIZE + 6.03 * * TLTA – 1.43 * WCTA + 0.757 * * CLCA – 2.37 * NITA – 1.83 * FUTL + 0.285 * INTWO – 1.72 * OENEG – 0.521CHIN, (3) whe e SIZE is he log o o al asse s o GNP P ice-le el index a io; TLTA is he o al liabili ies o o al asse s a io; WCTA is he wo king capi al o o al asse s a io; CLCA is Tempo a y dis ibu ion o he sample Yea Canada F ance Ge many I aly Uni ed kingdom USA To al 1995 76 119 161 47 311 789 1,503 1996 89 129 176 54 324 953 1,725 1997 97 133 183 59 344 1,056 1,872 1998 108 142 191 65 364 1,164 2,034 1999 153 147 202 70 443 1,353 2,368 2000 170 143 197 76 475 1,541 2,602 2001 186 234 252 113 506 1,585 2,876 2002 208 250 258 129 534 1,611 2,990 2003 199 245 246 129 519 1,519 2,857 2004 191 241 237 128 505 1,463 2,765 2005 187 223 234 123 489 1,424 2,680 2006 169 189 215 112 453 1,231 2,369 To al 1,833 2,195 2,552 1,105 5,267 15,689 28,641 Obse a ions pe coun y Coun y Obse a ions To al Dis essed Z-sco e Dis essed O-sco e Canada 1,833 516 122 F ance 2,195 436 149 Ge many 2,552 568 246 I aly 1,105 429 80 Uni ed kingdom 5,267 643 425 USA 15,689 2,343 1,128 To al 28,641 4,935 2,150 Sou ce: own Tab. 1: Sample desc ip ion EM_4_2016.indd 107EM_4_2016.indd 107 30.11.2016 16:31:4530.11.2016 16:31:45 108 2016, XIX, 4 Finance he cu en liabili ies o cu en asse s a io; NITA is he ne income o o al asse s a io; FUTL is he unds om ope a ions o o al liabili ies a io; INTWO is equal o one in ne income is nega i e in he p e ious 2 yea s o ze o o he wise; OENEG is equal o one i o al liabili ies a e g ea e han o al asse s o ze o o he wise; whe e NI is he ne income o yea . We iden i y i ms in i nancial dis ess when he bank up cy p obabili y is g ea e han o equal o 50%. These wo models ha e been widely used o iden i y i ms in i nancial di i cul ies in bo h Ame ican and in e na ional s udies (Diche , 1998; G i i n & Lemmon, 2002; Bhaga , Moyen, & Suh, 2005; Geo ge & Hwang, 2010; Lopez e al., 2012). On a e age, i ms in i nancial dis ess ep esen 17% o he obse a ions when we use he Z-Sco e model and 13% when we use he O-Sco e model. 2. Empi ical Analysis 2.1 Me hodology Shyam-Sunde and Mye s (1999) p esen a es o he pecking o de heo y based on i nancing de i ci unde he p emise ha his de i ci is co e ed en i ely by issuing new deb . Thus hey p opose he ollowing ela ionship: (4) whe e ΔDi is he amoun o ne deb issued o wi hd awn; DEFi is he i nancing de i ci ; ei is he andom e o e m. Acco ding o Shyam-Sunde and Mye s (1999), a simple e sion o he pecking o de heo y p edic s α = 0 and βPO = 1. This me hod o assessing he Pecking o de heo y has been widely c i icized (F ank & Goyal, 2003; Lea y & Robe s, 2010). In he case o i ms in i nancial dis ess, and acco ding o Chi inko and Singha (2000), his model does no admi he possibili y o simul aneously issuing deb and equi y as we p opose in his a icle. In his ega d, Liang and Ba hala (2009) i nd ha he βPO coe i cien was posi i e and signi i can , bu conside ably lowe han 1 in i ms in i nancial dis ess. This esul migh well e l ec ha i ms in i nancial dis ess co e hei i nancing de i ci s no only wi h deb bu wi h equi y. In o de o sol e his p oblem, we will in oduce he i nancing de i ci squa e (DEF2) in o he equa ion (4): (5) The use o a quad a ic e m allows es ing no only he simul aneous issue o deb and equi y, bu also he exis ence o a hie a chy di e en o ha p oposed by he pecking o de heo y (Lopez e al., 2012). I i ms in dis ess depend less on deb s (and mo e on equi y), o use deb issues dec easingly o co e hei i nancing de i ci , when his i