Fear the walking dead: the epidemic of zombie firms in Portugal
Abstract
This study explores firms called “zombies” in the Portuguese economy. These firms are defined as old firms and have persistent problems in paying interest on bank loans and sidetrack labor productivity from healthy companies. The theme of zombie firms has been analyzed from the case of Japan in the 1990s to the most recent on OECD countries (i.e. Portugal). Several authors have studied this theme, the most well-known being Caballero et al. (2008). First the idea of Caballero et. al (2008) being applied, then the simplified formula of McGowan et al. (2018) was used to compute the regression. The present study shows that, between 2010 and 2018, in the Portuguese economy we had zombie firms following a less restricted and a more restricted analysis. In the first, on average, during the years of analysis we have about 12.8% while in the second this number decreases to 7.4%. In the less restrictive analysis, firms need to be, at least, 10 years old to remove bias using newest firms (for example start-ups) and the interest coverage rate is less than 1 per cent. for at least 3 consecutive years. In the more restrictive, it has been implemented to previous financial ratios such as: Return on Assets and the total debt ratio. It was also confirmed by using Ordinary Least Squares, Fixed Effects and Random Effects models, as well as a Probit model, that zombie companies tend to be related to specific sectors of the economy, certain areas of the country and their size (number of employees). These zombie companies have significant implications for healthy companies operating in the same industry, reducing employment and profit margin.
Full text
i
João A onso Pe ei a Pi a
Fea he walking dead: he epidemic o zombie i ms in Po ugal
Mas e s in Economics
Supe iso s: P o esso Dou o Fe nando Manuel Almeida Alexand e
P o esso Dou o Miguel Ângelo Reis Po ela
Oc obe 2019
ii
DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS
Es e é um abalho académico que pode se u ilizado po e cei os desde que espei adas as eg as e
boas p á icas in e nacionalmen e acei es, no que conce ne aos di ei os de au o e di ei os conexos.
Assim, o p esen e abalho pode se u ilizado nos e mos p e is os na licença abaixo indicada.
Caso o u ilizado necessi e de pe missão pa a pode aze um uso do abalho em condições não
p e is as no licenciamen o indicado, de e á con ac a o au o , a a és do Reposi ó iUM da Uni e sidade
do Minho.
Licença concedida aos u ilizado es des e abalho
A ibuição-NãoCome cial-SemDe i ações
CC BY-NC-ND
h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0/
iii
Acknowledgmen s
The accomplishmen o his mas e 's hesis had impo an suppo and incen i es wi hou which i would
no be possible and o which I will be o e e g a e ul.
Fi s ly, I wan o hank he eache s Fe nando Alexand e and Miguel Po ela o hei guidance and o al
a ailabili y, as well as o he incen i es, opinions and c i icisms in sol ing doub s and p oblems ha
came om he ealiza ion o his wo k.
I also lea e he e a wo d o hanks o all he eache s o he School o Economics and Managemen o
he knowledge ansmi ed and o all he wo ds o encou agemen .
To my colleagues and iends who ha e always suppo ed me in accomplishing his wo k wi h wo ds and
wisdom needed o o e come he p oblems ha a ose.
Finally, a special hanks o my pa en s, sis e and o he membe s o my amily, especially my uncles
who ha e always been p esen and ga e uncondi ional suppo , o he inancial and emo ional suppo ,
encou agemen and pa ience in helping wi h all he di icul ies ha a ose. To hem I dedica e his wo k.
Thank you so much,
João A onso Pe ei a Pi a
i
STATEMENT OF INTEGRITY
I he eby decla e ha ing conduc ed his academic wo k wi h in eg i y. I con i m ha I ha e no used
plagia ism o any o m o undue use o in o ma ion o alsi ica ion o esul s along he p ocess leading o
i s elabo a ion.
I u he decla e ha I ha e ully acknowledged he Code o E hical Conduc o he Uni e si y o Minho.
Medo dos
zombies
: a epidemia de emp esas
zombies
em Po ugal
Resumo
Es e es udo explo a as emp esas denominadas “zombies” na economia po uguesa. Es as emp esas
são de inidas sendo emp esas an igas e que em p oblemas pe sis en es em paga os ju os de
emp és imos bancá ios e des iam a p odu i idade do abalho de emp esas saudá eis.
O ema das emp esas
zombies
já oi analisado desde o caso do Japão na década de 90 a é aos mais
ecen es sob e países da OCDE (nomeadamen e Po ugal). Vá ios au o es es uda am es e ema sendo o
mais conhecido Caballe o e al. (2008). P imei o oi usado a ideia de Caballe o e al. (2008) sendo
aplicada a ó mula simpli icada de McGowan e al. (2018).
O p esen e es udo mos a que, en e 2010 e 2018, na economia po uguesa emos emp esas
zombies
seguindo uma análise menos es i a e ou a mais es i a. Na p imei a, em média, du an e os anos de
analise emos ce ca de 12.8% enquan o que na segunda es e núme o diminui pa a 7.4%.
Na análise menos es i a oi somen e omado em conside ação o ac o de a emp esa e , pelo menos,
10 anos de idade pa a emo e da análise as no as emp esas (
s a -ups
, po exemplo) e a axa de
cobe u a dos ju os se in e io a 1 po , pelo menos, 3 anos consecu i os. Na mais es i a, oi
implemen ado a an e io es ácios inancei os ais como: o e o no sob e o A i o e o ácio o al da di ida.
Foi ambém con i mado eco endo aos modelos Mínimos Quad ados O diná ios, E ei os ixos e E ei os
alea ó ios e, ainda, um modelo P obi , que as emp esas
zombies
endem a es a elacionados com
se o es especí icos da economia, com ce as zonas do país e o seu amanho (núme o de emp egados).
Es as emp esas
zombies
em implicações signi ica i as nas emp esas saudá eis a ope a no mesmo
se o , eduzindo o emp ego e a ma gem de luc o.
Pala as cha e: capi al, emp ego, emp esas
zombie
, endi idamen o, inanciamen o, ácios inancei os
Classi icação JEL: E22; E24; G32; G33; O16.
i
Fea he walking dead: he epidemic o zombie i ms in Po ugal
Abs ac
This s udy explo es i ms called “zombies” in he Po uguese economy. These i ms a e de ined as old
i ms and ha e pe sis en p oblems in paying in e es on bank loans and side ack labo p oduc i i y
om heal hy companies.
The heme o zombie i ms has been analyzed om he case o Japan in he 1990s o he mos ecen
on OECD coun ies (i.e. Po ugal). Se e al au ho s ha e s udied his heme, he mos well-known being
Caballe o e al. (2008). Fi s he idea o Caballe o e . al (2008) being applied, hen he simpli ied
o mula o McGowan e al. (2018) was used o compu e he eg ession.
The p esen s udy shows ha , be ween 2010 and 2018, in he Po uguese economy we had zombie
i ms ollowing a less es ic ed and a mo e es ic ed analysis. In he i s , on a e age, du ing he yea s
o analysis we ha e abou 12.8% while in he second his numbe dec eases o 7.4%.
In he less es ic i e analysis, i ms need o be, a leas , 10 yea s old o emo e bias using newes i ms
( o example s a -ups) and he in e es co e age a e is less han 1 pe cen . o a leas 3 consecu i e
yea s. In he mo e es ic i e, i has been implemen ed o p e ious inancial a ios such as: Re u n on
Asse s and he o al deb a io.
I was also con i med by using O dina y Leas Squa es, Fixed E ec s and Random E ec s models, as
well as a P obi model, ha zombie companies end o be ela ed o speci ic sec o s o he economy,
ce ain a eas o he coun y and hei size (numbe o employees). These zombie companies ha e
signi ican implica ions o heal hy companies ope a ing in he same indus y, educing employmen and
p o i ma gin.
Keywo ds: capi al, employmen , inancial a ios, inancing, indeb ness, zombie i ms
JEL Classi ica ion: E22; E24; G32; G33; O16.
ii
Index
Acknowledgmen s ..................................................................................................................... iii
Resumo ........................................................................................................................................