nancing de i ci inc eases, he βPO coe i cien in hese i ms would be posi i e, bu conside ably lowe han 1, and he ϒ coe i cien would be nega i e. Thus he pe cen age o deb issued o i nance he i nancing de i ci would dec ease as his i nancing de i ci inc eased. As ou s udy simul aneously analyzed heal hy i ms and hose in i nancial di i cul y, we modi i ed equa ion (5) so ha he model o be es ima ed would be as ollows: (6) whe e ΔDi is he ne deb issued o o al asse s (F ank & Goyal, 2003; Lemmon & Zende , 2010; Liang & Ba hala, 2009; Shyam-Sunde & Mye s, 1999); DEFi is he i nancing de i ci di ided by o al asse s. This a iable includes di idend paymen s, ne in es men and changes in wo king capi al, and is educed by ope a ing cash l ows a e in e es s and axes; DIF is a dummy a iable ha akes alue 1 o i ms in dis ess and 0 o heal hy i ms. Fo his, as we showed ea lie , we ollowed wo al e na i e app oaches: he Al man Z-Sco e (DIFZ) and he Ohlson O-Sco e (DIFO). εi ep esen s he andom e o e m. We also included dummy a iables o coun y, yea and sec o . The β1 and ϒ1 coe i cien s, espec i ely, show he linea and quad a ic e ec s o heal hy i ms. The (β1 + β2) coe i cien s show he linea e ec o i ms in i nancial dis ess, and he (ϒ1 + ϒ2) coe i cien s show he quad a ic ela ionship o i ms in dis ess. To es he signi i cance o he (β1 + β2) and (ϒ1 + ϒ2) coe i cien s, i is necessa y o pe o m a join signi i cance es unde he null hypo heses H0: β1 + β2 = 0 and H0: ϒ1 + ϒ2 = 0. I i ms in i nancial dis ess dec easingly used deb o co e hei i nancing de i ci as his i nancing de i ci ises, (β1 + β2) can be expec ed o be posi i e and signi i can , EM_4_2016.indd 108EM_4_2016.indd 108 30.11.2016 16:31:4530.11.2016 16:31:45 109 4, XIX, 2016 Finance bu conside ably less han 1, and (ϒ1 + ϒ2) can be nega i e and signi i can . To es he obus ness o he analysis, we in oduced he a iables used by F ank and Goyal (2003) in o he model (6), as p oposed by Agca and Mozumda (2004). This con ols o o he ac o s (apa om i nancing de i ci ) whose ele ance has been demons a ed in p e ious s udies o i m i nancing decisions. The esul ing model would be as ollows: (7) whe e T e e s o he angibili y o asse s o o al asse s; MTB is he ma ke - o-book a io; LS is he na u al loga i hm o sales and P is he e u n on asse s. Tab. 2 p esen s summa y s a is ics o he sample. We es ima ed he models (6) and (7) using he gene alized me hod o momen s (Sys em GMM). This me hod allows con olling o po en ial p oblems o endogenei y h ough he use o ins umen s, by including he lagged igh -hand side a iables. 2.2 Resul s Tab. 3 shows he esul s o he analyses. In model (a), he Al man Z-Sco e was used o iden i y he i ms in i nancial dis ess, while in model (b) he Ohlson O-Sco e was used. In models (a) and (b), he quad a ic e m o he DEF a iable was in oduced. In he case o heal hy i ms, he DEF a iable has a posi i e and signi i can coe i cien , bu conside ably less han 1. The e o e, he pecking o de heo y Mean S anda d de ia ion Minimum Maximum ΔD 0.0137 0.1108 -0.9713 0.6989 DEF -0.0351 0.1761 -0.9877 2.4510 DIV 0.0138 0.0215 0.0000 0.2970 I 0.0691 0.0994 -0.9310 0.8892 ΔWK 0.0078 0.1287 -1.8135 1.3271 CA 0.1258 0.1646 -1.7118 1.1355 T 0.3226 0.2318 0.0002 0.9968 MTB 1.7125 1.2493 0.1845 19.1626 LS 12.6820 2.0845 2.8865 19.2228 P 0.0604 0.1237 -0.9779 0.6554 ΔT 0.0180 0.0957 -0.9566 0.8456 ΔMTB -0.0347 0.9701 -16.4099 14.4661 ΔLS 0.0904 0.3254 -4.8684 5.9904 ΔP -0.0009 0.0805 -0.9131 0.9371 LIQ 0.4898 0.2313 0.0032 0.9982 NDTS -0.0013 0.0337 -0.5281 0.6804 