Abs ac ....................................................................................................................................... i
Lis o Tables .............................................................................................................................. ix
Lis o Figu es .............................................................................................................................. x
Lis o abb e ia ions and ac onyms ............................................................................................ xi
1.In oduc ion .............................................................................................................................. 1
2. Mac oeconomic Si ua ion o Po ugal ..................................................................................... 3
3. Li e a u e Re iew .................................................................................................................... 7
4. Theo e ical amewo k and hypo heses ................................................................................. 10
5. Da a ........................................................................................................................................ 13
6 – Me hods ............................................................................................................................... 15
6.1. – O dina y Leas Squa es ............................................................................................... 16
6.2. – Fixed E ec s Model .................................................................................................... 17
6.3. – Random E ec s Model ................................................................................................ 17
6.4. – Hausman Tes .............................................................................................................. 18
6.5. - P obi Model ................................................................................................................. 18
7. Empi ical indings ................................................................................................................. 19
7.1. P o ile o Zombie i ms .................................................................................................. 19
7.2. Reg essions used ............................................................................................................ 20
7.3. Less es ic i e analysis .................................................................................................. 21
7.4. Mo e es ic i e analysis ................................................................................................. 25
7.5. Discussion o esul s ....................................................................................................... 29
8. Conclusion ............................................................................................................................. 31
iii
Re e ences ................................................................................................................................. 33
Appendix ................................................................................................................................... 36
ix
Lis o Tables
Table 1 - Numbe o zombies in Po ugal (less es ic i e) ................................................................... 21
Table 2 - Zombie i ms and non-zombies’ pe o mance ....................................................................... 23
Table 3 - Numbe o zombies in Po ugal ( es ic i e) .......................................................................... 26
Table 4 - Zombies i ms and non-zombies’ pe o mance ..................................................................... 26
Table 5 - Fi ms in he Po uguese economy om 2010 un il 2018 ...................................................... 36
Table 6 –
NUTS
2 dis ibu ion in he Po uguese economy ................................................................. 36
Table 7 –
NUTS
3 dis ibu ion in he Po uguese economy ................................................................. 36
Table 8 - NACE code and desc ip ion ................................................................................................. 37
Table 9 - NACE dis ibu ion in he Po uguese economy ...................................................................... 38
Table 10 - Numbe o employees in each i m .................................................................................... 38
Table 11 - P obi model ..................................................................................................................... 39
Table 12 - Ma gins e ec s .................................................................................................................. 40
5
Figu e 3 - Bi hs, Dea hs and Su i o s i ms in Po ugal
No es: The blue line is i ms’ bi hs; he o ange line is i ms’ dea hs; he g ey line is i ms ha su i ed 1 yea a e hey en e
he ma ke ; and he yellow line is i ms ha su i ed 2 yea s a e hey en e he ma ke .
Sou ce: po da a.p 2018, Bank o Po ugal, PORDATA
In he hi d igu e we can see ha , in e ms o i ms’ bi hs, Po ugal has had, a e he 2008
c isis, he numbe o bi hs dec ease a ound 60.000 un il 2012, in which hese bi hs inc eased again.
This can be a sign o he lowe in e es a es and he ise o loans as shown be o e. The i ms’ dea hs
e lec wha is e i ied in he bi hs o companies, in he sense ha , i he numbe o bi hs inc eases,
he numbe o dea hs inc eases. Howe e , be ween 2008 and 2012 he e was a g ea e numbe o
dea hs han bi hs. Cu en ly he numbe s a e e y close. Also, in his igu e, we can obse e he i ms
ha su i ed a leas 1 and 2 yea s a e en e ing he ma ke . Unsu p isingly, i ms ha en e he ma ke
end no o su i e o long, and i is cu ious ha abou hal o he i ms c ea ed in he p e ious wo
yea s a e no in he ma ke in he ollowing yea s. This is an impo an poin ela ed o zombie i ms
because since zombie i ms may, some imes, exi he ma ke , he numbe o i ms’ dea hs should be
some imes highe han i is shown.
E en hough is no shown, conclusions abou insol en and in eco e ing i ms can be gi en.
Acco ding o a s udy om a c edi and isk managemen consul an (“In oT us ”), Po ugal had in 2018
almos wice he closing o i ms’ compa a i e o he p e ious yea , ha ing he insol ency and eco e ing
i ms a dec ease. This may be a sign ha i ms a e su i ing hanks o subsidized bank c edi .
0
30 000
60 000
90 000
120 000
150 000
180 000
210 000
240 000
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Bi hs Deads Su i o s o 1 yea s Su i o s o 2 yea s
6
Figu e 4 - Numbe o employees: To al and by i m size
No es: The blue line is o al i ms in Po ugal; he o ange line is SME; and he g ey line is big i ms.
Sou ce: po da a.p 2018, Bank o Po ugal, PORDATA
Finally, his las igu e shows us ha , in Po ugal, he numbe o employees in o al i ms in
Po ugal dec eased since 2008 un il 2013, aising again o numbe close o he 2008 pe iod. Be ween
SME and big i ms, we can see ha SME had he bigges impac on he dec ease o numbe o
employees and big i ms main ain he same along he pe iod o analysis. Also, i is in e es ing o see
ha SME i ms domina e he ma ke in ela ion o big i ms, which is a cha ac e is ic known o he
Po uguese economy. As said be o e and acco ding o he li e a u e e ision, SMEs had a big impac on
he numbe o zombies ha ming he labo ma ke , so his is also a eason o he Po uguese economy
ha ing so much zombie i ms in ela ion o o he coun ies in he Eu opean Union.
0
500 000
1 000 000
1 500 000
2 000 000
2 500 000
3 000 000
3 500 000
4 000 000
4 500 000
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Dimension To al Dimension PME Dimension La ge
7
3. Li e a u e Re iew
The impac o zombie i ms is a big deal o he economy o a coun y, so he s udy and he
comp ehension o he way ha heal hy companies e ol ed in o zombies is an impo an opic o s udy in
Po ugal, since we ha e a la ge pe cen age o zombies, and i is con ained in he OECD esea ch
p og am.
The phenomenon o zombie i ms began in he ea ly 90’s wi h he collapse o he Japanese
asse p ice bubble, while he e was a pe iod o s agna ion o Japanese i ms known as he “los
decade”. These i ms dis o ed he ma ke and caused signi ican in e up ions in he economy’s
eco e y. Cu iously, ew we e he i ms ha decla ed bank up cy, and many o hem eco e ed a he
beginning o he 21s cen u y. Acco ding o Fukuda and Nakamu a (2001), educing he employee
s eng h and selling ixed asse s we e bene icial o e i ing zombie i ms. Also, ex e nal suppo s
including deb elie and capi al educ ion we e he o he impo an ac o s o he eco e y o hese
i ms.
The s udy om Hoshi and Kashyap (2004) was one o he i s o ale o his ype o i ms,
saying ha he size o bad loans in Japan was a ound 7% o GDP. Tha said, Caballe o e al. (2008), one
o he mos impo an pape ega ding zombie i ms, ied o iden i y hese i ms based on whe he hey
a e ecei ing subsidized c edi , no by looking a hei p oduc i i y o p o i abili y. Also, hey s uggle o
ob ain he da a necessa y o s udy his heme mainly because banks and he bo owe s a en’ a ailable
o p o ide he aw da a ha allows he hem o de e mine i hey a e o no a zombie i m. Tha said, he
s udy om Caballe o e al. (2008) concluded ha i banks le zombie i ms ali e, hey hinde e icien
human esou ce alloca ion and channel he in es men om iable i ms o un iable ones. When hese
un iable i ms go h ough a low pe o mance phase, banks end o p o ide mo e c edi o hem mos ly
when hey wo k oge he o ming a ype o alliance smoo hing (in Japan we can see he kei e su
alliance) he access o c edi .
Lam, Schipke, Tan and Tan (2017) is a s udy accomplished by obse ing i ms in China,
iden i ying he zombies, and explains he cen al ole ha hese i ms ha e wi h s a e-owned en e p ises
aligned o low p oduc i i y and highe deb . The go e nmen had ied a ious e o ms o educe he
8
dele e aging bu wi h no success. The empi ical esul s o his pape s sugges ha , o accele a e he
es uc u ing p ocess, i equi es a mo e holis ic and coo dina ed s a egy, ecognizing losses, educing
he implici suppo and liquida ing zombie i ms.
Recen s udies ha e s udied zombie i ms in he con ex o he Eu opean Union, such as
McGowan e al. (2018), de ining hese i ms as hose ha ha e, a leas , en yea s old, excluding s a -
ups, and hose ha ca no co e he in e es paymen s o , a leas , h ee consecu i e yea s. Also, his
pape es ic s he sample o he p e-c isis pe iod (2003-2007) inding ha an inc ease in zombie sha e
a he indus y le el is associa ed wi h lowe in es men and employmen g ow h o he a e age non-
zombie. So, he esou ces apped in zombie i ms was a policy issue al eady be o e he c isis ha is,
oday, no sol ed.
In he case o Po ugal, hose s udies old us ha , on a e age, zombies a e olde , la ge (bo h in
e ms o numbe o employees and u no e ) and a e much less p oduc i e han hei non-zombie
coun e pa s (Gou eia and Coelho, 2018). McGowan e al. (2018) also shows he sha e o capi al sunk
in zombie i ms in 2013, and in e ms o in e na ional compa ison, Po ugal has a ound 14%, much
mo e han F ance (a ound 6%) o he Uni ed Kingdom (a ound 7,5%). These numbe s a e dec easing,
and i is good o he economy as shown in Os e hold and Gou eia (2018), in which a educ ion in he
capi al sunk in zombie i ms e ches posi i e ex e nali ies o non-zombies.
P e ious pape s s udied he ac o o indeb edness o some Po uguese i ms om 1990 o
1995 and ound ou ha he g ow h o a i m has a posi i e ela ion wi h he indeb edness, which means
ha he i ms wi h highe asse g ow h a es a e he ones ha a e mo e indeb ed (Jo ge and A mada
2001).
S o z, Koe e and Se ze (2017) s udied, be ween he yea s o 2010 and 2014, (pos c isis and
when Po ugal had he (EFAP), he impac o bank s ess on he dele e aging p ocess o non- inancial
small and medium-sized en e p ises denomina ed SMEs, ocusing hei wo k in he eu o a ea wi h wo
g oups: he i s one includes Po ugal, G eece, I eland, Spain and Slo enia called by he au ho “ he
pe iphe y eu o a ea”, and he second denomina ed “eu o a ea co e” wi h F ance and Ge many. They
concluded ha he in e ac ion be ween he week SMEs and weak banks is a possible cause o
mis ep esen a ion in he dele e aging e o s o eu o a ea pe iphe y economies. Fo he same yea ,
9
F ance and Ge many had no signi ican impac o bank s ess on zombie i m le e age sugges ing ha
non- inancing co po a ions dele e aging is indeed associa ed wi h weak banks, possibly because hese
banks ha e an incen i e o keep hese i ms ali e (e e g een loans) o a oid ha ing o decla e non-
pe o ming loans.
Ano he ecen esea ch conduc ed by OECD, he main inancie o he zombie i ms a e weake
banks, sugges ing ha he zombie i m p oblem may pa ly s em om bank o bea ance (McGowan e
al. 2017). Also, Oli ei a (2008) said ha “ he p incipal o m o ex e nal inancing is h ough he use o
he banking sys em (…)”, bu some au ho s concluded ha he owne s hemsel es also in es in hei
business. This ype o bank o bea ance happens when banks seek o a oid ecogni ion o he loss o
c edi g an ed o suck companies (Alexand e e al., 2017).
The economic and inancial analysis o a i m is a oll ha allows, acco ding o Ca doso Mo ei a
(2001), o know he si ua ion o a company h ough i s pas and an icipa e u u e si ua ions, so policy
make s should be awa e o his phenomenon because i can condi ion he coun y’s economic
de elopmen , and some economis such as Lang ield and Pagano (2015), concluded ha an inc ease in
he size o he banking sys em is associa ed o mo e isk and less economic g ow h.
10
4. Theo e ical amewo k and hypo heses
As men ioned be o e, he li e a u e cha ac e izes a zombie i m as hose wi h mo e han a
decade o exis ence, ha do no gene a e enough e enues in hei egula business, pe sis en ly
dependen on bank c edi and hose who usually pay high sala ies conside ing he p oduc i i y o he
sec o in which hey wo k.
Fi s ly, o de ine a zombie i m, i was i s ly ied o use he ini ial idea in he Caballe o e al.
(2008) whe e hey de ine zombie i ms as hose po en ially ecei ing subsidized bank c edi and no by
looking a hei p oduc i i y o p o i abili y so ha we can e alua e he e ec o zombies on he economy.
To do his hey obse ed in e es paymen s made by he i m compa ing hem o an es ima ed
benchma k R* based on he i m deb s uc u e and ma ke in e es a e. Those i ms wi h nega i e
in e es a es a e ecei ing subsidized c edi and a e zombies. Bu he s udy om Caballe o e al. (2008)
is e y da a demanding wi h a da ase ha co e s he pe iod om 1981 o 2002, eaching, some yea s,
o 2500 i ms. Also, we canno p ecisely dis inguish be ween di e en o ms o deb held by companies
in ou da abase, such as bank loans and deb secu i ies issued. The e o e, we also canno obse e
ac ual in e es paymen s on di e en o ms o deb . Obse able o e all in e es expenses may no
necessa ily show he ac ual paymen s du ing a ce ain yea .
Knowing he p e iously p esen di icul y, McGowan e al. (2018) de ine zombie i ms based on
a simpli ied o mula om Caballe o e al. (2008) adding, o he de ini ion o he in e es co e age a io, a
c i e ion o he age o he i m. In Amadeus, we ha e he da e o inco po a ion o each i m so doing
basic compu a ions we go only i ms ha ha e only en yea s o mo e in he ma ke . Tha said, a
zombie mus ha e an in e es co e age a io less han one o h ee consecu i e yea s. Ini ially, he le el
o EBIT is obse ed and di ided by he in e es paid by i ms. This will gi e us he in e es co e age a io
and i i is lowe han one o h ee consecu i e yea s we can say, in p ac ice, his means ha a i m
mus ake on addi ional deb o co e i s in e es paymen s.
I is ypical o iden i y zombies based on a single weak pe o mance me ic, bu his easily
includes i ms who a e g owing and whose weak p o i abili y is o en empo a y. Fo example,
in es men in ol es a necessa y ade-o be ween sho e m cos s and u u e p oduc i i y and
p o i abili y g ow h, which may cause pe o mance me ics o supposedly decline be o e e enue
11
ca ches up. S a -up companies a e pa icula ly ulne able o his. So, hey used he age o he i ms o
di e en ia e a eal zombie om an inno a i e s a -up who may s ill ha e compa a i ely high ope a ing
cos s and low e enue bu since his da abase doesn’ ha e how long i ms had been on he ma ke , we
can’ di e en ia e zombies’ i ms om s a -ups.
A e ha , he empi ical amewo k uses pooled c oss-sec ion mic o da a o explo e he
dis o iona y e ec s o zombie i ms on he pe o mance o non-zombie i ms. Tha said, on he
eg essions non-zombies should ha e also en yea s bu he in e es co e age a io highe han one o
h ee consecu i e yea s and he ollowing inancial a ios.
Following Schi a di e al. (2017) a c i e ion o o al deb o o al asse s a io, also known as,
o al deb a io will be implemen ed and i a i m has a alue lowe han 40% i would be conside ed a
non-zombie (being es ed wi h o he a ios as well o see he impac in he numbe o zombies).
Thi dly, i will be used he Re u n o Asse s (ROA) o analyze he p o i abili y (o “quali y”) o he
i m o eplace Tobin’s q, as men ioned on Miyajima and Ya eh (2007). Tobin’s q is used o exp ess he
ela ionship be ween ma ke alua ion and in insic alue, in o he wo ds, i is a means o es ima ing
whe he a gi en i m o ma ke is o e alued o unde alued. The use o ROA is simila o Tobin’s q
because i ells in es o s an idea o how e ec i e he i m is in con e ing he money i in es s in o ne
income, so he highe he ROA, he be e , because mo e money is made wi h less in es men .