DEBT 0.5313 0.1983 0.0054 0.9984 LOGSIZE 5.5391 0.8878 2.8948 8.7543 Sou ce: own No e: ΔD is he ne deb issued o o al asse s; DEF is he i nancing de i ci di ided by o al asse s; DIV is di idend pay- men s o o al asse s; I is he ne in es men o o al asse s; CA e e s o cash l ow o o al asse s; ΔWK is he change in wo king capi al o o al asse s; T e e s o he angibili y o asse s o o al asse s; MTB is he ma ke - o-book a io; LS is he na u al loga i hm o sales; P is he e u n on asse s; LIQ e e s o he cu en asse s o o al asse s; NDTS is non-deb ax shields o o al asse s; DEBT is he le e age a io; LOGSIZE is he loga i hm o o al asse s. Tab. 2: Sample s a is ics EM_4_2016.indd 109EM_4_2016.indd 109 30.11.2016 16:31:4530.11.2016 16:31:45 110 2016, XIX, 4 Finance does no appea o ha e a highe explana o y powe in he i nancing decisions adop ed by he heal hy i ms in ou sample. The DEF2 a iable is no signi i can and he e o e a quad a ic model would no be sui able o hese i ms. Rega ding i ms in i nancial dis ess, model (a) shows ha he join signi i cance es (β1 + β2), unde he null hypo hesis H0: β1 + β2 = 0, is posi i e and signi i can . The join signi i cance es o he quad a ic componen (ϒ1 + ϒ2), unde he null hypo hesis H0: ϒ1 + ϒ2, is nega i e and signi i can . The e o e, unlike in heal hy i ms, a conca e quad a ic ela ionship can be obse ed in i ms in i nancial dis ess. So, i ms in i nancial dis ess would use deb dec easingly as hei i nancing de i ci inc eases. In ac , he linea coe i cien is conside ably less 1, so he i ms in i nancial dis ess in ou sample did no s ic ly ollow he pecking o de heo y. This coe i cien , as we p oposed, indica es ha di e en i nancial esou ces a e used o co e he i nancing de i ci . Model (b) shows he same (a) (b) (c) (d) DEF 0.1662 ** (2.48) 0.1641 ** (2.32) 0.0953 * (1.83) 0.0613 * (1.71) DEF*DIFZ 0.2713 ** (2.36) 0.2995 *** (4.46) DEF *DIFO 0.2353 *** (2.73) 0.2142 *** (3.63) DEF2 -0.0927 (-0.67) -0.0413 (-0.28) -0.0876 (-0.96) 0.0324 (0.49) (DEF*DIFZ)2 -0.1174 (-0.95) -0.1820 * (-1.79) (DEF *DIFO)2 -0.2470 (-1.47) -0.2517 *** (-2.66) ΔT 0.4591 *** (6.24) 0.5002 *** (8.94) ΔMTB 0.0011 (0.14) -0.0053 (-1.37) ΔLS 0.0225 (1.18) 0.0053 (0.6) ΔP -0.2113 *** (-3.6) -0.0837 * (-1.81) CONS 0.0540 (1.48) -0.0057 (-0.25) -0.0083 (-0.30) -0.0024 (-0.31) (β1 + β2) 0.4375 *** 0.3994 *** 0.3949 *** 0.2755 *** (ϒ1 + ϒ2) -0.2101 *** -0.2883 *** -0.2696 *** -0.2192 *** m20.951 0.452 0.171 0.12 HANSEN 74.26 (0.179) 104.57 (0.194) 151.4 (0.131) 263.47 (0.212) Sou ce: own No e: Coe i cien s associa ed wi h each a iable. In b acke s, T-s uden ; *** indica es a le el o signi i cance o 0.01, ** indica es a le el o signi i cance o 0.05, * indica es a le el o signi i cance o 0.1; m2 is he 2nd o de se ial co ela ion s a is ic. Hansen is he o e -iden i ying es ic ion es (p- alue in b acke s). (β1 + β2) is he es s o join signi i cance unde he null hypo heses H0: β1 + β2 = 0. (ϒ1 + ϒ2) is he es s o join signi i cance unde he null hypo heses H0: ϒ1 + ϒ2 = 0. Time-dummy a iables, coun y dummy a iables and sec o dummy a iables a e also included in he es ima ions al- hough he esul s a e no shown in he ables o ocus on he main esul s ob ained. Tab. 3: Resul s (DIF a iable non-lagged) EM_4_2016.indd 110EM_4_2016.indd 110 30.11.2016 16:31:4530.11.2016 16:31:45 111 4, XIX, 2016 Finance esul , whe e he Ohlson O-Sco e was used o iden i y i ms in dis ess. The es o he join signi i cance (β1 + β2) is posi i e and signi i can and (ϒ1 + ϒ2) is nega i e and signi i can . In models (c) and (d), he F