Summing up, a i m is conside ed a non-zombie, whene e hey ha e he in e es co e age a io
highe han one o , a leas , h ee consecu i e yea s; ha e mo e han en yea s in business, he o al
deb a io is lowe han 40% and he ROA is highe han 0. Ou non-zombie dummy is hus equal o 0,
whene e he i m ul ills c i e ia all hese c i e ia o he cu en pe iod, and 1 o he wise.
Following he speci ica ion in Caballe o e al. (2008) and McGowan e al. (2018), we es
whe he zombies en ail nega i e spillo e e ec s on iable i ms. We depend on panel da a om 2010 o
2018 o es ima e a educed- o m equa ion o see he impac o zombie conges ion as:
𝑌𝑖𝑠𝑡 =𝛽0+𝛽1𝑛𝑜𝑛𝑧𝑜𝑚𝑏𝑖𝑒𝑖𝑠𝑡 +𝛽2𝑛𝑜𝑛𝑧𝑜𝑚𝑏𝑖𝑒𝑖𝑠𝑡 ∗𝑍𝑠𝑡 +𝛽3𝑟𝑒𝑣𝑒𝑛𝑢𝑒𝑖𝑠𝑡 +𝛽4𝑠𝑖𝑧𝑒𝑖𝑠𝑡 +
𝜀𝑖𝑠𝑡 (1)
12
Whe e Y deno es he p o i ma gin and employmen g ow h o i m
i
, in indus y
s
, a yea
.
The
dummy nonzombie akes he alue o 1 o non-zombie i ms and 0 o he wise. This a iable is de ined
as shown be o e whene e he i m ul ills all ou c i e ia. Z is he sha e o indus y capi al sunk in
zombie i ms. Wi h he a iable o 𝛽3 i is possible o see he impac o sales in he eg ession. The
a iable β4 is a i m con ol a iable, so i is a dummy a iable ha de ines he size o he i m (1 o 10,
11 o 19, 20 o 49, 50 o 99, 100 o 249 and 250+) in e ms o employmen .
I is expec ed ha 𝛽2will be nega i e, implying ha mo e esou ces a e sunken in zombie i ms,
o he p o i ma gin and employmen g ow h since zombie i ms educe he abili y and capi al o
nonzombie i ms o g ow. As well, 𝛽1 may be nega i e i zombie i ms ecei e la ge amoun s o
subsidized c edi bu , i ’s shown in he li e a u e ha i could be posi i e due o zombie i ms no being
able o spend as much as heal hy i ms. Fo he dummy a iable o size, a aise in he numbe o
employees he highe is expec ed o be he numbe o zombies. Ope a ing Re enue may also in luence
he capi al and employmen g ow h and should ell us i he i m is doing good in e m o sales o i hey
jus ecei e he subsidized c edi .
13
5. Da a
In he li e a u e (see Caballe o e al 2008; McGowan e al. 2018; Alexand e e al. 2017), i has
been obse ed ha he su i al o zombie i ms may dis o compe i ion and weaken ma ke e iciency.
Heal hy ma ke s a e cha ac e ized by a p ocess o c ea i e des uc ion, whe e insol en o unp o i able
i ms educe hei sha e o labo and success ul i ms in es and c ea e new jobs. When zombies
pa icipa e in he ma ke , hey aise demand o labo and in ensi y compe i ion o ma ke sha e. This
has he consequence o lowe ing p oduc p ices and inc easing wages, e ec i ely conges ing g ow h
condi ions o mo e p omising i ms.
To de ine a zombie i ms, se e al c i e ia a e used, om leas es ic ed o he mos es ic ed.
As said be o e, his s udy will p ima ily use he Caballe o e al. (2008) o mula and he simples o m
used by McGowan e al. (2018) wi h he addi ion o o he s inancial a iables, using a ailable
in o ma ion o de e mine which i ms a e ecei ing subsidized c edi using i m-le el da a om Bu eau
an Dijk’s (B D) Amadeus da abase. This is a da abase o compa able inancial in o ma ion o public
and p i a e companies among coun ies in Eu ope. The pu pose o using his da abase is ha i
encompasses da a om Po uguese companies wi h signi ican ele ance o he business and om
inancial a ios ha a e ou inely used in he inancial analysis o companies. Also, his da abase is in
panel da a and has he pe iod which will be analyzed (2010-2018). I would be be e i we could use
he
Sis ema de Con as In eg adas das Emp esas
(SCIE) om he
Ins i u o Nacional de Es a ís ica
(INE)
bu due o some es ic ions i was no possible o use.
The only p oblem in his pe iod is ha i coincides in a subs an ial pa wi h he Economic and
Financial Assis ance P og am (EFAP) ha occu ed in Po ugal be ween 2011 and 2014, a e he
wo ldwide c isis o 2008. Also, no analyzing he pe iod be o e 2008 is a big sh inkage because we
don’ obse e wha happened in a good economic pe iod. Thus, in addi ion o a sho pe iod o ime (8
yea s o analysis), i is an a ypical pe iod, in which economic ac i i y is ex ao dina ily e ac ed.
Ne e heless, as said be o e, i would be in e es ing o analyze he impac o he EFAP in he con ex o
zombie i ms, bu a longe pe iod o da ase would be necessa y.
Ano he impo an poin o his da abase is ha i analyzes bi hs and dea hs o companies as
well as demog aphic indica o s. Fo he cu en yea , he popula ion is cons i u ed by all he companies
14
ha ca y ou an ac i i y o p oduc ion o goods and/o se ices, in Po ugal. In addi ion, he da abase
has in o ma ion o be able o pe o m a dimensional analysis o each i m, along wi h a sec o ial analysis
in which i ms will be di ided by se en egions acco ding o
NUTS
2 and wen y- i e egions acco ding o
NUTS
3.
To s udy he dimension o he i m, his s udy will ha e i e size classes ha a e: i ms wi h less
han en employe s, i ms wi h less han i y employe s bu o e han en, i ms wi h less han one
hund ed bu o e i y employe s, i ms wi h lowe han wo hund ed and i y employe s bu o e han
one hund ed and big en e p ises wi h o e wo hund ed and i y employe s. This size o he i ms was
made acco dingly o Alexand e e al. (2017) and acco ding o a a iable a ailable in Amadeus (I should
be no ed ha his a iable is acco ding o he Eu opean Union classi ica ion based on o al asse s and
ope a ing e enue). Fo he sec o ial analysis i will be di ided by le e s acco ding o he sec o in
ques ion, i.e. each le e will ep esen a sec o o ac i i y wi h a o al o se en een sec o s.
Using he da a a ailable in Amadeus, a se o 378,887 i ms was ob ained o Po ugal bu a e
some es ic ions and adjus men s he numbe was educed o 245,015 mainly due o missing alues.
To be e explain his dec ease in he numbe o obse a ions, i ’s compu ed in S a a ha , i impo an
inancial a iables a e missing, such as, o al asse s, capi al, p o i o EBIT, he i ms a e emo ed. This
command, o example, o he yea o 2010 dec eases he numbe o i ms om he 378,887 o
194,237. Ano he poin is, i he i m has mo e han 7 a iables missing ( he maximum is 11) i ge s
emo ed (in his poin , o he yea o 2010 we ha e 191,151 i ms). O he es ic ion was i he i m
has yea s missing, his means i he i m ha e epo s o 2010 and 2012 bu misses he yea o 2011,
i will ge 1 (equals o 1 yea missing) un il he maximum ha is 9. A his s age we can no ice mo e
i ms d opped in he ea lie da a such as 2018, 2017 and 2016. ( o example, in 2017, p e iously we
had 314,198 i ms and now only 259,654 i ms). This can be explained because, in Po ugal, i ms end
o epo he yea n in he n+1 yea o , in some cases, hey send w ong da a and he egula o s ask o
co ec his. Then i ms ha do no ha e ecen yea da a (2018) a e dele ed. Finally, we ha e ou
da abase eady o he analysis wi h an inc ease in he numbe o i ms since 2010, om 153,795 o
219,749 in 2018. This can cause a p oblem in he analysis since i was no possible o e i y he eal
eason o his dec ease (see Table 5 on appendix).
21
𝐿𝑛(𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡)𝑖𝑠𝑡 =β0+β1nonzombieis +β2nonzombieis ∗Zs +β3 e enueis +
β4size+𝜂𝑖+εis (8)
I is expec ed ha 𝛽2 will be nega i e since zombie conges ion educe he abili y o incen i es
o non-zombie i ms o g ow. The coe icien o he non-zombie 𝛽1 will be ha d o unde s and because,
as said be o e, i can be ei he nega i e o some cases o posi i e, depending on he subsidized c edi
ha i ms ecei e.
7.3. Less es ic i e analysis
As said be o e, a less es ic i e c i e ia will be used o see how many zombies we ha e in he
Po uguese economy. In his pa , i was only used wo c i e ia: he i ms’ age (highe han 10 yea s)
and he in e es co e age a io (less han one o 3 consecu i e yea s).
Using he S a a o compu e hese c i e ia, we go , o Po ugal, an a e age o zombie i ms,
be ween 2010 and 2018, o 12.8% (see Table 1). Also, we can obse e a dec ease since 2010, om
13% o 10.8% in 2018. Also, om 2011 un il 2013, he numbe o zombies aises in 2.7 pe cen ual
poin s. This can be explained mainly because o wo easons. Ei he hese i ms exi ed he ma ke o
had a es uc u ing. Acco ding o he li e a u e, when hese es ic ions we e used, he alue o zombie
i ms was a ound 10%. As we can see, he numbe o zombie i ms in Po ugal is sligh ly highe han he
esul s gi en by he li e a u e. A o his is c ucial o deeply analyze he zombies wi h some mo e c i e ia
and o see whe he and whe e zombies ha e mo e impac on he Po uguese economy.
Table 1 - Numbe o zombies in Po ugal (less es ic i e)
Yea
Zombies
2010
13.0%
2011
12.9%
2012
15.0%
2013
15.6%
2014
14.1%
2015
12.6%
2016
11.4%
22
2017
11.1%
2018
10.8%
Sou ce: Au ho ’s own calcula ions.
In he eg essions (see Table 2) we can no ice a di e ence in he numbe o obse a ions, om
899,937 o he p o i ma gin a iable and 929,326 o he employmen g ow h. This was a cons an
p oblem wi h his da abase because, o example, he p o i ma gin a iable has a lo o missing alues
in Amadeus. Tha said, obse ing he R squa ed, which is 21.7% o he p o i ma gin, 21.7% o he
a ia ion can be explained by he independen a iables in he model and o he Log o he employmen
we ha e a highe R squa ed in which his eg ession can explain 65.6% o he employmen g ow h.
Fo he p o i ma gin eg ession, he p- alue o almos all a iables is, app oxima ely, 0 so since
i is a low p- alue (below 0.05 o 95% con idence le el) his indica es ha we can ejec he null
hypo hesis. In o he wo ds, a change in he independen a iable is associa ed wi h changes in he
esponse a he popula ion le el and all a iables a e s a is ically signi ican . Only he size when he
numbe o employees is be ween 50 and 100, we ha e 0.165 and o he signi icance le el o 0.05 i is
no s a is ically signi ican . Fo he employmen g ow h, we only ha e one p- alue di e en om 0 bu i
is below 0.05, so we ejec he null hypo hesis.
The es ima ed coe icien s a e posi i e o 𝛽1 and o he 𝛽2 we ha e a posi i e sign in he p o i
ma gin eg ession bu a (low) nega i e esul s in he employmen g ow h. This means ha we ha e an
inc ease in he p o i ma gin and employmen g ow h when he i ms a e no zombie. Also, he
in e ac ion be ween he nonzombie i ms and he sha e o indus y capi al sunk in zombies’ i ms,
educes he p o i ma gin bu aise he employmen g ow h sligh ly.
We can also see he impac o he dimension o he i m. Acco ding o he compu a ions, he
i ms wi h high numbe o employees,
ce e is pa ibus
, ansla es in o highe p o i ma gin and highe
employmen g ow h. To no ice ha i ms wi h ewe han 20 employees ha e nega i e signs and
he e o e ha e a nega i e impac on p o i . The emaining i ms ha e posi i e alues, no ably la ge
companies (wi h mo e han 250 employees) a e hose wi h a highe coe icien alue.
A e ha i is impo an o pay a en ion o he homoscedas ici y. I is a e m o designa ing
cons an a iance o he e o s e ms o dis inc obse a ions. I he homoscedas ici y assump ion is no
alid, we can lis some e ec s on he model:
1- The s anda d e o s o he es ima o s ob ained by OLS a e inco ec and he e o e he
23
s a is ical in e ence is no alid;
2- We can no longe say ha OLS is he bes es ima o s o β, al hough hey may s ill be non-
bias.
Doing he Whi e es o de ec he e oscedas ici y (absence o homoscedas ici y), we can see ha
he null hypo hesis o cons an a iance can be ejec ed a 5% le el o signi icance. The implica ion o
he abo e inding is ha he e is he e oscedas ici y in he esiduals. This can be due o measu emen
e o , model misspeci ica ions o subpopula ion di e ences. Consequences o he he e oscedas ici y a e
ha he OLS es ima es a e no longe BLUE (Bes Linea Unbiased Es ima o ). S anda d e o s will be
un eliable, which will u he cause bias in es esul s and con idence in e als. To sol e his issue he
eg ession was co ec ed by using he obus command.
Table 2 - Zombie i ms and non-zombies’ pe o mance
OLS
FE
RE
OLS
FE
RE
VARIABLES
P o i
Ma gin
P o i
Ma gin
P o i
Ma gin
Ln(emp)
Ln(emp)
Ln(emp)
Non-zombie
30.30***
20.01***
23.33***
0.199***
-0.018***
0.002***
(0.065)
(0.081)
(0.072)
(0.002)
(0.001)
(0.001)
Non-zombie * Zombie
Sha e
0.0246***
-0.105***
-0.050***
-0.002***
0.0001**
-0.0001**
(0.002)
(0.003)
(0.003)
(7.40e-05)
(5.55e-05)
(5.57e-05)
Ope a ing Re enue
0.01***
0.151***
0.02***
0.004***
0.009***
0.007***
(0.002)
(0.02)
(0.006)
(0.0008)
(0.0004)
(0.0002)
Employees 10<X<=20
-0.623***
0.274**
-0.0269
1.607***
0.594***
0.787***
(0.065)
(0.111)
(0.086)
(0.002)
(0.002)
(0.002)
Employees 20<X<=50
-0.267***
0.701***
0.264**
2.340***
1.127***
1.483***
(0.075)
(0.172)
(0.113)
(0.002)
(0.003)
(0.003)
Employees 50<X<=100
0.108
1.003***
0.529***
3.177***
1.671***
2.198***
(0.134)
(0.286)
(0.191)
(0.005)
(0.005)
(0.005)
24
Employees
100<X<=250
0.987***
1.822***
1.364***
3.948***
2.241***
2.942***
(0.181)
(0.433)
(0.273)
(0.007)
(0.008)
(0.007)
Employees 250<X
1.372***
1.559**
1.351***
5.276***
2.913***
3.846***
(0.267)
(0.716)
(0.437)
(0.009)
(0.013)
(0.012)
Obse a ions
899,937
899,937
899,937
929,326
929,326
929,326
R-squa ed
0.217
0.081
0.692
0.194
Numbe o Fi ms
138,923
138,923
140,321
140,321
No es: Non-zombie deno es i ms in Po ugal ha a e no zombies, wi h mo e han 10 yea s and an in e es co e age a io
highe han 1 o 3 consecu i e yea s. Zombie sha e deno es he capi al sunk in zombie i ms. Ln(emp) means he Log o he
employmen . Ope a ing Re enue is ep esen ed in millions. This panel is om da a collec ed om 2010 un il 2018. S anda d
e o s in pa en heses. *** deno es s a is ical signi icance a he 1% le el, ** signi icance a he 5% le el, * signi icance a he
10% le el.
Sou ce: Au ho ’s own calcula ions.