ank and Goyal a iables (F ank & Goyal, 2003) we e in oduced. In his case, he p e ious esul s we e main ained since i ms in i nancial dis ess show a conca e quad a ic ela ionship, whils in heal hy i ms only a linea e ec is obse ed. Fo he con ol a iables in oduced, ΔT coe i cien is posi i e and signi i can , showing he use ulness o angible asse s as colla e al o suppo g ea e le el o le e age. As p e ious s udies ha e shown g ea e p o i abili y has a nega i e e ec on le e age (F ank & Goyal, 2003; Mackay & Go don, 2005). To check he obus ness o ou esul s, he p e ious models we e es ima ed again, in oducing he DIF a iable lagged one yea . We used his o con i m he e ec o i nancing de i ci on ne deb issued one yea a e a i m expe iences i nancial dis ess. The esul s, no shown in his pape , a e simila o hose ob ained in Tab. 3. 2.3 Analysis o Equi y Financing The esul s o he p e ious analysis show a conca e quad a ic ela ionship be ween ne deb issued and i nancing de i ci o i ms in i nancial dis ess due o he ac ha his de i ci is co e ed by using di e en i nancial esou ces. As we explained ea lie , he main sou ce a ailable o hese i ms migh well be equi y i nancing. I i ms in dis ess do no ollow he hie a chy o he pecking o de heo y, a g ea p obabili y o issuing equi y can be expec ed han in heal hy i ms. To es his idea, we p opose a disc e e choice analysis based on a logis ic model in which he dependen a iable akes alue 1 i he i m issues equi y and alue 0 o he wise. Howe e , he inclusion o wo g oups o i ms (heal hy and dis essed) makes i e y p obable ha he homoscedas ici y o andom e o s will no be ul i lled because o he exis ence o di e ences in he deg ee o esidual a ia ion be ween bo h g oups o i ms. Unlike linea models, in non-linea models his ac gi es ise o signi i can biases in he es ima ion o he model pa ame e s (Ya chew & G iliches, 1985). To o e come his p oblem in he cu en s udy, we pe o med an analysis using he He e ogeneous Choice Models (HCM) applied o a logis ic unc ion. These models con ol o he di e ences in he andom e o a iance be ween he g oups, which allows a oiding he biases in he es ima ions (Williams, 2009). The p oposed model is as ollows: (8) whe e Λ (.) ep esen s a logis ic p ocess; he dependen a iable “y” akes alue 1 i he e is a ne inc ease in ex e nal equi y o a leas 5% o o al asse s, o he wise alue 0 (Ho akimian, Ople , & Ti man, 2001; Lea y & Robe s, 2010; Vanacke & Maniga , 2010); zi is a ec o o a iables used o de e mine he e o a iances linked o ce ain ϒ pa ame e s. To selec he con ol a iables included in xi, p e ious s udies on i nancing and equi y issuance was ollowed (De Haan & Hinloopen, 2003; De Jong & Veld, 2001); DIF is a dummy a iable ha akes alue 1 o i ms in i nancial dis ess and 0 o heal hy i ms (as in p e ious analyses, he Al man Z-Sco e and Ohlson O-Sco e we e used); P is he e u n on asse s; LIQ e e s o he cu en asse s o o al asse s; DIV is di idend paymen s o o al asse s; NDTS is non-deb ax shields o o al asse s (Pindado e al., 2006); DEBT is he le e age a io; LOGSIZE is he loga i hm o o al asse s; PE is a dummy a iable ha akes alue 1 i he i m has used equi y i nancing du ing he p e ious yea and 0 o he wise. We also included dummy a iable o coun y, yea and sec o . Summa y s a is ics o he a iables is showed in Tab. 2. Tab. 4 shows he esul s o he analysis. Models (a) and (b), which include he a iables wi hou lags show ha he e is a g ea e p obabili y o equi y i nancing in i ms in i nancial dis ess as he ma ginal e ec s o he DIFZ and DIFO a