As s a ed be o e, wo models will be used o deeply analyze his s udy. Fi s ly, i was applied in
S a a he Fixed e ec s model o he p o i ma gin a iable o check he impac o he independen
a iables ha can a y o e ime. Secondly, a Random e ec s model was conduc ed using andom
alues. The FE old us ha , by obse ing he p- alue which is he p obabili y ha he null hypo hesis o
he ull model is ue, and since is lowe han 0.05 (in his case is app oxima ely 0) we ha e a leas
some o he pa ame e s being nonze o. The - alues es he hypo hesis ha each coe icien is di e en
om 0 and in his case, we ejec he - alue o e e y a iable because i is highe han 1.96 ( o a 95%
o con idence), so he a iables ha e a signi ican in luence on p o i ma gin. Also, he " ho" alue
(0.53685) ell us ha 53.7% o he a iance is due o di e ences ac oss panels. The ho is also known
as he in e class co ela ion and ells how s ongly he obse a ions wi hin each esemble each o he .
Ano he ele an alue is he coe icien o he eg esso s ega ding he size o he i m. Fo example,
p o i ma gin aises 2 pe cen ual poin s when he i ms wi h mo e han 250 employees inc eases by one
uni .
In bo h, p o i ma gin and employmen g ow h eg essions we ha e a -s a is ic in whe e he null
hypo hesis is ha all he coe icien s in he model excep o he in e cep a e ze o, so:
25
𝐻0:𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑖𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡𝑠 𝑛𝑜𝑡 𝑠𝑦𝑠𝑡𝑒𝑚𝑎𝑡𝑖𝑐 s. 𝐻1:𝐻𝑜 𝑖𝑠 𝑛𝑜𝑡 𝑡𝑟𝑢𝑒 (9)
Since he alue is 0, we ejec he null hypo hesis ( ejec 𝐻0), so his means ha we ha e, a
leas , one 𝛽 di e en om 0 and all independen a iables ( oge he ) a e s a is ically signi ican
Meanwhile he RE says ha he p- alue is, again, app oxima ely 0. The wo- ail p- alues a e all 0
and his es s he hypo hesis ha each coe icien is di e en om 0, so i can be said ha all he
a iables ha e a signi ican in luence on p o i ma gin. The " ho" in his case is 0.429, 10.8 pe cen age
poin s lowe han he FE.
Using he Hausman es , we can decide he e o e which model is mos sui able o his analysis.
Tha said, we ha e:
𝐻0: Random e ec s model is p e e ed
𝐻1: Fixed e ec s model is p e e ed
Tha said and obse ing ou esul s we ha e a p- alue o , app oxima ely, 0, so we ejec he 𝐻0
hypo hesis.
Wi h hese esul s, ou eg essions indica e ha he p esen o hese zombies’ i ms, in he
Po uguese economy, may ha e ampli ied he nega i e consequences o he 2008 c isis bu may also
slow down he economy eco e y by dis o ing he money ha could wen o heal hy i ms.
7.4. Mo e es ic i e analysis
A e s udying he numbe o zombies in Po ugal and doing a es ic i e eg ession, i is
impo an o see he impac o he inancial a ios when we conside a i m o be zombie. Using he
c i e ia desc ibed ea lie in he sec ion 4, a i m needs o ill all he c i e ia o become a zombie so, in
his poin we will ha e a smalle numbe o zombies, in a e age, 6.6% o o al i ms in Po ugal.
Also, i is impo an o no ice (see Table 3) ha , as he less es ic i e analysis, he numbe
dec eased om 2010, wi h 7.4% o he Po uguese being zombies o 5.5% in 2018, bu i aises om
2011 o 2013. This canno be explained p ope ly bu again, he ac ha we ha e a lo o missing da a,
mainly in he oldes yea s, can in luence he da a.
26
Table 3 - Numbe o zombies in Po ugal ( es ic i e)
Yea
Zombies
2010
7.4%
2011
6.8%
2012
7.8%
2013
8.2%
2014
7.3%
2015
6.5%
2016
5.8%
2017
5.6%
2018
5.5%
Sou ce: Au ho ’s own calcula ions.
In he eg essions (see Table 4) we can also no ice almos he same di e ence in he numbe o
obse a ions, om 899,937 o he p o i ma gin a iable and 929,326 o he employmen g ow h.
He eupon, obse ing he R squa ed, 11.5% o he p o i ma gin can be explained by he independen
a iable in his model and o he employmen g ow h we ha e a highe R squa ed jus like be o e, which
is be e , because his eg ession can explain 65.7% on he employmen g ow h. The R2 is almos he
same using he mo e es ic i e analysis.
Table 4 - Zombies i ms and non-zombies’ pe o mance
OLS
FE
RE
OLS
FE
RE
VARIABLES
Ln(emp)
Ln(emp)
Ln(emp)
P o i Ma gin
P o i Ma gin
P o i
Ma gin
Non-zombie
0.217***
-0.011***
0.005***
27.15***
13.28***
16.89***
27
(0.003)
(0.002)
(0.002)
(0.091)
(0.098)
(0.091)
Non-zombie *
Zombie Sha e
-0.001***
-7.78e-05
-0.0002***
0.193***
0.112***
0.149***
(7.16e-05)
(5.17e-05)
(5.21e-05)
(0.002)
(0.003)
(0.003)
Ope a ing
Re enue
0.004***
0.009***
0.007***
0.015***
0.191***
0.026***
(0.0008)
(0.0004)
(0.0002)
(0.003)
(0.020)
(0.006)
Employees
10<X<=20
1.612***
0.594***
0.787***
0.201***
0.328***
0.364***
(0.002)
(0.002)
(0.002)
(0.069)
(0.114)
(0.090)
Employees
20<X<=50
2.345***
1.127***
1.482***
0.624***
0.940***
0.901***
(0.003)
(0.003)
(0.003)
(0.079)
(0.176)
(0.120)
Employees
50<X<=100
3.181***
1.670***
2.198***
0.759***
1.320***
1.096***
(0.005)
(0.005)
(0.005)
(0.142)
(0.294)
(0.201)
Employees
100<X<=250
3.950***
2.241***
2.941***
1.481***
2.471***
1.998***
(0.007)
(0.008)
(0.007)
(0.192)
(0.444)
(0.289)
Employees
250<X
5.277***
2.912***
3.845***
1.570***
2.609***
1.989***
(0.010)
(0.013)
(0.012)
(0.284)
(0.735)
(0.464)
Obse a ions
929,326
929,326
929,326
899,937
899,937
899,937
R-squa ed
0.691
0.193
0.115
0.031
Numbe o
Fi ms
140,321
140,321
138,923
138,923
No es: Non-zombie deno es i ms in Po ugal ha a e no zombies, wi h mo e han 10 yea s, an in e es co e age a io highe
han 1 o 3 consecu i e yea s, ROA below 0 and o al deb a io below 40%. Zombie sha e deno es he capi al sunk in
zombie i ms. Ln(emp) means he Log o he employmen . Ope a ing Re enue is ep esen ed in millions. This panel is om
da a collec ed om 2010 un il 2018. S anda d e o s in pa en heses. *** deno es s a is ical signi icance a he 1% le el, **
signi icance a he 5% le el, * signi icance a he 10% le el.
Sou ce: Au ho ’s own calcula ions.
28
In his model we ha e go he same p- alues (app oxima ely ze o o all a iables) as be o e so
ou sample da a p o ide enough e idence o ejec he null hypo hesis o he en i e se . The da a a o
he hypo hesis ha he e is a non-ze o co ela ion. The be a (𝛽1) in his model has he same signal as
he es ic i e model, being posi i e o bo h dependen a iables. So e en wi h he addi ional
es ic ions, we ha e a posi i e impac o he non-zombie i ms on he p o i ma gin and he employmen
g ow h. The sha e o indus ial capi al sunk in zombie i ms wi h he in e ac ion wi h he non-zombies is
almos he same as he p e ious one, being nega i e o he employmen and posi i e o he p o i
ma gin. This means ha , ha ing mo e capi al sunk in zombie i ms can con ibu e sligh ly o a highe
p o i ma gin o non-zombie i ms. This was no expec ed bu i can be explained because i zombie
i ms can ge subsidized c edi .
In bo h, p o i ma gin and employmen g ow h eg essions we ha e a -s a is ic in whe e he null
hypo hesis is ha all he coe icien s in he model excep o he in e cep a e ze o, so:
𝐻0:𝛽1=0,𝛽2=0,𝛽3=0,𝛽4=0 s. 𝐻1:𝐻𝑜 𝑖𝑠 𝑛𝑜𝑡 𝑡𝑟𝑢𝑒 (10)
Since he alue is 0, we ejec he null hypo hesis ( ejec 𝐻0), so his means ha we ha e, a
leas , one 𝛽 di e en om 0 and all independen a iables a e s a is ically signi ican .
The numbe o employees, in his eg ession, ell us he same conclusion as he p e ious one.
So, a aise in he numbe o employees ha each i m has, con ibu es posi i ely o a aise in ei he he
p o i ma gin o he employmen g ow h
To hese eg essions i is, as well, applied he Fixed e ec s model and he Random e ec s
model. To he p o i ma gin, he FE model p o ides almos he same esul s as he less es ic i e
eg essions p o ided wi h a p- alue being app oxima ely 0. To no ice ha he wo- ail p- alues a y in he
size o he i m, mainly when i ms ha e mo e han 250 o be ween 10 and 20 employees, bu s ill less
han 0.05 so we can say ha i is a signi ican in luence on p o i ma gin. The " ho" is 0.557 and
ano he poin is ha ou non-zombie a iable wi h es ic i e inancial a ios has a - alue o 139 ( he
highe he - alue, he highe he ele ance o he a iable).
Again, using he RE model he p- alue is app oxima ely 0 as well as all wo- ail p- alues on all
a iables. Obse ing he alue on he Hausman es , we again should ejec he 𝐻0.
To he employmen g ow h, FE model show us he same esul s as he p e ious models wi h
29
app oxima ely 0 o he p- alue and he wo- ail p- alues bu a " ho" o 0.902. The di e ence a e he
alues o he i m size ega ding he - alues wi h a ound 300 in each ca ego y o he numbe o he
employees ( he highes is wi hin he g oup o mo e han 20 bu less han 50), so i can be said ha his
a iable has a signi ican in luence on he employmen g ow h.
The RE model on he employmen , he impo an hing o highligh is he ac ha ou non-
zombie a iable and he in e ac ion be ween he non-zombie and he capi al sunk in he zombie i ms
ha e di e en alues o he wo- ail p- alues wi h 0.01 and 0.03 espec i ely. Rega ding he
employmen g ow h hese wo a iables ha e a signi ican impac .
Using he Hausman es , p e ious shown, o he employmen g ow h, we should ejec he 𝐻0
hypo hesis since he p- alue o his es is, app oxima ely, 0.
7.5. Discussion o esul s
Looking a he eg essions and he esul s ob ained, he eg essions, ei he he less es ic i e o
he mo e es ic i e, ha e almos he same impac in he p o i ma gin and employmen g ow h
Po uguese i ms.
E en wi h a smalle R squa e, he ixed e ec s model should be conside ed as he mos sui able
o his eg ession. Looking a his model we can conclude ha , o he es ic i e analysis, o example,
p o i inc eases by 0.2 pe cen age poin s when a i m is non-zombie ins ead o a zombie, ha means
ha , non-zombie i ms ha e highe p o i s han zombies (as expec ed). In his eg ession, he log o he
employmen is nega i e and sugges s ha employmen g ow h ends o slow i he i m is non-zombie.
Also, he ope a ing e enue is posi i e o bo h dependen a iables which means ha , an inc ease in
hese a iables will inc ease he ope a ing e enue o he i m.
Fo he mo e es ic i e analyzes, ou a iable o non-zombie dec eases ou alues o he
es ima o . So, he p o i ma gin inc eases by 0.27 pe cen age poin s when i is a non-zombie i m, his
means ha a non-zombie should epo highe p o i ma gin ha a zombie i m. Again, he log o he
employmen sugges s ha he employmen g ow h ends o slow i he i m is non-zombie (nega i e
sign). The ope a ing e enue in his eg ession is also posi i e o bo h a iables.
30
Ou eg essions showed ha , i he i m has, o example, mo e employees, he p o i ma gin
and he employmen g ow h end o inc ease,
ce e is pa ibus
. Also, ou a iable o he non-zombie is
always posi i e indica ing ha hey ha e a posi i e impac on he wo dependen a iables and he
in e cep ion be ween hese and he capi al sunk in zombie i ms p o ide i al in o ma ion ha zombie
i ms can co up he ma ke .
The li e a u e old us ha zombie i ms, in a e age, a e olde and less p oduc i e and wi h his
da abase we can ob ain almos he same esul s (Gou eia and Coelho, 2017). O he au ho s such as
Os e hold and Gou eia (2018), p esen ed a s udy whe e hey show ha zombies, in Po ugal, a e
dec easing as almos as he esul s p esen ed, and he educ ion in he capi al sunk in zombie i ms
e ches posi i e ex e nali ies o non-zombies.
O he pape s like McGowan e al. (2018), ha do no alk di ec ly abou he Po uguese
economy, ind ha an inc ease in zombie sha e a he indus y le el is associa ed wi h lowe in es men
and employmen g ow h o he a e age non-zombie and acco ding o bo h eg essions his is ue o
Po ugal. Also, Ba ne e al. (2014) p o ide a ele an s udy abou he employmen beha io ha allows
us o see ha zombie i ms end o be mo e size dependen .
Ba os e al. (2017) also analyzed he zombies using es ic ions on he sec o s. They had he
same esul s p esen ed on he poin 7.1., whe e zombies end o be om eal es a e, accommoda ion
and ood se ices and cons uc ion. Rega ding Po ugal we ha e o he pape s ha alk abou whe e he
zombies end o concen a e in e ms o zones and, as we could saw he No h o Po ugal ha e mo e
i ms bu less zombies ega ding, o example, he Alga e o he au onomous egions o Madei a and
Azo es.
Since he da abase used was no he bes bu he esul s we e almos he same as p e ious and
ele an pape s (wi h some di e ences), hese eg essions can be accep ed and may help in o de o
analyze mo e deeply zombie i ms. O he impo an aspec o no ice is ha he inancial a ios may ha e
a huge ole when ying o de ine and desc ibe zombie i ms.
37
PT11B - Al o Tamega
10,142
0.58
29.41
PT11C - Tamega e Sousa
60,301
3.48
32.88
PT11D - Dou o
24,773
1.43
34.31
PT11E - Te as de T as-os-Mon es
14,775
0.85
35.16
PT150 - Alga e
80,746
4.65
39.82
PT16B - Oes e
60,870
3.51
43.33
PT16D - Regiao de A ei o
59,985
3.46
46.78
PT16E - Regiao de Coimb a
68,685
3.96
50.74
PT16F - Regiao de Lei ia
65,359
3.77
54.51
PT16G - Viseu Dao La oes
38,445
2.22
56.73
PT16H - Bei a Baixa
11,661
0.67
57.40
PT16I - Medio Tejo
35,550
2.05
59.45
PT16J - Bei as e Se a da Es ela
30,290
1.75
61.19
PT170 - A ea Me opoli ana de Lisboa
508,769
29.33
90.52
PT181 - Alen ejo Li o al
13,387
0.77
91.29
PT184 - Baixo Alen ejo
15,079
0.87
92.16
PT185 - Lezi ia do Tejo
36,112
2.08
94.24
PT186 - Al o Alen ejo
15,045
0.87
95.11
PT187 - Alen ejo Cen al
24,581
1.42
96.53
PT200 - Regiao Au onoma dos Aco es
22,555
1.30
97.83
PT300 - Regiao Au onoma da Madei a
37,708
2.17
100.00
To al
1,734,879
100.00
No es: Fi ms om da abase di ided in
NUTS
3 p esen ed in he Po uguese economy.
Sou ce: Au ho ’s own calcula ions acco ding o Amadeus da abase.