iables a e posi i e and signi i can . These esul s suppo he exis ence o a conca e ela ionship be ween he ne deb issued and i nancing de i ci ob ained in he p e ious analysis. The absence o a s ic i nancial hie a chy implies he simul aneous use o di e en sou ces o i nancing. This analysis demons a es ha equi y i nancing could be an al e na i e o deb issuance as a sou ce o unds o i ms in i nancial dis ess as his would allow hem o a oid excessi e deb a ios o deb es uc u ing. EM_4_2016.indd 111EM_4_2016.indd 111 30.11.2016 16:31:4630.11.2016 16:31:46 112 2016, XIX, 4 Finance Wi h ega d o he con ol a iables, he esul s a e simila o hose o p e ious s udies (De Haan & Hinloopen, 2003; Vanacke & Maniga , 2010). P o i abili y, liquidi y, he le e age a io and p io equi y i nancing ha e a nega i e and signi i can e ec on he p obabili y o issuing equi y, whils i m size and non-deb ax shields ha e a posi i e in l uence on his p obabili y. Like p e ious s udies, di idends do no a ec he p obabili y o equi y i nancing. To check he obus ness o ou esul s, he p e ious models we e es ima ed again, including all he a iables lagged one pe iod o a oid possible endogenei y p oblems (De Haan & Hinloopen, 2003). The esul s, no shown in his pape , a e e y simila o hose ob ained in Tab. 4, as i ms in i nancial dis ess con inue showing a g ea e p obabili y o issuing equi y. Conclusions This s udy ocused on analyzing he i nancial decisions o i ms in i nancial dis ess. A s ic hie a chy o i nancing sou ces does no appea o be applicable in hese i ms. The s udy analyses he exis ence o a conca e quad a ic ela ionship be ween i nancing de i ci and ne deb issued, which p o ides addi ional e idence o p e ious esea ch on he capi al s uc u e o i ms expe iencing i nancial di i cul ies. The analysis was pe o med using a sample o 3,337 non- i nancial i ms lis ed on he s ock exchanges in Ge many, Canada, he Uni ed S a es, F ance, I aly and he Uni ed Kingdom du ing he pe iod be ween 1995 and 2006. The es ima es we e based on Sys em GMM me hodology o panel da a, which makes i possible o con ol o endogenei y p oblems, and on HCM models applied o a logis ic unc ion, which con ol o he exis ence o di e ences in he deg ee o esidual a ia ion be ween heal hy and dis essed i ms. We ound e idence ha nei he he ade- o no he s ic hie a chy sugges ed by he pecking o de heo y would be applicable in i ms in i nancial dis ess. Ou esul s show ha as i nancing de i ci g ows, hese i ms use deb dec easingly and ha e a g ea e p obabili y o issuing equi y. This leads o a conca e quad a ic ela ionship be ween i nancing de i ci (a) (b) DIFZ0.0405 (8.25) *** DIFO 0.0201 (3.29) *** P -0.0483 (-7.09) *** -0.0597 (-7.66) *** LIQ -0.0009 (-0.21) -0.0139 (-3.12) *** DIV -0.0306 (-0.61) -0.0567 (-1.09) NDTS 0.0734 (3.14) *** 0.0854 (3.61) *** DEBT -0.0511 (-9.43) *** -0.0381 (-7.15) *** LOGSIZE -0.0049 (-3.93) *** -0.0042 (-3.23) *** PE 0.0981 (13.78) *** 0.1017 (14.07) *** Pseudo R20.2128 0.2181 Sou ce: own No e: Ma ginal e ec s (inc emen al e ec s o dummy a iables) associa ed wi h each a iable. In b acke s, T-s a is ic; *** indica es signi i cance a he 1% le el, ** indica es signi i cance a he 5% le el, * indica es signi i cance a he 10% le el. Time-dummy a iables, coun y dummy a iables and sec o dummy a iables a e also included in he es ima ions al hough he esul s a e no shown in he ables o ocus on he main esul s ob ained. Tab. 4: Ma ginal e ec s – HCM – logis ic models EM_4_2016.indd 112EM_4_2016.indd 112 30.11.2016 16:31:4630.11.2016 16:31:46