Table 8 - NACE code and desc ip ion
1
A
AGRICULTURE, FORESTRY AND FISHING
2
B
MINING AND QUARRYING
3
C
MANUFACTURING
4
D
ELECTRICITY, GAS, STEAM AND AIR CONDITIONING SUPPLY
5
E
WATER SUPPLY; SEWERAGE, WASTE MANAGEMENT AND REMEDIATION ACTIVITIES
6
F
CONSTRUCTION
7
G
WHOLESALE AND RETAIL TRADE; REPAIR OF MOTOR VEHICLES AND MOTORCYCLES
8
H
TRANSPORTATION AND STORAGE
9
I
ACCOMMODATION AND FOOD SERVICE ACTIVITIES
10
J
INFORMATION AND COMMUNICATION
11
K
FINANCIAL AND INSURANCE ACTIVITIES
12
L
REAL ESTATE ACTIVITIES
13
M
PROFESSIONAL, SCIENTIFIC AND TECHNICAL ACTIVITIES
14
N
ADMINISTRATIVE AND SUPPORT SERVICE ACTIVITIES
15
O
PUBLIC ADMINISTRATION AND DEFENCE; COMPULSORY SOCIAL SECURITY
38
16
P
EDUCATION
17
Q
HUMAN HEALTH AND SOCIAL WORK ACTIVITIES
18
R
ARTS, ENTERTAINMENT AND RECREATION
19
S
OTHER SERVICE ACTIVITIES
20
T
ACTIVITIES OF HOUSEHOLDS AS EMPLOYERS; U0NDIFFERENTIATED GOODS- AND
SERVICES-PRODUCING ACTIVITIES OF HOUSEHOLDS FOR OWN USE
21
U
ACTIVITIES OF EXTRATERRITORIAL ORGANISATIONS AND BODIES
No es: NACE dis ibu ion using le e s ( om A o U) o de ine each sec o in he economy
Sou ce: NACE acco ding o Eu os a
Table 9 - NACE dis ibu ion in he Po uguese economy
NACE code
Numbe o Fi ms
Pe cen age
Cumula i e
A
61,901
3.57
3.57
B
3,788
0.22
3.79
C
212,061
12.22
16.01
D
1,900
0.11
16.12
E
4,747
0.27
16.39
F
172,889
9.97
26.36
G
473,301
27.28
53.64
H
96,557
5.57
59.21
I
151,872
8.75
67.96
J
40,006
2.31
70.27
K
26,573
1.53
71.80
L
81,294
4.69
76.48
M
169,302
9.76
86.24
N
53,648
3.09
89.33
O
142
0.01
89.34
P
23,180
1.34
90.68
Q
106,861
6.16
96.84
R
19,465
1.12
97.96
S
35,392
2.04
100.00
To al
1,734,879
100.00
No es: NACE dis ibu ion using he le e s p e iously de ined o see how many i ms he e is on his da abase
Sou ce: Au ho ’s own calcula ions acco ding o Amadeus da abase.
Table 10 - Numbe o employees in each i m
Numbe o employees
F equency
Pe cen age
Cumula i e
X<=10
1,477,335
85.15
85.15
10<X<=20
128,174
7.39
92.54
39
20<X<=50
86,841
5.01
97.55
50<X<=100
24,103
1.39
98.94
100<X<=250
12,559
0.72
99.66
X>250
5,867
0.34
100.00
To al
1,734,876
100.00
No es: Numbe o employees in each i m di ided pe ca ego ies.
Sou ce: Au ho ’s own calcula ions acco ding o he Amadeus da abase
Table 11 - P obi model
Va iables
Pa ame e s
10<X<=20
-0.4609***
(0.0070)
20<X<=50
-0.5175***
(0.0084)
50<X<=100
-0.4239***
(0.0143)
100<X<=250
-0.3910***
(0.0191)
250<X
-0.3621***
(0.0273)
Ag icul u e, o es y and
ishing
-0.1075***
(0.0106)
Mining and qua ying
0.2579***
(0.0285)
Manu ac u ing
-0.0559***
(0.0059)
Elec ici y, gas, s eam and ai
condi ioning supply
-0.3041***
(0.0603)
Wa e supply. sewe age,
was e managemen and
emedia ion ac i i ies
-0.3081***
(0.0445)
Cons uc ion
-0.0785***
(0.0063)
T anspo a ion and s o age
-0.1598***
(0.0073)
Accommoda ion and ood
se ice ac i i ies
0.4826***
(0.0056)
In o ma ion and
communica ion
-0.1754***
(0.0150)
Financial and insu ance
ac i i ies
-0.3356***
(0.0208)
Real es a e ac i i ies
0.1038***
40
(0.0078)
P o essional, scien i ic and
echnical ac i i ies
-0.3580***
(0.0074)
Adminis a i e and suppo
se ice ac i i ies
-0.1899***
(0.0124)
Public adminis a ion and
de ense
0.6649***
(0.1751)
Educa ion
0.2327***
(0.0143)
Human heal h and social
wo k ac i i ies
-0.5441***
(0.0097)
A s, en e ainmen and
ec ea ion
0.1704***
(0.0187)
O he se ice ac i i ies
0.3100***
(0.0113)
No h
-0.1147***
(0.0043)
Alga e
-0.0257***
(0.0081)
Cen e
-0.1534***
(0.0048)
Alen ejo
-0.0927***
(0.0078)
Azo es
-0.0111
(0.0156)
Madei a
0.2194***
(0.0105)
pseudo-R2
0.05
log likelihood
-340,804.38
N
935,458
No es: P obi model using he size o he i m, NACE code and he
NUTS
2 dis ibu ion. N is he numbe o obse a ions. :
Robus s anda d e o s in pa en heses. ***deno es s a is ical signi icance a he 1% le el, ** signi icance a he 5% le el, *
signi icance a he 10% le el.
Sou ce: Au ho ’s own calcula ions acco ding o he Amadeus da abase
Table 12 - Ma ginal e ec s o he P obi model (Table 11)
Va iables
Ma ginal e ec s
X<=10
0.1437***
(0.0004)
10<X<=20
0.0656***
(0.0008)
20<X<=50
0.0589***
(0.0009)
50<X<=100
0.0703***
(0.0018)
41
100<X<=250
0.0747***
(0.0026)
250<X
0.0787***
(0.0039)
Ag icul u e, o es y and ishing
0.1090***
(0.0019)
Mining and qua ying
0.1914***
(0.0076)
Manu ac u ing
0.1188***
(0.0010)
Elec ici y, gas, s eam and ai
condi ioning supply
0.0770***
(0.0086)
Wa e supply. sewe age, was e
managemen and emedia ion ac i i ies
0.0765***
(0.0063)
Cons uc ion
0.1144***
(0.0011)
Wholesale and e ail ade
0.1301***
(0.0006)
T anspo a ion and s o age
0.0997***
(0.0011)
Accommoda ion and ood se ice
ac i i ies
0.2572***
(0.0015)
In o ma ion and communica ion
0.0970***
(0.0025)
Financial and insu ance ac i i ies
0.0727***
(0.0028)
Real es a e ac i i ies
0.1529***
(0.0017)
P o essional, scien i ic and echnical
ac i i ies
0.0697***
(0.0009)
Adminis a i e and suppo se ice
ac i i ies
0.0946***
(0.0020)
Public adminis a ion and de ense
0.3181***
(0.0613)
Educa ion
0.1847***
(0.0037)
Human heal h and social wo k ac i i ies
0.0483***
(0.0009)
A s, en e ainmen and ec ea ion
0.1689***
(0.0046)
O he se ice ac i i ies
0.2057***
(0.0030)
No h
0.1191***
(0.0006)
Alga e
0.1369***
(0.0016)
Cen e
0.1119***
(0.0007)
AML
0.1424***
42
(0.0006)
Alen ejo
0.1234***
(0.0014)
Azo es
0.1400***
(0.0033)
Madei a
0.1951***
(0.0027)
N
935,458
No es: Using he size o he i m, NACE dis ibu ion and
NUTS
2 o compu e he ma ginal e ec s. N is he numbe o
obse a ions.
Sou ce: Au ho ’s own calcula ions acco ding o he Amadeus da abase